Artificial Intelligence and Machine Learning-Based Medical Image Analysis

Paranasal sinus pathologies, particularly those affecting the maxillary sinuses, pose significant challenges in diagnosis and treatment due to the complex anatomical structures and diverse disease manifestations. The aim of this study is to investigate the use of deep learning techniques, particularly generative adversarial networks (GANs), in combination with convolutional neural networks (CNNs), for the classification of sinus pathologies in medical imaging data. The dataset is composed of images obtained through computed tomography (CT) scans, covering cases classified into “Moderate”, “Severe”, and “Normal” classes. The lightweight GAN is applied to augment a dataset by creating synthetic images, which are then used to train and test the ResNet-50 and ResNeXt-50 models. The model performance is optimized using random search to perform hyperparameter tuning, and the evaluation is conducted extensively for various metrics like accuracy, precision, recall, and the F1-score. The results demonstrate the effectiveness of the proposed approach in accurately classifying sinus pathologies, with the ResNeXt-50 model achieving superior performance with accuracy: 91.154, precision: 0.917, recall: 0.912, and F1-score: 0.913 compared to ResNet-50. This study highlights the potential of GAN-based data augmentation and deep learning techniques in enhancing the diagnosis of maxillary sinus diseases. Full article

Deep learning-based segmentation models have made a profound impact on medical procedures, with U-Net based computed tomography (CT) segmentation models exhibiting remarkable performance. Yet, even with these advances, these models are found to be vulnerable to adversarial attacks, a problem that equally affects automatic CT segmentation models. Conventional adversarial attacks typically rely on adding noise or perturbations, leading to a compromise between the success rate of the attack and its perceptibility. In this study, we challenge this paradigm and introduce a novel generation of adversarial attacks aimed at deceiving both the target segmentation model and medical practitioners. Our approach aims to deceive a target model by altering the texture statistics of an organ while retaining its shape. We employ a real-time style transfer method, known as the texture reformer, which uses adaptive instance normalization (AdaIN) to change the statistics of an image’s feature.To induce transformation, we modify the AdaIN, which typically aligns the source and target image statistics. Through rigorous experiments, we demonstrate the effectiveness of our approach. Our adversarial samples successfully pass as realistic in blind tests conducted with physicians, surpassing the effectiveness of contemporary techniques. This innovative methodology not only offers a robust tool for benchmarking and validating automated CT segmentation systems but also serves as a potent mechanism for data augmentation, thereby enhancing model generalization. This dual capability significantly bolsters advancements in the field of deep learning-based medical and healthcare segmentation models. Full article

Multimodal image registration is an important component of medical image processing, allowing the integration of complementary information from various imaging modalities to improve clinical applications like diagnosis and treatment planning. We proposed a novel multistage neural network for three-dimensional multimodal medical image registration, which addresses the challenge of larger rigid deformations commonly present in medical images due to variations in patient positioning in different scanners and rigid anatomical structures. This multistage network combines