Next Article in Journal
Fruiting Characteristics and Molecular-Assisted Identification of Korla Fragrant Pear Bud Mutation Materials
Next Article in Special Issue
Temperature Dynamics in Early Pregnancy: Implications for Improving In Vitro Fertilization Outcomes
Previous Article in Journal
Sparsity-Based Nondestructive Evaluations of Downhole Casings Technique Using the Uniform Linear Array
Previous Article in Special Issue
Efficient Sleep–Wake Cycle Staging via Phase–Amplitude Coupling Pattern Classification
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

Artificial Intelligence in Biomaterials: A Comprehensive Review

1
Department of Computer Engineering, Ankara University, Ankara 06830, Turkey
2
Institute of Nuclear Sciences, Ankara University, Ankara 06830, Turkey
3
Department of Artificial Intelligence and Data Engineering, Ankara University, Ankara 06830, Turkey
4
Fonet Information Technologies, Ankara 06520, Turkey
5
Faculty of Medicine and Health Technology, Tampere University, 33720 Tampere, Finland
6
VTT Technical Research Centre of Finland, 33101 Tampere, Finland
*
Author to whom correspondence should be addressed.
Appl. Sci. 2024, 14(15), 6590; https://doi.org/10.3390/app14156590
Submission received: 14 June 2024 / Revised: 22 July 2024 / Accepted: 26 July 2024 / Published: 28 July 2024
(This article belongs to the Special Issue Machine Learning in Biomedical Applications)

Abstract

:
The importance of biomaterials lies in their fundamental roles in medical applications such as tissue engineering, drug delivery, implantable devices, and radiological phantoms, with their interactions with biological systems being critically important. In recent years, advancements in deep learning (DL), artificial intelligence (AI), machine learning (ML), supervised learning (SL), unsupervised learning (UL), and reinforcement learning (RL) have significantly transformed the field of biomaterials. These technologies have introduced new possibilities for the design, optimization, and predictive modeling of biomaterials. This review explores the applications of DL and AI in biomaterial development, emphasizing their roles in optimizing material properties, advancing innovative design processes, and accurately predicting material behaviors. We examine the integration of DL in enhancing the performance and functional attributes of biomaterials, explore AI-driven methodologies for the creation of novel biomaterials, and assess the capabilities of ML in predicting biomaterial responses to various environmental stimuli. Our aim is to elucidate the pivotal contributions of DL, AI, and ML to biomaterials science and their potential to drive the innovation and development of superior biomaterials. It is suggested that future research should further deepen these technologies’ contributions to biomaterials science and explore new application areas.

1. Introduction

AI and ML has become a powerful tool in various scientific fields, and its applications have permeated different industries. In recent years, the biomaterial field has also harnessed the potential of machine learning to advance research and development [1,2]. This review article aims to explore the applications of ML in the biomaterial field, focusing on its impact, challenges, and future prospects. Biomaterials are essential in medicine, facilitating the creation of innovative medical devices, tissue engineering frameworks, and drug delivery mechanisms. The intricate relationship between biomaterials and living tissues requires a deep understanding of their properties, interactions, and performance. ML provides an avenue to enhance this understanding and improve the design and optimization of biomaterials [3].
One key application of ML in the biomaterial field is predictive modeling. By training algorithms on extensive datasets, researchers can develop models capable of predicting the behavior or characteristics of biomaterials. For example, ML algorithms can be trained to predict the biocompatibility of a material or forecast its degradation rate. These predictive models aid in the selection of biomaterials for specific applications, reducing the time and cost associated with trial-and-error experiments [4,5,6,7]. ML has also made notable advancements in the design and synthesis of biomaterials [8]. Traditional trial-and-error methods can be time-consuming and resource intensive. ML algorithms can expedite this process by analyzing large databases of material properties, identifying correlations, and generating new material designs with tailored properties [9]. This ability to generate novel biomaterials opens avenues for innovation and accelerates the discovery of materials with enhanced biocompatibility and performance [10]. ML also plays a critical role in image analysis and interpretation in the biomaterial field. Medical imaging techniques produce vast amounts of complex data, which can be challenging to interpret accurately. By utilizing ML algorithms, researchers can analyze and classify images, aiding in the diagnosis, monitoring, and treatment of diseases. Additionally, ML algorithms can be applied to image-based analysis of biomaterials, enabling rapid evaluation of material performance and integrity [11]. ML offers significant benefits in biomaterials research, yet it faces challenges such as limited data availability and quality. Developing robust ML models requires standardized, large-scale datasets, and expertise in model interpretation and validation due to their complexity. Ethical considerations, including data privacy and bias, are critical in healthcare applications. Looking forward, ML’s future in biomaterials research promises advancements as datasets improve, enhancing model accuracy and reliability. Moreover, integrating ML with emerging technologies like nanotechnology and bioinformatics will synergistically advance biomaterials research [12,13].
In the upcoming sections, we explore quick explanations of key AI and ML subfields, such as SL, US, RL, and DL. Following this, we delve into various ML methods commonly employed in biomaterials research.
  • Artificial Intelligence and Machine Learning Methods
AI involves the emulation of human intellect in machines designed to replicate human thinking and actions. These advanced systems can carry out tasks that usually need human intelligence, like visual interpretation, speech understanding, decision making, and translating languages. AI includes various subfields such as machine learning, natural language processing, robotics, and computer vision, among others [14]. Figure 1 illustrates the hierarchical relationship between AI, ML, and DL. AI includes a wide array of technologies aimed at replicating human intelligence, such as ML, which uses algorithms to allow computers to learn from data and make predictions. A specific branch of ML and DL utilizes multi-layered neural networks to examine different aspects of data.
ML focuses on enabling computers to learn from data and identify meaningful patterns without being explicitly programmed for each task [15]. Recent advances in ML have driven the development of intelligent systems that exhibit human-like cognitive abilities, impacting various aspects of business and personal life. These systems play a significant role in electronic markets by enhancing decision-making processes for productivity, engagement, and employee retention adapting to user preferences through trainable assistant systems and transforming traditional financial markets with sophisticated trading agents [16].
The advanced problem-solving capabilities of these systems, broadly termed as AI, are grounded in analytical models that generate predictions, recommendations, and other valuable outputs. Initially, building such models involved manually programming known relationships and decision rules into systems. However, with the advent of more powerful programming frameworks, increased data availability, and accessible computing resources, the development of analytical models has increasingly relied on ML techniques. ML reduces the need for humans to explicitly encode their knowledge, allowing for more efficient development of intelligent systems [17].
Over recent decades, the field of ML has seen significant progress, particularly with the evolution of artificial neural networks (ANNs) into deep neural network architectures, commonly known as DL [18,19]. These advancements have enabled DL models to outperform humans in specific, controlled environments [20]. Despite their potential, deploying these models in real-world business contexts presents challenges, such as selecting the appropriate implementation options, addressing biases and data drift, mitigating the “black-box” nature of models, and reusing preconfigured models as services.

1.1. Supervised Learning (SL)

SL is a fundamental paradigm in ML, where the objective is to predict the label or value of new instances based on a training dataset with known labels. The central idea is to learn a mapping from input features to output labels, utilizing a labeled dataset to train the model, as shown in Figure 2. This process involves the development of an inference function that can generalize from the training data to make accurate predictions on unseen data. SL encompasses two main types of problems: classification and regression.
  • Classification: When the output variable is categorical, the problem is known as classification. The goal is to assign an input to one of several predefined classes. This can involve binary classification, where there are only two classes, or multiclass classification, where there are more than two classes. To address these issues, commonly employed classification algorithms include logistic regression, support vector machines, decision trees, and neural networks. Classification model performance is frequently assessed using metrics such as accuracy, precision, recall, and the F1 score [21].
  • Regression: When the output variable is continuous, the problem is referred to as regression. The objective in this context is to forecast a continuous value from the input features. Frequently used regression algorithms include linear regression, ridge regression, lasso regression, least absolute deviations (LAD), partial least squares (PLS), and more sophisticated models like random forests and gradient boosting machines. The effectiveness of regression models is generally evaluated using metrics like mean squared error (MSE), mean absolute error (MAE), and R-squared (R2) [22].
In supervised learning, the inference function is obtained by minimizing a loss function that quantifies the difference between predicted outputs and actual labels in the training set. This loss function is adapted to the specific problem type. For classification tasks, commonly used loss functions are cross-entropy loss and hinge loss, while for regression tasks, MSE is often employed. The choice of loss function influences the optimization process and the resulting model’s performance [23].
The training process consists of modifying the model parameters to reduce the loss function on the training data. This is generally accomplished using optimization algorithms like gradient descent and its variations (e.g., stochastic gradient descent, Adam). The optimization method iteratively adjusts the model parameters in a direction that decreases the loss, enabling the model to capture the data’s underlying patterns [18].
A critical challenge in supervised learning is ensuring that the model generalizes well to unseen data. Overfitting occurs when a model learns the noise and details in the training data to the extent that it performs poorly on new data. Techniques such as cross-validation, regularization (e.g., L1 and L2 regularization), and early stopping are employed to mitigate overfitting and enhance generalization [15].
This ML method is a widely used approach in machine learning, enabling the prediction of labels or values for new instances based on historical data. By developing robust inference functions, carefully selecting algorithms, and conducting rigorous evaluations, supervised learning models can achieve high accuracy and generalization, making them indispensable tools across various fields.

1.2. Unsupervised Learning (UL)

UL is a branch of ML employed when the data under analysis are unlabeled, meaning that the classes or target values are unknown. In unsupervised learning, only the input data are provided during the training phase, and the goal is to identify hidden patterns or structures within these data. This method seeks to uncover intrinsic relationships within the data without the guidance of pre-existing labels shown in Figure 3.
  • Clustering: Clustering is one of the primary techniques in UL, where the algorithm groups similar data points into clusters. The goal is to ensure that data points within the same cluster are more alike to each other than to those in different clusters. This technique is widely used in market segmentation, social network analysis, and bioinformatics [24]. Common clustering algorithms include:
    • K-means clustering: Partitions the data into k clusters by minimizing the variance within each cluster [25].
    • Hierarchical clustering: Builds a tree of clusters through either a bottom-up (agglomerative) or top-down (divisive) approach [26].
  • Dimensionality reduction: This method decreases the number of random variables by identifying a set of principal variables. Dimensionality reduction is essential for managing high-dimensional data, enhancing computational efficiency, and boosting model performance.
    Techniques include:
    • Principal component analysis (PCA): Identifies the directions (principal components) in which the data varies the most [27].
    • t-Distributed stochastic neighbor embedding (t-SNE): A non-linear method for dimensionality reduction that is especially useful for visualizing high-dimensional datasets [28].
  • Association rules: This method is used to find interesting relationships or associations among variables in large databases. Association rules are widely used in market basket analysis to identify items that frequently co-occur in transactions [29].
One of the primary challenges in unsupervised learning is the lack of ground truth, making it difficult to evaluate the model’s performance objectively. Unlike supervised learning, where accuracy can be directly measured against known labels, unsupervised learning often relies on indirect metrics such as cluster cohesion and separation or the interpretability of reduced dimensions [30]. Another consideration is the choice of the algorithm and the number of clusters in clustering algorithms, which often requires domain knowledge and experimentation to determine the most meaningful grouping [31].
UL is a powerful tool for discovering hidden patterns and structures within unlabeled data. By employing techniques such as clustering, dimensionality reduction, and association rule learning, unsupervised learning algorithms can reveal valuable insights and drive decision making in various domains. Despite the challenges associated with evaluation and algorithm selection, unsupervised learning remains a crucial component of the machine learning landscape.

1.3. Reinforcement Learning (RL)

RL is a learning paradigm focused on controlling a system to optimize a numerical performance metric, which represents a long-term goal. Unlike supervised learning, RL provides partial feedback about actions during training, and these actions can have long-term impacts on the system’s future states. In RL, an agent interacts with an environment by observing its current state, performing actions, and receiving rewards based on those actions. The goal is to learn a policy—a mapping from states to actions—that maximizes the total cumulative reward over time [32]. This policy dictates the agent’s behavior by determining which action to take in a given state. Value functions, such as the state-value function (V) and the action-value function (Q), represent the expected return or cumulative reward for a specific policy. These functions help the agent make decisions to optimize long-term rewards [32].
RL agents encounter the exploration–exploitation dilemma, where they must balance exploring new actions to find better strategies and exploiting known actions to maximize immediate rewards. Techniques such as epsilon-greedy policies and upper confidence bound (UCB) methods address this trade-off [32]. During training, the agent engages with the environment through trial and error, adjusting its policy based on received rewards to optimize cumulative rewards over time. This training process often utilizes algorithms like Q-learning, SARSA, and deep Q-networks (DQN), which update the agent’s value estimates based on observed rewards and state transitions [33].
RL presents several challenges, including the need for efficient exploration strategies, dealing with delayed rewards, and addressing the curse of dimensionality in high-dimensional state and action spaces. Additionally, ensuring the safety and ethical behavior of reinforcement learning agents in real-world applications remains a significant concern [33].
Reinforcement learning presents an effective framework for training agents to interact with dynamic environments and optimize long-term performance objectives. By learning from feedback and exploring various strategies, reinforcement learning agents can autonomously acquire complex behaviors and adapt to diverse tasks in real-world settings.

1.4. Deep Learning (DL)

DL, a subset of machine learning, specializes in training artificial neural networks with multiple layers to learn data representations at various levels of abstraction. These networks, called deep neural networks (DNNs), have achieved significant success in tasks such as image recognition, natural language processing, and speech recognition. Unlike traditional ML approaches, DL does not require feature extraction before training.
DL models are inspired by the structure and function of the human brain. They are composed of interconnected layers of artificial neurons, where each neuron computes a weighted sum of its inputs followed by a non-linear activation function, as shown in Figure 4. Through forward propagation, input data pass through multiple layers of neurons, eventually producing an output [18]. DL architectures often include multiple hidden layers, enabling the network to learn hierarchical representations of the input data. These architectures encompass convolutional neural networks (CNNs) for image processing, recurrent neural networks (RNNs) for sequential data such as time series analysis using long short-term memory (LSTM) and bidirectional LSTM (BiLSTM), and transformer models for natural language processing [19].
DL models are trained using large datasets through a process called backpropagation, where the error between the predicted output and the ground truth is propagated backward through the network to update the model’s parameters. Training deep networks often requires powerful computational resources, and techniques like mini-batch gradient descent and adaptive optimization algorithms are commonly employed to accelerate convergence [34]. Despite its successes, deep learning faces several challenges. DL models often require large amounts of labeled data to achieve optimal performance. Techniques for training deep models with limited labeled data, such as transfer learning and semi-supervised learning, are areas of active research [35]. DNN are often considered black-box models, making it challenging to interpret their decisions. Addressing the interpretability of deep learning models is crucial, especially in applications where transparency and accountability are essential. Training deep networks can be computationally intensive, requiring specialized hardware like GPUs and TPUs. Developing efficient algorithms and hardware architectures for deep learning remains a focus of research and development [19].
DL, similar to ML, encompasses various learning paradigms categorized as supervised, UL, and RL, as illustrated in Figure 5. In SL, models are trained on labeled datasets, allowing them to learn the mapping from inputs to desired outputs, which is essential for tasks like classification and regression. UL involves training models on unlabeled data to uncover hidden patterns or intrinsic structures, often used in clustering and dimensionality reduction. Reinforcement learning trains models to make sequences of decisions by rewarding desired outcomes, making it useful for applications like robotics and game playing [18].
Deep learning has emerged as an effective approach for learning representations from complex data and solving a wide range of tasks across various domains. With ongoing research efforts aimed at addressing its challenges and expanding its capabilities, deep learning continues to drive innovation and advance the state-of-the-art in artificial intelligence.

2. Artificial Intelligence and Machine Learning in the Biomaterial Field

The study of advanced materials such as polymers, metals, composites, and inorganics requires a multifaceted approach that leverages modern computational techniques and experimental validations. This methodology section outlines the integrated use of ML, high-throughput simulations, and experimental methods to predict, optimize, and design these materials for various applications.
Machine learning stands out as a transformative tool, capable of dissecting vast datasets, unraveling intricate patterns, and making precise prognoses regarding material properties. Within the domain of polymers, ML models prognosticate diverse properties, spanning from gene delivery efficacy to cellular responses. Techniques like artificial neural networks and genetic algorithms fine-tune structure–property relationships, fostering the development of avant-garde polymeric materials.
In the realm of metals, ML techniques forecast critical properties such as glass-forming ability and mechanical behavior. Various models, including random forest classifiers and CNNs, navigate extensive datasets to craft predictive models, guiding the creation of bespoke metallic alloys and compounds. Composites undergo scrutiny through a fusion of ML and high-fidelity simulations like finite element analysis, effectively modeling elastic and mechanical attributes [36,37,38]. Deep learning paradigms, notably three-dimensional CNNs, showcase substantial progress in predicting the performance of high-contrast composites, expediting the design of bioinspired materials tailored for specific applications [39]. Inorganic materials similarly benefit from an integrative approach, combining high-throughput molecular dynamics simulations with ML to prognosticate properties such as Young’s modulus and glass-forming ability [40]. Ensemble ML models and Gaussian process regression prove adept at handling sparse datasets, furnishing dependable predictions and facilitating the design of biomedical glasses optimized for healthcare exigencies [41].
The search strategies applied are given in Table 1. In order to include similar studies, the words and word groups “all: Artificial Intelligence (AI)”, “Machine Learning (ML)”, “biomaterials”, “Deep Learning (DL)”, “polymers”, “metals”, “composites”, and “inorganic materials” were also included for detailed search. Since the desired results could not be achieved with the search method, starting with the most recent publications, the search method was expanded and applied to all years.
This methodology harmonizes computational prowess with empirical validation, providing a holistic framework for material exploration and optimization. Utilizing the synergies of ML and high-throughput simulations expedites the discovery of materials tailored to meet diverse industrial and scientific requisites.
The integration of AI and ML techniques revolutionizes material research paradigms, augmenting our capacity to predict, design, and optimize materials. This section explores numerous studies employing AI and ML techniques to push the boundaries of biomaterial science, covering polymers, metals, composites, and inorganic materials, as depicted in Figure 6.

2.1. Metal Materials

The use of metals in various applications, including biomaterials, has seen significant advancements, particularly with the integration of ML and data mining techniques in recent years. Predictive modeling in corrosion resistance and mechanical properties, such as tensile strength and ductility, has greatly improved the reliability and performance of metal components. Material design and discovery have been enhanced through ML in alloy design and additive manufacturing, optimizing parameters for metal 3D printing processes. Process optimization has seen significant improvements, with ML models predicting outcomes of heat treatment and refining casting and forging processes by minimizing defects. AI also plays a crucial role in failure analysis and prevention, predicting fatigue and fracture points, and modeling crack propagation to aid in maintenance and design improvements. Surface engineering has benefited from ML in designing advanced coatings and optimizing surface treatments to achieve desired characteristics [42].
In the field of biomaterials, specific subfields have emerged, such as biomedical implants using titanium alloys and stainless steel, bioactive metals like magnesium alloys and Nitinol for biodegradable and flexible medical devices, and metallic scaffolds for tissue engineering, providing structures that support cell growth and tissue regeneration [43]. Table 2 highlights the broad range of applications and subfields where ML and AI are making substantial contributions to the field of metal materials.
Researchers extensively explored the vast realm of chemical databases, including Citrination and The Materials Project, to meticulously examine the descriptors associated with metals. This comprehensive investigation aimed to unravel the intricate chemical properties inherent in different metal compositions, shedding light on their diverse behaviors and characteristics [44]. Delving into the realm of biomedical applications, scientists undertook a meticulous review of nanoparticles’ predictive capabilities using advanced machine learning methodologies. Through the utilization of random forest classifiers, Yousefi et al. achieved a remarkable level of accuracy in forecasting material behaviors relevant to biomedical contexts, signifying a significant leap forward in the field of material science [45]. The exploration of quantum-mechanical properties inherent in inorganic complexes prompted researchers to embark on a thorough examination of various feature selection methods. This rigorous investigation aimed to elucidate the complex structure–property relationships prevalent in transition metal chemistry, offering valuable insights into the fundamental mechanisms governing these intricate systems [46]. Leveraging the synergistic potential of cellular automata and machine learning, scientists endeavored to simulate the intricate processes underlying the growth of aluminum alloys. By meticulously analyzing the interplay between porosity and solidification parameters, researchers sought to optimize material properties during the fabrication process, thereby enhancing the overall efficiency and effectiveness of alloy production techniques [47]. In a demonstration of machine learning’s prowess in materials science, He et al. embarked on a quest to streamline the synthesis and characterization of metal-organic frameworks (MOFs). Through the judicious application of advanced ML methodologies, they successfully identified and prioritized potential candidates, expediting the discovery process and facilitating the exploration of novel MOF structures with diverse applications [48].
In the field of metals, QSAR modeling and ML techniques show promise in predicting cytotoxicity and other properties. These techniques can handle diverse data rapidly, especially in image analysis and phenotypic recognition. However, challenges such as data standardization and collaboration issues hinder the adoption of ML in these fields. Standardizing data collection and promoting collaboration among researchers are crucial steps to improve data sharing and fully utilize ML in biomaterials and tissue engineering research. In a comprehensive study focused on predicting cell viability in the presence of metal oxide nanomaterials, researchers developed quasi-QSAR models to assess toxicity and biocompatibility. This endeavor not only shed light on the potential health implications of exposure to these materials but also underscored the pivotal role of machine learning in advancing our understanding of nanotoxicology [49].
Exploring novel methodologies in transition metal chemistry, researchers pioneered the utilization of innovative topological representations, such as RAC (Ring Analysis with Connectivity). This groundbreaking approach aimed to enhance the predictive capabilities of machine learning models, enabling more accurate forecasts of various properties critical to the field of inorganic chemistry [50]. Through the development of a sophisticated machine learning model, researchers endeavored to predict the adsorption energies of CH4-related species on Cu-based alloys. This endeavor offered invaluable insights into the design of selective catalytic reactions, potentially revolutionizing the field of catalysis by facilitating the development of more efficient and sustainable processes [51]. Identifying key structural features influencing the mechanical properties of MOFs, researchers harnessed the power of artificial neural networks to expedite the generation of MOFs with desired mechanical properties. This innovative approach aimed to minimize the need for laborious experimental procedures, thereby accelerating the pace of materials discovery and optimization [42]. Utilizing a diverse array of machine learning models, researchers endeavored to predict and optimize the hardness of high entropy alloys. By meticulously analyzing the intricate relationships between alloy composition and mechanical properties, scientists aimed to develop robust predictive models capable of guiding the design and fabrication of high-performance alloys for various applications [52].
Figure 7 shows that by using ML surrogate models to predict properties and uncertainties and employing a DOE utility function to balance exploration and exploitation, the method streamlines the selection and synthesis of promising candidates. Iterative incorporation of new data continuously improves the model’s accuracy, while the inclusion of a feature pool in the process allows for more precise training from compositions and physical features. This approach accelerates the development of high entropy alloys by optimizing the experimental process and reducing the time and resources required for material discovery [52].
Employing sophisticated algorithms such as random forest and symbolic regression, researchers gained valuable insights into predicting mechanical properties and phases of amorphous metallic alloys. This comprehensive analysis aimed to elucidate the underlying mechanisms governing the behavior of these complex materials, paving the way for enhanced predictive capabilities and deeper understanding in the field of materials science [53]. A comprehensive review explored the integration of data-driven science, particularly machine learning, to accelerate the design and discovery of MOFs. By synthesizing insights from molecular simulations and ML algorithms, researchers sought to decipher the complex structure–performance relationships inherent in MOFs, offering valuable guidance for future research endeavors in this rapidly evolving field [54].
An ANN model was used to predict water contact angles (WCA) and protein adsorption on self-assembled monolayers (SAMs) based on the chemical structures of SAM molecules. After training the ANN with data from 145 SAMs, it successfully predicted WCA and protein adsorption. This analysis highlighted the importance of various structural parameters for these properties, aiding material design [6].
From optimizing alloy design and additive manufacturing processes to enhancing failure analysis and surface engineering, these technologies have enabled more efficient and precise material development. The integration of ML and DOE-based iterative design loops, as shown in Figure 7, exemplifies how predictive models and experimental optimization can accelerate the discovery and application of high entropy alloys and metal-organic frameworks. These approaches not only streamline the selection and synthesis of promising candidates but also continuously improve through iterative data incorporation.

2.2. Polymer Materials

The integration of ML with biomaterial research has significantly advanced the design and discovery of novel polymer-based biomaterials. One notable study utilized LAD regression to analyze and predict polymers for safe and effective gene delivery. This was achieved by conducting a large-scale series of 12,000 transfection experiments based on the Michael addition polymerization of acrylates with amines [55]. QSAR models have also been employed to predict protein adsorption and cellular response on biodegradable polymers designed for tissue engineering applications. These models used descriptors calculated through molecular operating environments or Dragon and were optimized using artificial neural networks [56].
High-throughput methods have played a crucial role in understanding the structure–function relationships of polymers. For instance, ToF SIMS analysis was correlated with cell adhesion properties using PLS regression, highlighting the importance of surface functionalities in controlling human embryoid body cell adhesion [57]. Additionally, a collection of 496 polyacrylate microarrays was created to investigate the effects of hydrophobicity, hydrophilicity, and crosslinking density on polymer properties and their influence on cellular responses [58].
To further understand the interactions between polymer properties and cellular behaviors, the nonlinear Bayesian neural network model, BRANNGP, was employed to study the correlation between human embryonic stem cell (hES) adhesion and the attributes of polyacrylate surfaces [59]. Over the past decade, significant progress has been made in using polymers as scaffolds in tissue engineering to mimic the functions of natural tissues and create artificial tissues and organ substitutes [60]
The Polymer Genome Project exemplifies the integration of high-throughput computation, targeted experimentation, and ML to design novel polymer dielectrics for capacitive energy storage. This project has resulted in the creation of an online knowledgebase that stores accumulated data and prediction models, facilitating efficient material selection and design, as illustrated in Figure 8 [61].
Additionally, Bayesian network and ML techniques have been applied to identify key factors influencing the growth of small cracks in polycrystalline materials, combining high-resolution experiments with crystal plasticity simulations to reliably predict crack propagation [62].
The design of self-assembling dipeptide hydrogels has been enhanced by ML, which efficiently predicts hydrogels’ physical characteristics. This combinatorial approach has shown potential in various biomedical applications, such as drug delivery and tissue engineering [63]. Melt electrowriting (MEW) has been employed to create 3D substrates with customizable porous structures, which can influence cell morphology and control cell phenotype at the single-cell level. Additionally, a machine-learning-based metrology framework has been developed to investigate how these substrate architectures affect cell shape and the distribution of focal adhesion proteins [64].
Data-driven approaches have also been applied to the study of organic photovoltaic polymers. Methods such as self-organizing mapping and tailored polymer descriptor generators have identified key factors like photon energy loss and the number of fluorine atoms, correlating strongly with power conversion efficiency [65]. As shown in Figure 9 the data processing pipeline begins with a preprocessed dataset, where relevant features are selected, and input data include SMILES notation and other molecular properties. These inputs form the training set for self-organizing maps (SOM), which are then trained and validated. A molecular descriptor generator (MDG) creates molecular descriptors, and quality measurements assess model performance. Extracted molecular descriptors undergo feature scaling and normalization. The output includes heatmaps to visualize property–property correlations, U-matrix maps for unified distance visualization in SOM, cluster maps for displaying clustering results, and information-projected cluster maps to visualize property–structure relationships.
Understanding how polymers influence the immune system is crucial for creating biomaterials that do not trigger adverse foreign body responses. ML has been employed to link polymer chemistry with immune-instructive behaviors [66]. Deep learning models have been developed for predicting and designing the secondary structure content of structural proteins. These models utilize convolutional layers, BiLSTM units, and other techniques to process sequential data, achieving high accuracy in protein structure prediction [67].
Similarly, a deep learning framework has been proposed to predict the thermal stability (Tm values) of collagen sequences, essential for collagen-based biomaterials in tissue repair and regeneration [68]. Figure 9 indicates mapping of amino acid sequences into latent space and the architecture of a machine learning model designed to predict TM values.
Figure 9. LSTM model for prediction of the thermal stability. The figure is reproduced from [68].
Figure 9. LSTM model for prediction of the thermal stability. The figure is reproduced from [68].
Applsci 14 06590 g009
Machine learning has revolutionized polymeric biomaterial design by addressing the limitations of traditional trial-and-error methods, enabling faster and more efficient material discovery. Current efforts highlight the potential of ML in developing new polymeric biomaterials, despite challenges such as data standardization and parameter characterization [69].
AI and ML have revolutionized polymeric biomaterial design by addressing the limitations of traditional trial-and-error methods, enabling faster and more efficient material discovery. Current efforts highlight the potential of ML in developing new polymeric biomaterials, despite challenges such as data standardization and parameter characterization. Table 3 summarizes the various applications of ML and AI in polymer materials, underscoring the transformative impact of these technologies on the field.

2.3. Composite Materials

The intersection of composite materials and advanced computational techniques like AI and ML is reshaping the landscape of materials science. Unlike traditional methods, which often rely on time-consuming and costly experimental procedures, ML and AI offer a streamlined approach to discovering and optimizing new composites. Researchers can predict material behaviors, uncover new material combinations, and enhance properties with unprecedented accuracy and speed by using a vast number of datasets and sophisticated algorithms. This technological fusion not only accelerates the development cycle but also opens up new possibilities for creating high-performance composites that meet specific, demanding criteria across various industries.
One of the studies delved into the creation of statistical models to capture elastic localization relationships in high-contrast composite materials. By systematically evaluating various feature selection and regression approaches, the research concluded that combining basic and engineered feature descriptors, particularly through ensemble methods like random forests, markedly improves prediction performance and efficiency [70]. Another investigation introduced a 3D convolutional neural network (CNN) to model elastic homogenization in high-contrast composites. Figure 10 presents a schematic illustration of a sample 3D CNN configuration. This deep learning approach significantly outperformed traditional physics-inspired methods, reducing the mean absolute scaled error (MASE) by up to 54% [71] The study highlighted the potential of advanced CNN architectures and regularization techniques to further enhance model performance, emphasizing deep learning’s transformative impact on materials science.
Researchers proposed a bioinspired approach to hierarchical composite design by integrating machine learning with finite element analysis. This model accurately predicted mechanical properties and generated new microstructural patterns, validated through additive manufacturing and tensile testing. This method demonstrated machine learning’s capability to significantly speed up the exploration and optimization of high-performing materials [72]. An innovative AI-based method was introduced for optimizing the design of 2D nanocomposites for shear crack resistance. In the study, the proposed combined a deep CNN with a global optimizer to explore the design space efficiently, showcasing the AI tool’s ability to predict and design composites resilient to various fracture conditions, thus proving its versatility and scalability [73]. The training set was split into 90% for training and 10% for validation to prevent overfitting, with the loss function showing significant initial decreases and converging around 10−5 after 1000 epochs, as depicted in Figure 11b. The training process flowchart is shown in Figure 11a, and the regression values for toughness and strength are presented in Figure 11c,d, respectively.
In a study focused on high-entropy alloys (HEAs), machine learning, particularly ANNs, was employed to predict and optimize the composition of HEAs for maximum hardness. Integrating simulated annealing with the ANN, the study successfully discovered novel HEA compositions with superior hardness, surpassing previously reported values, thus illustrating the potential of ML in alloy design [74]. Another study utilized AI to predict fracture patterns in crystalline solids, demonstrating the potential for generative methods to support reverse engineering of mechanical properties. The study suggested integrating adversarial training and advanced neural networks to design crack-insensitive materials, adaptable to various force fields and material properties [75]. Additionally, Wu et al. developed a neural network called “βLow” to recommend new low-modulus β titanium alloys for biomedical applications, as shown in Figure 12. Despite limited data, the network identified a promising alloy (Ti-12Nb-12Zr-12Sn) that is cost-effective and biocompatible, highlighting the efficiency of ML in accelerating alloy design for specific applications [76].
A systematic framework was proposed for phase prediction in HEAs, utilizing a genetic algorithm to select ML models and material descriptors. This optimized classification model achieved high accuracy in identifying solid-solution HEAs and their phases, demonstrating the efficacy of combining genetic algorithms with active learning approaches for iterative model improvement [77]. In response to the global challenge of diagnosing respiratory infections, a study combined surface-enhanced Raman spectroscopy (SERS) with ML for virus detection. This method enabled rapid, accurate classification of various viruses, highlighting the potential of ML in enhancing diagnostic techniques for timely and effective epidemic response [78]. The integration of multi-scale simulations and machine learning in researching HEAs was examined, addressing challenges like data quality and model interpretability. This approach highlighted the importance of combining experimental and theoretical research to enhance the precision of multi-scale simulations and the interpretability of ML models [79]. Using synchrotron microfocus X-ray diffraction, researchers explored a machine-learning-based approach to quickly predict fiber orientation metrics in biological composites. This method, validated with real experimental data, showed promise for automating data processing, as well as accelerating the analysis of large datasets in biomaterials research [80].
Another study aimed to identify suitable biomaterials for bone scaffolds using machine learning models, particularly ANNs. The research effectively ranked biomaterials, demonstrating high accuracy in selecting candidates for bone tissue engineering, thus illustrating the potential of ML in biomaterial selection and prioritization [81]. Lastly, an innovative approach using a CNN was introduced to generate spinodal structures for biomaterials, ideal for bone repair due to their stochastic and interconnected pore channels. As shown in Figure 13, the proposed model provided flexibility and computational efficiency, improving the development of stochastic biomaterials for biomedical applications [82].
The applications of AI and ML in composite materials research, as detailed in Table 4, illustrates a paradigm shift in the field. These technologies enable unprecedented precision and efficiency in material design, structural health monitoring, manufacturing optimization, and sustainability efforts. By leveraging AI and ML, researchers can rapidly innovate and improve material properties, leading to more advanced and capable composites. This integration not only enhances current practices but also opens up new possibilities for future research and development, ensuring that composite materials continue to evolve and meet the demands of various high-tech industries.

2.4. Inorganic Materials

The landscape of inorganic materials research is undergoing a revolutionary transformation driven by the integration of ML and AI. Unlike traditional methods, which often rely heavily on trial and error, these advanced computational techniques offer a more efficient and systematic approach to material discovery and optimization.
The study presents a machine learning framework designed to predict various properties of inorganic materials, including both crystalline and amorphous structures. The framework utilizes a diverse set of chemical attributes and a unique grouping method to enhance predictive accuracy. The results demonstrate its effectiveness in predicting properties such as band gap energy and glass-forming ability, showcasing the potential of ML in materials science [83]. Dataset’s Chemical Diversity Limits the Generalizability of Machine Learning Predictions: Highlighting the limitations of existing datasets like QM9 introduces the PC9 dataset from the PubChemQC project, which offers greater chemical diversity. Testing various ML models, including Kernel ridge regression and neural networks on both datasets revealed that PC9 provided higher energy prediction accuracy, emphasizing the importance of chemical diversity in training data for generalizable ML predictions [84]. Another study combined high-throughput molecular dynamics simulations with machine learning to predict the elastic properties of silicate glasses. The ANN emerged as the most accurate model, while the LASSO algorithm provided a simpler, more interpretable model. The method’s versatility allows its application to other properties and systems, illustrating its broad potential in the field [85]. Addressing the challenge of sparse datasets, this research employed Gaussian process regression (GPR) to predict Young’s modulus in silicate glasses. GPR outperformed neural networks in avoiding overfitting and provided a reliable quantification of prediction accuracy, demonstrating its advanced capability in developing functional materials [86]. Focusing on biomedical glasses used in healthcare, the research introduces an ensemble machine learning model to predict and control the dissolution behavior of these materials in biological fluids. This approach aims to meet the demand for glasses with precise degradation rates, offering a path towards optimized glass compositions and improved performance in medical applications [41]. In Figure 14, the primary applications of ML models for biomedical glasses are detailed, including the prediction of time-dependent biomedical glass dissolution behavior and the optimization of biomedical glass composition.
Drawing inspiration from millennia of natural design exploration, one study employed advanced computational algorithms to create innovative designs based on diatom structures. Central to the study is the training of a VQGAN neural network with diatom structure data, enabling it to meticulously replicate the intricate architectural details through extensive training epochs. Furthermore, the integration of the VQGAN model with CLIP, a versatile image classifier, allows for the interpretation of textual prompts to generate corresponding 3D models. These models are then physically realized using additive manufacturing technology. The methodology not only explores various biological design cues but also implements systematic optimization to achieve specific design objectives, integrating a diverse array of material design sets [87]. Another research study focuses on detecting cancer biomarkers by analyzing odor-evoked calcium responses in the rat olfactory bulb [88]. By employing neural networks and transfer learning from simulated data, this research aims to improve the accuracy of recognition. This approach has the potential to contribute to more effective cancer detection methods and diagnostic tools, enhancing early diagnosis and treatment strategies. The neural networks are trained to recognize specific patterns in the calcium responses, which correlate with the presence of cancer biomarkers, thereby offering a novel and innovative way to identify cancer at its early stages. Figure 15 indicates that exhaled air is sampled and analyzed via intravital calcium imaging in the olfactory bulb glomeruli, producing raw video data. These data are pre-processed into a tensor format and then encoded by a neural network into an encoding vector. The process includes encoding exhaled air samples from healthy individuals and those with cancer biomarkers, followed by calibration. Unknown air samples are encoded and tested, with a k-nearest neighbors (kNN) algorithm classifying the samples as either healthy or cancerous, thereby enhancing detection accuracy.
Table 5 provides a comprehensive overview, summarizing the broader scope of ongoing research efforts in inorganic materials science beyond the specific studies previously discussed. Ultimately, it highlights the collective advancements achieved through AI and ML applications, which have significantly improved predictive modeling and enabled the creation of novel materials with specific, several properties.

3. Discussion

The integration of ML, DL, and AI into biomaterials research has revolutionized the field, offering unprecedented capabilities for material design, optimization, and predictive modeling. This section synthesizes the findings from various studies and discusses the broader implications, challenges, and future directions for the application of these technologies in biomaterials.
Simulation systems are also crucial in the training of AI and ML. The simulation systems used must be well-established in the literature. Particularly in data acquisition, Monte Carlo-based simulation systems play a significant role in the radiological data acquisition of biomaterials [89,90,91]. These simulations are critical for predicting the interactions of biomaterials with biological systems and understanding the outcomes of these interactions [89,90,91,92]. Monte Carlo methods are used to comprehend the behaviors and probability distributions of complex systems, providing the necessary data to optimize the performance of biomaterials [93,94,95]. Additionally, although the process of experimental data acquisition is laborious, it is essential for understanding how biomaterials behave under real-world conditions [96,97,98]. Investigating the radiological properties of biomaterials in laboratory conditions enables the development of production and application processes [95,97,98]. These laboratory studies contribute to the development of reliable and effective products by examining the behavior of biomaterials under various radiological conditions. By combining experimental data with simulation results, more comprehensive and reliable models can be created [95,98,99]. In this way, all stages necessary for ensuring the safe and effective use of biomaterials in clinical applications are carefully planned and optimized. Consequently, the combination of simulation and experimental data, which allows AI and ML to conduct broader and more in-depth research in the field of biomaterials, will lead to significant advancements in material science and medicine.

3.1. Impact on Biomaterials Research

Machine learning techniques have significantly advanced the study of polymers, metals, composites, and inorganic materials. For instance, the use of databases like Citrination and The Materials Project has enabled researchers to explore extensive chemical descriptors, leading to a better understanding of the properties and behaviors of different metal compositions [100]. In the realm of biomedical applications, ML algorithms like random forest classifiers have achieved remarkable accuracy in predicting the behavior of nanoparticles, facilitating their use in medical contexts [101].
Similarly, ML has transformed the design and discovery of polymer-based biomaterials. Techniques such as LAD regression and QSAR models have optimized gene delivery systems and predicted protein adsorption on biodegradable polymers [102,103]. High-throughput methods and projects like the Polymer Genome have further accelerated the discovery and optimization of novel polymers with desired properties [104]. In composite materials, ML approaches, including statistical models and deep learning, have enabled the prediction of elastic localization, structural properties, and mechanical behaviors [105,106]. These advancements have facilitated the design of bioinspired hierarchical composites and nanocomposites optimized for specific applications. For inorganic materials, ML frameworks have predicted various properties, including band gap energy and glass-forming ability [83]. The importance of chemical diversity in training data for generalizable predictions has been highlighted, emphasizing the need for diverse datasets to improve ML model accuracy.

3.2. Challenges and Limitations

Despite the significant advancements, several challenges and limitations hinder the full potential of ML and AI in biomaterials research.
  • The development of robust ML models requires large, high-quality datasets [107]. However, the field of biomaterials often faces limitations in terms of standardized, large-scale datasets. Ensuring data quality and consistency is crucial for the reliability of ML predictions.
  • ML models, particularly deep learning models, are often considered “black-box” systems, making it challenging to interpret their decisions [108]. Enhancing the interpretability of these models is essential, especially in applications where transparency and accountability are critical.
  • Training complex ML models, such as deep neural networks, requires significant computational power and specialized hardware like GPUs and TPUs [109]. This requirement can limit the accessibility of ML techniques to researchers with limited resources.
  • The use of ML in biomaterials research raises ethical concerns, including data privacy, bias in training data, and the potential for unintended consequences [110]. Addressing these issues is vital to ensure the responsible application of ML technologies.
  • While ML can significantly accelerate material discovery and optimization, it is crucial to integrate these computational approaches with experimental validation [111]. Ensuring that ML predictions are experimentally validated can enhance the reliability and applicability of the findings.
  • In biomaterials research, the biocompatibility of materials and potential toxicity issues are of paramount importance [8]. However, utilizing ML and AI techniques to evaluate and predict these factors presents notable challenges [12]. Moreover, the standards for sensitivity and accuracy required for biomedical applications are high. ML and AI methods play a crucial role in meeting these standards, yet their application in biomaterials research can be complex and demanding.
  • Biomaterials often exhibit complex structures, making it difficult to fully characterize these materials [112]. ML models must effectively navigate this complexity to provide comprehensive insights.

3.3. Future Directions

The future of ML and AI in biomaterials research holds immense potential. As more comprehensive and standardized datasets become available, the accuracy and reliability of ML models will continue to improve. Advancements in computational power and algorithms will facilitate the development of more complex models, enabling the exploration of intricate material behaviors and interactions [113]. For example, combining ML with high-throughput screening techniques can expedite the discovery of novel materials with tailored properties.
Interdisciplinary collaboration will be essential to address the challenges and maximize the potential of ML in biomaterials [114]. Researchers from materials science, computer science, biology, and ethics must work together to develop robust, interpretable, and ethically sound ML models [115,116].

4. Conclusions

The application of ML, DL, and AI has revolutionized biomaterials research, offering powerful tools for the design, optimization, and predictive modeling of various materials. From polymers and metals to composites and inorganic materials, these technologies have significantly advanced our understanding and capabilities in biomaterial science. ML has enabled the development of predictive models that can accurately forecast material properties, reducing the reliance on traditional trial-and-error methods. This has led to the discovery of novel materials with enhanced performance, optimized for specific applications in medicine, engineering, and beyond.
Despite the remarkable progress, several challenges remain, including data availability and quality, model interpretability, computational resource requirements, and ethical considerations. Addressing these challenges will be crucial to fully realize the potential of ML and AI in biomaterials research. Looking ahead, the future of ML and AI in biomaterials is promising. As the field continues to evolve, the integration of these technologies with experimental methods and other emerging technologies will further accelerate the discovery and optimization of innovative materials. Interdisciplinary collaboration and a focus on ethical and responsible AI use will be key to ensuring that these advancements benefit society.
Utilizing the capabilities of machine learning and artificial intelligence, researchers can expand the horizons of biomaterials research, facilitating the creation of advanced materials that could revolutionize multiple industries and significantly improve human health and well-being.

Author Contributions

Conceptualization, Y.G., F.E. and M.S.G.; Methodology, Y.G. and F.E.; Software, Y.G. and M.S.G.; Validation, K.A. and T.A.; Formal Analysis, Y.G. and F.E.; Investigation, M.S.G.; Resources, Y.G., F.E., K.A. and T.A.; Writing—Original Draft Preparation, Y.G. and F.E.; Writing—Review and Editing, M.S.G.; Visualization, Y.G. and F.E.; Supervision, M.S.G. and S.A. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Acknowledgments

This work was partly supported by Fonet Information Technologies.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Cecen, A.; Dai, H.; Yabansu, Y.C.; Kalidindi, S.R.; Song, L. Material structure-property linkages using three-dimensional convolutional neural networks. Acta Mater. 2018, 146, 76–84. [Google Scholar] [CrossRef]
  2. Vasilevich, A.; Boer, J.d. Robot-scientists will lead tomorrow’s biomaterials discovery. Curr. Opin. Biomed. Eng. 2018, 6, 74–80. [Google Scholar] [CrossRef]
  3. Suwardi, A.; Wang, F.; Xue, K.; Han, M.Y.; Teo, P.; Wang, P.; Wang, S.; Liu, Y.; Ye, E.; Li, Z.; et al. Machine Learning-Driven Biomaterials Evolution. Adv. Mater. 2022, 34, 2102703. [Google Scholar] [CrossRef] [PubMed]
  4. Lazarovits, J.; Sindhwani, S.; Tavares, A.J.; Zhang, Y.; Song, F.; Audet, J.; Krieger, J.R.; Syed, A.M.; Stordy, B.; Chan, W.C.W. Supervised Learning and Mass Spectrometry Predicts the in Vivo Fate of Nanomaterials. ACS Nano 2019, 13, 8023–8034. [Google Scholar] [CrossRef] [PubMed]
  5. Hakimi, O.; Krallinger, M.; Ginebra, M.-P. Time to kick-start text mining for biomaterials. Nat. Rev. Mater. 2020, 5, 8. [Google Scholar] [CrossRef]
  6. Kwaria, R.J.; Mondarte, E.A.Q.; Tahara, H.; Chang, R.; Hayashi, T. Data-Driven Prediction of Protein Adsorption on Self-Assembled Monolayers toward Material Screening and Design. ACS Biomater. Sci. Eng. 2020, 6, 4949–4956. [Google Scholar] [CrossRef] [PubMed]
  7. Zhou, Y.; Ping, X.; Guo, Y.; Heng, B.C.; Wang, Y.; Meng, Y.; Jiang, S.; Wei, Y.; Lai, B.; Zhang, X.; et al. Assessing Biomaterial-Induced Stem Cell Lineage Fate by Machine Learning-Based Artificial Intelligence. Adv. Mater. 2023, 35, e2210637. [Google Scholar] [CrossRef] [PubMed]
  8. Al-Kharusi, G.; Dunne, N.J.; Little, S.; Levingstone, T.J. The Role of Machine Learning and Design of Experiments in the Advancement of Biomaterial and Tissue Engineering Research. Bioengineering 2022, 9, 561. [Google Scholar] [CrossRef] [PubMed]
  9. Ke, R.; Li, B. A Review for Machine Learning Applications in Characterizing Biomaterials and Biological Materials Properties. Am. J. Biomed. Sci. Res. 2021, 13, 432–436. [Google Scholar]
  10. Kerner, J.; Alan Dogan, H.v.R. Machine learning and big data provide crucial insight for future biomaterials discovery and research. Acta Biomater. 2021, 130, 54–65. [Google Scholar] [CrossRef]
  11. Pratap, A.; Sardana, N. Machine learning-based image processing in materials science and engineering: A review. Mater. Today Proc. 2022, 62, 7341–7347. [Google Scholar] [CrossRef]
  12. Vinoth, A.; Datta, S. Computational intelligence-based design of biomaterials. Comput. Methods Mater. Sci. 2022, 22, 229–262. [Google Scholar] [CrossRef]
  13. Mateu-Sanz, M.; Fuenteslópez, C.V.; Uribe-Gomez, J.; Haugen, H.J.; Pandit, A.; Ginebra, M.P.; Hakimi, O.; Krallinger, M.; Samara, A. Redefining biomaterial biocompatibility: Challenges for artificial intelligence and text mining. Trends Biotechnol. 2024, 42, 402–417. [Google Scholar] [CrossRef] [PubMed]
  14. Mitchell, M. Artificial Intelligence: A Guide for Thinking Humans. Farrar, Straus and Giroux: New York, NY, USA, 2019; ISBN 9780374715236. [Google Scholar]
  15. Bishop, C.M. Pattern Recognition and Machine Learning; Springer: Berlin/Heidelberg, Germany, 2006. [Google Scholar]
  16. Shrestha, Y.R.; Ben-Menahem, S.M.; Von Krogh, G. Organizational decision-making structures in the age of artificial intelligence. Calif. Manag. Rev. 2021, 63, 73–96. [Google Scholar] [CrossRef]
  17. Janiesch, C.; Zschech, P.; Heinrich, K. Machine learning and deep learning. Electron Mark. 2021, 31, 685–695. [Google Scholar] [CrossRef]
  18. Goodfellow, I.; Bengio, Y.; Courville, A. Deep Learning; MIT Press: Cambridge, MA, USA, 2016. [Google Scholar]
  19. LeCun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature 2015, 51, 436–444. [Google Scholar] [CrossRef]
  20. Madani, K.; Velay, M.; Mosser, L.; Debayle, J. Enhancing human decision making using deep learning for complex problems. IEEE Access 2018, 6, 14697–14705. [Google Scholar]
  21. Hastie, T.; Tibshirani, R.; Friedman, J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction; Springer: Berlin/Heidelberg, Germany, 2009. [Google Scholar]
  22. Montgomery, D.C.; Peck, E.A.; Vining, G.G. Introduction to Linear Regression Analysis; John Wiley & Sons: Hoboken, NJ, USA, 2012. [Google Scholar]
  23. Murphy, K.P. Machine Learning: A Probabilistic Perspective; MIT Press: Cambridge, MA, USA, 2012. [Google Scholar]
  24. Jain, A.K. Data clustering: 50 years beyond K-means. Pattern Recognit. Lett. 2010, 31, 651–666. [Google Scholar] [CrossRef]
  25. MacQueen, J. Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability; University of California Press: Berkeley, CA, USA, 1967; Volume 1, No. 14. [Google Scholar]
  26. Johnson, S.C. Hierarchical clustering schemes. Psychometrika 1967, 32, 241–254. [Google Scholar] [CrossRef]
  27. Jolliffe, I.T. Principal Component Analysis; Springer: Berlin/Heidelberg, Germany, 2002. [Google Scholar]
  28. Van der Maaten, L.; Hinton, G. Visualizing data using t-SNE. J. Mach. Learn. Res. 2008, 9, 2579–2605. [Google Scholar]
  29. Agrawal, R.; Imieliński, T.; Swami, A. Mining association rules between sets of items in large databases. In Proceedings of the 1993 ACM SIGMOD International Conference on Management of Data, Washington, DC, USA, 26–28 May 1993. [Google Scholar]
  30. Theodoridis, S.; Koutroumbas, K. Pattern Recognition; Academic Press: Cambridge, MA, USA, 2008. [Google Scholar]
  31. Xu, R.; Wunsch, D. Survey of clustering algorithms. IEEE Trans. Neural Netw. 2005, 16, 645–678. [Google Scholar] [CrossRef] [PubMed]
  32. Sutton, R.S.; Barto, A.G. Reinforcement Learning: An Introduction; MIT Press: Cambridge, MA, USA, 2018. [Google Scholar]
  33. Amodei, D.; Olah, C.; Steinhardt, J.; Christiano, P.; Schulman, J.; Mané, D. Concrete Problems in AI Safety. arXiv 2016, arXiv:1606.06565. [Google Scholar]
  34. Ruder, S. An overview of gradient descent optimization algorithms. arXiv 2016, arXiv:1609.04747. [Google Scholar]
  35. Yosinski, J.; Clune, J.; Bengio, Y.; Lipson, H. How transferable are features in deep neural networks? In Advances in Neural Information Processing Systems; MIT Press: Cambridge, MA, USA, 2014; pp. 3320–3328. [Google Scholar]
  36. Li, J.; Yao, W.; Lu, Y.; Chen, J.; Sun, Y.; Hu, X. High fidelity FEM based on deep learning for arbitrary composite material structure. Compos. Struct. 2024, 340, 118176. [Google Scholar] [CrossRef]
  37. Ford, E.; Maneparambil, K.; Rajan, S.; Neithalath, N. Machine learning-based accelerated property prediction of two-phase materials using microstructural descriptors and finite element analysis. Comput. Mater. Sci. 2021, 191, 110328. [Google Scholar] [CrossRef]
  38. Chahar, R.S.; Mukhopadhyay, T. Multi-fidelity machine learning based uncertainty quantification of progressive damage in composite laminates through optimal data fusion. Eng. Appl. Artif. Intell. 2023, 125, 106647. [Google Scholar] [CrossRef]
  39. Zhou, K.; Sun, H.; Enos, R.; Zhang, D.; Tang, J. Harnessing deep learning for physics-informed prediction of composite strength with microstructural uncertainties. Comput. Mater. Sci. 2021, 197, 110663. [Google Scholar] [CrossRef]
  40. Xiong, J.; Zhang, T.-Y.; Shi, S.-Q. Machine learning prediction of elastic properties and glass-forming ability of bulk metallic glasses. MRS Commun. 2019, 9, 576–585. [Google Scholar] [CrossRef]
  41. Han, T.; Stone-Weiss, N.; Huang, J.; Goel, A.; Kumar, A. Machine learning as a tool to design glasses with controlled dissolution for healthcare applications. Acta Biomater. 2020, 107, 286–298. [Google Scholar] [CrossRef]
  42. Moghadam, P.Z.; Rogge, S.M.; Li, A.; Chow, C.-M.; Wieme, J.; Moharrami, N.; Aragones-Anglada, M.; Conduit, G.; Gomez-Gualdron, D.A.; Van Speybroeck, V.; et al. Structure-Mechanical Stability Relations of Metal-Organic Frameworks via Machine Learning. Matter 2019, 1, 219–234. [Google Scholar] [CrossRef]
  43. Festas, A.; Ramos, A.; Davim, J. Medical devices biomaterials—A review. Proc. Inst. Mech. Eng. Part L J. Mater. Des. Appl. 2020, 234, 218–228. [Google Scholar] [CrossRef]
  44. Jones, D.E.; Ghandehari, H.; Facelli, J.C. A review of the applications of data mining and machine learning for the prediction of biomedical properties of nanoparticles. Comput. Methods Programs Biomed. 2016, 132, 93–103. [Google Scholar] [CrossRef]
  45. Yousefi, M.; Rahmani, K.; Rajabi, M.; Reyhani, A.; Moudi, M. Random forest classifier for high entropy alloys phase diagnosis. Afrika Matematika 2024, 35, 57. [Google Scholar] [CrossRef]
  46. Janet, J.P.; Kulik, H.J. Resolving Transition Metal Chemical Space: Feature Selection for Machine Learning and Structure–Property Relationships. J. Phys. Chem. 2017, 121, 8939–8954. [Google Scholar] [CrossRef]
  47. Hu, Y.; Xie, J.; Liu, Z.; Ding, Q.; Zhu, W.; Zhang, J.; Zhang, W. CA method with machine learning for simulating the grain and pore growth of aluminum alloys. Comput. Mater. Sci. 2018, 142, 244–254. [Google Scholar] [CrossRef]
  48. He, Y.; Cubuk, E.D.; Allendorf, M.D.; Reed, E.J. Metallic Metal–Organic Frameworks Predicted by the Combination of Machine Learning Methods and Ab Initio Calculations. J. Phys. Chem. Lett. 2018, 9, 4562–4569. [Google Scholar] [CrossRef]
  49. Gajewicz, A.; Puzyn, T.; Rasulev, B. Towards understanding mechanisms governing cytotoxicity of metal oxides nanoparticles: Hints from nano-QSAR studies. Nanotoxicology 2017, 11, 751–758. [Google Scholar] [CrossRef]
  50. Nandy, A.; Duan, C.; Janet, J.P.; Gugler, S.; Kulik, H.J. Strategies and Software for Machine Learning Accelerated Discovery in Transition Metal Chemistry. Ind. Eng. Chem. Res. 2018, 57, 13973–13986. [Google Scholar] [CrossRef]
  51. Toyao, T.; Suzuki, K.; Kikuchi, S.; Takakusagi, S.; Shimizu, K.-i.; Takigawa, I. Toward Effective Utilization of Methane: Machine Learning Prediction of Adsorption Energies on Metal Alloys. J. Phys. Chem. 2018, 122, 8315–8326. [Google Scholar] [CrossRef]
  52. Wen, C.; Zhang, Y.; Wang, C.; Xue, D.; Bai, Y.; Antonov, S.; Dai, L.; Lookman, T.; Su, Y. Machine learning assisted design of high entropy alloys with desired property. Acta Mater. 2017, 170, 109–117. [Google Scholar] [CrossRef]
  53. Shi, S.; Xıong, J. Data for: A machine-learning approach to predicting and understanding the properties of amorphous metallic alloys. Mendeley Data 2020. [Google Scholar] [CrossRef]
  54. Demir, H.; Daglar, H.; Gulbalkan, H.C.; Aksu, G.O.; Keskin, S. Recent advances in computational modeling of MOFs: From molecular simulations to machine learning. Coord. Chem. Rev. 2023, 484, 215112. [Google Scholar] [CrossRef]
  55. Anderson, D.; Akinc, A.; Hossain, N.; Langer, R. Structure/property studies of polymeric gene delivery using a library of poly(β-amino esters). Mol. Ther. 2005, 11, 426–432. [Google Scholar] [CrossRef] [PubMed]
  56. Smith, J.R.; Kholodovych, V.; Knight, D.; Welsh, W.J.; Kohn, J. QSAR Models for the Analysis of Bioresponse Data from Combinatorial Libraries of Biomaterials. QSAR Comb. Sci. 2005, 24, 99–113. [Google Scholar] [CrossRef]
  57. Yang, J.; Mei, Y.; Hook, A.L.; Taylor, M.; Urquhart, A.J.; Bogatyrev, S.R.; Langer, R.; Anderson, D.G.; Davies, M.C.; Alexander, M.R. 3-Polymer surface functionalities that control human embryoid body cell adhesion revealed by high throughput surface characterization of combinatorial material microarrays. Biomaterials 2010, 31, 8827–8838. [Google Scholar] [CrossRef] [PubMed]
  58. Hook, A.L.; Anderson, D.G.; Langer, R.; Williams, P.; Davies, M.C.; Alexander, M.R. High throughput methods applied in biomaterial development and discovery. Biomaterials 2010, 31, 187–198. [Google Scholar] [CrossRef]
  59. Epa, V.; Yang, J.; Mei, Y.; Hook, A.; Langer, R.; Anderson, D.; Davies, M.; Alexander, M.; Winkler, D. Modelling human embryoid body cell adhesion to a combinatorial library of polymer surfaces. J. Mater. Chem. 2012, 22, 20902–20906. [Google Scholar] [CrossRef]
  60. Khademhosseini, A.; Langer, R.A. A decade of progress in tissue engineering. Nat. Protoc. 2016, 11, 1775–1781. [Google Scholar] [CrossRef] [PubMed]
  61. Mannodi-Kanakkithodi, A.; Chandrasekaran, A.; Kim, C.; Huan, T.D.; Pilania, G.; Botu, V.; Ramprasad, R. Scoping the polymer genome: A roadmap for rational polymer dielectrics design and beyond. Mater. Today 2018, 21, 785–796. [Google Scholar] [CrossRef]
  62. Rovinelli, A.; Sangid, M.D.; Proudhon, H.; Ludwig, W. Using machine learning and a data-driven approach to identify the small fatigue crack driving force in polycrystalline materials. npj Comput. Mater. 2018, 4, 35. [Google Scholar] [CrossRef]
  63. Li, F.; Han, J.; Cao, T.; Lam, W.; Fan, B.; Tang, W.; Chen, S.; Fok, K.; Li, L. Design of self-assembly dipeptide hydrogels and machine learning via their chemical features. Proc. Natl. Acad. Sci. USA 2019, 116, 11259–11264. [Google Scholar] [CrossRef] [PubMed]
  64. Tourlomousis, F.; Jia, C.; Karydis, T.; Mershin, A.; Wang, H.; Kalyon, D.; Chang, R. Machine learning metrology of cell confinement in melt electrowritten three-dimensional biomaterial substrates. Microsyst. Nanoeng. 2019, 5, 15. [Google Scholar] [CrossRef] [PubMed]
  65. Huang, Y.; Zhang, J.; Jiang, E.S.; Oya, Y.; Saeki, A.; Kikugawa, G. Structure–Property Correlation Study for Organic Photovoltaic Polymer Materials Using Data Science Approach. J. Phys. Chem. C 2020, 124, 12871–12882. [Google Scholar] [CrossRef]
  66. Rostam, H.M.; Fisher, L.E.; Hook, A.L.; Burroughs, L.; Luckett, J.C.; Figueredo, G.P.; Mbadugha, C.; Teo, A.C.; Latif, A.; Kämmerling, L.; et al. Immune-Instructive Polymers Control Macrophage Phenotype and Modulate the Foreign Body Response In Vivo. Matter 2020, 2, 1564–1581. [Google Scholar] [CrossRef]
  67. Yu, C.H.; Chen, W.; Chiang, Y.H.; Guo, K.; Moldes, Z.M.; Kaplan, D.L.; Buehler, M.J. End-to-End Deep Learning Model to Predict and Design Secondary Structure Content of Structural Proteins. ACS Biomater. Sci. Eng. 2022, 8, 1156–1165. [Google Scholar] [CrossRef] [PubMed]
  68. Yu, C.-H.; Khare, E.; Narayan, O.P.; Parker, R.; Kaplan, D.L.; Buehler, M.J. ColGen: An end-to-end deep learning model to predict thermal stability of de novo collagen sequences. J. Mech. Behav. Biomed. Mater. 2022, 125, 104921. [Google Scholar] [CrossRef] [PubMed]
  69. McDonald, S.M.; Augustine, E.K.; Lanners, Q.; Rudin, C.; Brinson, L.C.; Becker, M.L. Applied machine learning as a driver for polymeric biomaterials design. Nat. Commun. 2023, 14, 4838. [Google Scholar] [CrossRef] [PubMed]
  70. Liu, R.; Yabansu, Y.C.; Agrawal, A.; Kalidindi, S.R.; Choudhary, A.N. Machine learning approaches for elastic localization linkages in high-contrast composite materials. Integr. Mater. Manuf. Innov. 2015, 4, 192–208. [Google Scholar] [CrossRef] [PubMed]
  71. Yang, Z.; Yabansu, Y.C.; Al-Bahrani, R.; Liao, W.-k.; Choudhary, A.N.; Kalidindi, S.R.; Agrawal, A. Deep learning approaches for mining structure-property linkages in high contrast composites from simulation datasets. Comput. Mater. Sci. 2018, 151, 278–287. [Google Scholar] [CrossRef]
  72. Gu, G.X.; Chen, C.-T.; Richmond, D.J.; Buehler, M.J. Bioinspired hierarchical composite design using machine learning: Simulation, additive manufacturing, and experiment. Mater. Horiz. 2018, 5, 939–945. [Google Scholar] [CrossRef]
  73. Yu, C.-H.; Qin, Z.; Buehler, M.J. Artificial intelligence design algorithm for nanocomposites optimized for shear crack resistance. Nano Futures 2019, 3, 035001. [Google Scholar] [CrossRef]
  74. Chang, Y.-J.; Jui, C.-Y.; Lee, W.-J.; Yeh, A.-C. Prediction of the Composition and Hardness of High-Entropy Alloys by Machine Learning. JOM-J. Miner. Met. Mater. Soc. 2019, 71, 3433–3442. [Google Scholar] [CrossRef]
  75. Hsu, Y.-C.; Yu, C.-H.; Buehler, M.J. Using Deep Learning to Predict Fracture Patterns in Crystalline Solids. Matter 2020, 3, 197–211. [Google Scholar] [CrossRef]
  76. Wu, C.-T.; Chang, H.-T.; Wu, C.-Y.; Chen, S.-W.; Huang, S.-Y.; Huang, M.; Pan, Y.-T.; Bradbury, P.; Chou, J.; Yen, H.-W. Machine learning recommends affordable new Ti alloy with bone-like modulus. Mater. Today 2020, 34, 41–50. [Google Scholar] [CrossRef]
  77. Zhang, Y.; Wen, C.; Wang, C.; Antonov, S.; Xue, D.; Bai, Y.; Su, Y. Phase prediction in high entropy alloys with a rational selection of materials descriptors and machine learning models. Acta Mater. 2020, 185, 528–539. [Google Scholar] [CrossRef]
  78. Ansah, I.B.; Leming, M.; Lee, S.H.; Yang, J.-Y.; Mun, C.; Noh, K.; An, T.; Kim, S.L.D.-H.; Kim, M.; Im, H.; et al. Label-free detection and discrimination of respiratory pathogens based on electrochemical synthesis of biomaterials-mediated plasmonic composites and machine learning analysis. Biosens. Bioelectron. 2023, 227, 115178. [Google Scholar] [CrossRef]
  79. Jiang, D.; Xie, L.; Wang, L. Current application status of multi-scale simulation and machine learning in research on high-entropy alloys. J. Mater. Res. Technol. 2023, 26, 1341–1374. [Google Scholar] [CrossRef]
  80. Sun, M.; Dong, Z.; Wu, L.; Yao, H.; Niu, W.; Xu, D.; Chen, P.; Gupt, H.S.; Zhang, Y.; Dong, Y.; et al. Fast extraction of three-dimensional nanofiber orientation from WAXD patterns using machine learning. IUCrJ 2023, 10, 297–308. [Google Scholar] [CrossRef]
  81. Javaid, S.; Gorji, H.T.; Soulami, K.B.; Kaabouch, N. Identification and ranking biomaterials for bone scaffolds using machine learning and PROMETHEE. Res. Biomed. Eng. 2023, 39, 129–138. [Google Scholar] [CrossRef]
  82. Wang, Z.; Dabaja, R.; Chen, L.; Banu, M. Machine learning unifies flexibility and efficiency of spinodal structure generation for stochastic biomaterial design. Sci. Rep. 2023, 13, 5414. [Google Scholar] [CrossRef]
  83. Ward, L.; Agrawal, A.; Choudhary, A.; Wolverton, C. A general-purpose machine learning framework for predicting properties of inorganic materials. npj Comput. Mater. 2016, 2, 16028. [Google Scholar] [CrossRef]
  84. Glavatskikh, M.; Leguy, J.; Hunault, G.; Cauchy, T.; Da Mota, B. Dataset’s chemical diversity limits the generalizability of machine learning predictions. J. Cheminform. 2019, 11, 69. [Google Scholar] [CrossRef] [PubMed]
  85. Yang, K.; Xu, X.; Yang, B.; Cook, B.; Ramos, H.; Krishnan, N.M.A.; Smedskjaer, M.M.; Hoover, C.; Bauchy, M. Predicting the Young’s Modulus of Silicate Glasses using High-Throughput Molecular Dynamics Simulations and Machine Learning. Sci. Rep. 2019, 9, 8739. [Google Scholar] [CrossRef] [PubMed]
  86. Bishnoi, S.; Singh, S.; Ravinder, R.; Bauchy, M.; Gosvami, N.N.; Kodamana, H.; Krishnan, N.M.A. Predicting Young’s modulus of oxide glasses with sparse datasets using machine learning. J. Non-Cryst. Solids 2019, 524, 119643. [Google Scholar] [CrossRef]
  87. Buehler, M.J. Diatom-inspired architected materials using language-based deep learning: Perception, transformation and manufacturing. In Perspectives on the Mechanics of Fracture & Biological Materials; LuLu Publishing: Morrisville, NC, USA, 2023. [Google Scholar]
  88. Kopeliovich, M.V.; Petrushan, M.V.; Matukhno, A.E.; Lysenko, L.V. Towards detection of cancer biomarkers in human exhaled air by transfer-learning-powered analysis of odor-evoked calcium activity in rat olfactory bulb. Heliyon 2024, 10, e20173. [Google Scholar] [CrossRef] [PubMed]
  89. Ekinci, F.; Bölükdemir, M.H. The Effect of the Second Peak formed in Biomaterials used in a Slab Head Phantom on the Proton Bragg Peak. J. Polytechnıc 2020, 23, 129–136. [Google Scholar]
  90. Ekinci, F.; Acici, K.; Asuroglu, T.; Soylu, B.E. MC TRIM Algorithm in Mandibula Phantom in Helium Therapy. In Healthcare. MDPI, 2023; Volume 11, p. 2523.
  91. Fatih, E.; Erkan, B.; Serdar, G.M.; Özlem, D. Analysing the effect of a cranium thickness on a Bragg peak range in the proton therapy: A TRIM and GEANT4 based study. Научнo-технические ведoмoсти Санкт-Петербургскoгo гoсударственнoгo пoлитехническoгo университета. Физикo-математические науки 2022, 15, 64–78. [Google Scholar]
  92. Ekinci, F.; Asuroglu, T.; Acici, K. Monte Carlo Simulation of TRIM Algorithm in Ceramic Biomaterial in Proton Therapy. Materials 2023, 16, 4833. [Google Scholar] [CrossRef]
  93. Ekinci, F.; Bostanci, E.; Güzel, M.S.; Dagli, Ö. A Monte Carlo Study for Soft Tissue Equivalency of Potential Polymeric Biomaterials Used in Carbon Ion Radiation Therapy. Nucl. Technol. 2023, 209, 1229–1239. [Google Scholar] [CrossRef]
  94. Ekinci, F. Investigation of tissue equivalence of phantom biomaterials in 4He heavy ion therapy. In Radiation Effects and Defects in Solids; Taylor and Francis: Oxfordshire, UK, 2022. [Google Scholar]
  95. Demirel, G.; Orhan, A.I.; Irmak, O.; Aydın, F.; Büyüksungur, A.; Bilecenoğlu, B.; Orhan, K. Effects of preheating and sonic delivery techniques on the internal adaptation of bulk-fill resin composites. Oper. Dent. 2021, 46, 226–233. [Google Scholar] [CrossRef]
  96. Demirel, İ.; Yücel, H. Development of a flexible composite based on vulcanized silicon casting with bismuth oxide and characterization of its radiation shielding effectiveness in diagnostic X-ray energy range and medium gamma-ray energies. Nucl. Eng. Technol. 2024, 56, 2570–2575. [Google Scholar] [CrossRef]
  97. Yücel, H.; Safi, A. Investigation of the suitability of new developed epoxy based-phantom for child’s tissue equivalency in paediatric radiology. Nucl. Eng. Technol. 2021, 53, 4158–4165. [Google Scholar] [CrossRef]
  98. Buyuksungur, S.; Tanir, T.E.; Buyuksungur, A.; Bektas, E.I.; Kose, G.T.; Yucel, D.; Beyzadeoglu, T.; Cetinkaya, E.; Yenigun, C.; Tönük, E.; et al. 3D printed poly(ε-caprolactone) scaffolds modified with hydroxyapatite and poly(propylene fumarate) and their effects on the healing of rabbit femur defects. Biomater. Sci. 2017, 5, 2144–2158. [Google Scholar] [CrossRef] [PubMed]
  99. Ekinci, F.; Aşlar, E. Thyroid and contralateral breast surface dose variation in mammography: A phantom study on the role of breast tissue composition. Eur. Phys. J. Plus 2024, 139, 330. [Google Scholar] [CrossRef]
  100. Zhou, T.; Song, Z.; Sundmacher, K. Big Data Creates New Opportunities for Materials Research: A Review on Methods and Applications of Machine Learning for Materials Design. Engineering 2019, 5, 1017–1026. [Google Scholar] [CrossRef]
  101. Zhu, X.; Li, Y.; Gu, N. Application of Artificial Intelligence in the Exploration and Optimization of Biomedical Nanomaterials. Nano Biomed. Eng. 2023, 15, 342–353. [Google Scholar] [CrossRef]
  102. Gubskaya, A.V.; Kholodovych, V.; Knight, D.; Kohn, J.; Welsh, W.J. Prediction of fibrinogen adsorption for biodegradable polymers: Integration of molecular dynamics and surrogate modeling. Polymer 2007, 48, 5788–5801. [Google Scholar] [CrossRef] [PubMed]
  103. Miryala, B.; Zhen, Z.; Potta, T.; Breneman, C.M.; Rege, K. Parallel synthesis and quantitative structure–activity relationship (QSAR) modeling of aminoglycoside-derived lipopolymers for transgene expression. ACS Biomater. Sci. Eng. 2015, 1, 656–668. [Google Scholar] [CrossRef]
  104. Baudis, S.; Behl, M. High-throughput and combinatorial approaches for the development of multifunctional polymers. Macromol. Rapid Commun. 2022, 43, 2100400. [Google Scholar] [CrossRef]
  105. Jin, H.; Zhang, E.; Espinosa, H.D. Recent advances and applications of machine learning in experimental solid mechanics: A review. Appl. Mech. Rev. 2023, 75, 061001. [Google Scholar] [CrossRef]
  106. Yang, Z.; Yu, C.H.; Buehler, M.J. Deep learning model to predict complex stress and strain fields in hierarchical composites. Sci. Adv. 2021, 7, eabd7416. [Google Scholar] [CrossRef] [PubMed]
  107. Lwakatare, L.E.; Raj, A.; Crnkovic, I.; Bosch, J.; Olsson, H.H. Large-scale machine learning systems in real-world industrial settings: A review of challenges and solutions. Inf. Softw. Technol. 2020, 127, 106368. [Google Scholar] [CrossRef]
  108. Rudin, C. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat. Mach. Intell. 2019, 1, 206–215. [Google Scholar] [CrossRef] [PubMed]
  109. Jeon, W.; Ko, G.; Lee, J.; Lee, H.; Ha, D.; Ro, W.W. Deep learning with GPUs. In Advances in Computers; Elsevier: Amsterdam, The Netherlands, 2021; Volume 122, pp. 167–215. [Google Scholar]
  110. de Kanter, A.F.J.; Jongsma, K.R.; Verhaar, M.C.; Bredenoord, A.L. The ethical implications of tissue engineering for regenerative purposes: A systematic review. Tissue Eng. Part B Rev. 2023, 29, 167–187. [Google Scholar] [CrossRef] [PubMed]
  111. Hong, Y.; Hou, B.; Jiang, H.; Zhang, J. Machine learning and artificial neural network accelerated computational discoveries in materials science. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2020, 10, e1450. [Google Scholar] [CrossRef]
  112. Xue, K.; Wang, F.; Suwardi, A.; Han, M.-Y.; Teo, P.; Wang, P.; Wang, S.; Ye, E.; Li, Z.; Loh, X.J. Biomaterials by design: Harnessing data for future development. Mater. Today Bio 2021, 12, 100165. [Google Scholar] [CrossRef] [PubMed]
  113. Singh, A.V.; Rosenkranz, D.; Ansari, M.H.D.; Singh, R.; Kanase, A.; Singh, S.P.; Kanase, A.; Singh, S.P.; Johnston, B.D.; Tentschert, J.; et al. Artificial Intelligence and Machine Learning Empower Advanced Biomedical Material Design to Toxicity Prediction. Adv. Intell. Syst. 2020, 2, 2000084. [Google Scholar] [CrossRef]
  114. Basu, B.; Gowtham, N.H.; Xiao, Y.; Kalidindi, S.R.; Leong, K.W. Biomaterialomics: Data science-driven pathways to develop fourth-generation biomaterials. Acta Biomater. 2022, 143, 1–25. [Google Scholar] [CrossRef]
  115. Kasun, M.; Ryan, K.; Paik, J.; Lane-McKinley, K.; Dunn, L.B.; Roberts, L.W.; Kim, J.P. Academic machine learning researchers’ ethical perspectives on algorithm development for health care: A qualitative study. J. Am. Med. Inform. Assoc. 2024, 31, 563–573. [Google Scholar] [CrossRef]
  116. Rudin, C.; Chen, C.; Chen, Z.; Huang, H.; Semenova, L.; Zhong, C. Interpretable machine learning: Fundamental principles and 10 grand challenges. Stat. Surv. 2022, 16, 1–85. [Google Scholar]
Figure 1. Hierarchy between AI, ML, and DL.
Figure 1. Hierarchy between AI, ML, and DL.
Applsci 14 06590 g001
Figure 2. Supervised learning.
Figure 2. Supervised learning.
Applsci 14 06590 g002
Figure 3. Unsupervised learning.
Figure 3. Unsupervised learning.
Applsci 14 06590 g003
Figure 4. Basic DL architecture.
Figure 4. Basic DL architecture.
Applsci 14 06590 g004
Figure 5. Deep learning methods.
Figure 5. Deep learning methods.
Applsci 14 06590 g005
Figure 6. Types of biomaterials.
Figure 6. Types of biomaterials.
Applsci 14 06590 g006
Figure 7. A schematic of the iterative design loop utilizing ML and design of experiments (DOE) for the accelerated development of high-entropy alloys (HEAs). The figure is reproduced from [52].
Figure 7. A schematic of the iterative design loop utilizing ML and design of experiments (DOE) for the accelerated development of high-entropy alloys (HEAs). The figure is reproduced from [52].
Applsci 14 06590 g007
Figure 8. Diagram of the ML to design the novel polymer genome application. The figure is reproduced from [61].
Figure 8. Diagram of the ML to design the novel polymer genome application. The figure is reproduced from [61].
Applsci 14 06590 g008
Figure 10. Example architecture of a 3D convolutional neural network. The figure is reproduced from [71].
Figure 10. Example architecture of a 3D convolutional neural network. The figure is reproduced from [71].
Applsci 14 06590 g010
Figure 11. The diagram illustrates (a) the CNN training flowchart for predicting composite toughness and strength, (b) MSE convergence over epochs, and regression values for (c) toughness and (d) strength with blue (training) and red (testing) dots. The figure is reproduced from [73].
Figure 11. The diagram illustrates (a) the CNN training flowchart for predicting composite toughness and strength, (b) MSE convergence over epochs, and regression values for (c) toughness and (d) strength with blue (training) and red (testing) dots. The figure is reproduced from [73].
Applsci 14 06590 g011
Figure 12. Schematic diagram of machine-learning-assisted Ti alloy discovery. The figure is reproduced from [76].
Figure 12. Schematic diagram of machine-learning-assisted Ti alloy discovery. The figure is reproduced from [76].
Applsci 14 06590 g012
Figure 13. Scheme for CNN-based high-throughput screening for spinodal structures integrates two CNNs: one for rapid spinodal structure generation and another for ultrafast property prediction, enabling efficient generation and screening of spinodal structures with target anisotropic elasticity. The figure is reproduced from [82].
Figure 13. Scheme for CNN-based high-throughput screening for spinodal structures integrates two CNNs: one for rapid spinodal structure generation and another for ultrafast property prediction, enabling efficient generation and screening of spinodal structures with target anisotropic elasticity. The figure is reproduced from [82].
Applsci 14 06590 g013
Figure 14. Diagram of the ML applications for biomedical glass. The figure is reproduced from [41].
Figure 14. Diagram of the ML applications for biomedical glass. The figure is reproduced from [41].
Applsci 14 06590 g014
Figure 15. Diagram of the bio-hybrid method for detecting cancer biomarkers in exhaled human air. The figure is reproduced from [88].
Figure 15. Diagram of the bio-hybrid method for detecting cancer biomarkers in exhaled human air. The figure is reproduced from [88].
Applsci 14 06590 g015
Table 1. Search strategies in databases.
Table 1. Search strategies in databases.
DatabaseSearch Strategy
Google Scholar(“Artificial Intelligence (AI)” OR “Machine learning (ML)” OR “biomaterials” OR “Deep Learning (DL)” OR “Polymers” OR “Metals” OR “Composites” OR “Inorganic Materials”) AND PUBYEAR > 2002 AND PUBYEAR < 2025
Web of Science(“Artificial Intelligence (AI)”)) AND ALL = (“Machine learning (ML)”)) AND ALL = (“Deep Learning (DL)” OR “biomaterials” OR “Polymers” OR “Metals” OR “Composites” OR “Inorganic Materials”) AND PUBYEAR > 2002 AND PUBYEAR < 2025
Scopus(“Artificial Intelligence (AI)” AND “Machine learning (ML)” OR “biomaterials” OR “Polymers” OR “Metals” OR “Composites” OR “Inorganic Materials”) AND (“Deep Learning (DL)”) AND PUBYEAR > 2002 AND PUBYEAR < 2025)
ScienceDirect(“Artificial Intelligence (AI)” AND “Machine learning (ML)” OR “biomaterials” OR “Polymers” OR “Metals” OR “Composites” OR “Inorganic Materials”) AND (“Deep Learning (DL)”) AND PUBYEAR > 2003 AND PUBYEAR < 2025
ProQuest(“Artificial Intelligence (AI)” AND “Machine learning (ML)” OR “biomaterials” OR “Polymers” OR “Metals” OR “Composites” OR “Inorganic Materials”) AND (“Deep Learning (DL)”) AND PUBYEAR > 2002 AND PUBYEAR < 2025
Engineering Village(“Artificial Intelligence (AI)” AND “Machine learning (ML)” OR “biomaterials” OR “Polymers” OR “Metals” OR “Composites” OR “Inorganic Materials”) AND (“Deep Learning (DL)”) AND PUBYEAR > 2002 AND PUBYEAR < 2025
Table 2. AI and ML applications in metal materials.
Table 2. AI and ML applications in metal materials.
CategoriesApplications
Predictive modelingCorrosion resistance: Predicting corrosion behavior under different conditions.
Mechanical properties: Forecasting tensile strength, hardness, and ductility.
Material design and discoveryAlloy design: Discovering new alloys with desired properties.
Additive manufacturing: Optimizing parameters for metal 3D printing processes.
Process optimizationHeat treatment: Predicting outcomes to optimize time and temperature profiles.
Casting and forging: Minimizing defects and optimizing process parameters.
Failure analysis and preventionFatigue and fracture: Analyzing data to predict failure points.
Crack propagation: Modeling initiation and growth of cracks.
Surface engineeringCoating design: Designing advanced coatings to enhance properties like wear resistance.
Surface treatment: Optimizing treatments such as anodizing, plating, and chemical vapor deposition.
Subfields in biomaterialsBiomedical implants: Titanium alloys and stainless steel for orthopedic and dental implants.
Bioactive metals: Magnesium alloys for biodegradable implants; Nitinol for flexible medical devices.
Metallic scaffolds: Porous metals for tissue engineering applications.
Table 3. Application of AI and ML in the field of polymers.
Table 3. Application of AI and ML in the field of polymers.
CategoriesApplications
Predictive modelingMechanical properties: Predicting tensile strength, elasticity, and impact resistance of polymers.
Material design and discoveryPolymer synthesis: Discovering new polymer structures with desired properties.
Drug delivery systems: Designing polymer-based carriers for controlled drug release.
Process optimizationPolymerization processes: Optimizing conditions for polymer synthesis such as temperature, pressure, and catalysts.
Extrusion and molding: Improving process parameters to enhance the quality and performance of polymer products.
Failure analysis and preventionDegradation and aging: Predicting the lifespan and degradation behavior of polymers under various environmental conditions.
Fatigue resistance: Modeling the resistance of polymers to cyclic loading and stress.
Surface engineeringSurface modification: Designing surface treatments to improve properties like hydrophobicity, adhesion, and biocompatibility.
Coating applications: Developing advanced polymer coatings for protection and functionalization.
Subfields in biomaterialsTissue engineering scaffolds: Creating biodegradable polymer scaffolds for cell growth and tissue regeneration.
Biodegradable polymers: Designing environmentally friendly polymers for medical implants and drug delivery systems.
Smart polymers: Developing stimuli-responsive polymers for applications in drug delivery, tissue engineering, and diagnostics.
Table 4. Composite material subfields with ML and AI.
Table 4. Composite material subfields with ML and AI.
CategoriesApplications
Material design and discoveryOptimization of fiber–matrix combinations: Discovering new fiber and matrix combinations to improve properties like strength and stiffness.
Tailoring interfaces: Designing interfaces at the nano and microscale to enhance load transfer and durability.
Structural health monitoringReal-time monitoring: Using ML algorithms to monitor the structural integrity of composites in real time.
Damage prediction: Predicting potential failure points and remaining useful life of composite structures.
Manufacturing process optimizationAutomated lay-up: Optimizing automated lay-up processes for consistent and high-quality composite production.
Cure monitoring: Using AI to optimize cure cycles in autoclave and out-of-autoclave processes.
Performance enhancementLightweighting: Designing lightweight composites for aerospace and automotive applications without compromising strength.
Impact resistance: Developing composites with improved impact resistance for use in critical applications.
Recycling and sustainabilityRecyclability assessment: Assessing and improving the recyclability of composite materials.
Sustainable composites: Designing eco-friendly composites with biodegradable or recyclable matrices and fibers.
Biomedical applicationsComposite implants: Designing composites for load-bearing implants with tailored mechanical properties.
Biocompatible composites: Developing composites that are compatible with biological tissues for implants and prosthetics.
Table 5. AI and ML applications in inorganic materials.
Table 5. AI and ML applications in inorganic materials.
CategoriesApplications
Electronic propertiesBand gap prediction: Using ML to predict band gaps of various inorganic compounds, aiding in the design of semiconductors.
Electrical conductivity: Forecasting electrical conductivity for materials used in electronic devices and sensors.
Thermal managementThermal conductivity: Predicting thermal conductivity to optimize materials for heat dissipation in electronics and other high-temperature applications.
Thermal expansion: Modeling thermal expansion coefficients to prevent material failure due to temperature changes.
Material discoveryCrystal structure prediction: Identifying new stable crystal structures with desired electronic and mechanical properties.
Doping optimization: Using AI to determine optimal doping levels and types to enhance material properties like conductivity and magnetism.
Synthesis optimizationChemical vapor deposition (CVD): Optimizing CVD processes for high-quality thin films and coatings.
Sol–gel processes: Enhancing sol–gel synthesis parameters to produce materials with improved purity and properties.
Mechanical enhancementHardness and strength: Developing materials with superior hardness and tensile strength for industrial applications.
Fracture toughness: Improving fracture toughness to create more durable materials.
Environmental applicationsCatalytic converters: Designing materials for efficient catalytic converters and pollution control systems.
Photocatalysis: Developing materials for photocatalytic applications like water splitting and CO2 reduction.
Energy applicationsBattery components: Enhancing materials used in batteries, such as anodes, cathodes, and electrolytes, for better performance and longevity.
Photovoltaic materials: Optimizing materials for solar cell applications, including perovskites and other semiconductors.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Gokcekuyu, Y.; Ekinci, F.; Guzel, M.S.; Acici, K.; Aydin, S.; Asuroglu, T. Artificial Intelligence in Biomaterials: A Comprehensive Review. Appl. Sci. 2024, 14, 6590. https://doi.org/10.3390/app14156590

AMA Style

Gokcekuyu Y, Ekinci F, Guzel MS, Acici K, Aydin S, Asuroglu T. Artificial Intelligence in Biomaterials: A Comprehensive Review. Applied Sciences. 2024; 14(15):6590. https://doi.org/10.3390/app14156590

Chicago/Turabian Style

Gokcekuyu, Yasemin, Fatih Ekinci, Mehmet Serdar Guzel, Koray Acici, Sahin Aydin, and Tunc Asuroglu. 2024. "Artificial Intelligence in Biomaterials: A Comprehensive Review" Applied Sciences 14, no. 15: 6590. https://doi.org/10.3390/app14156590

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop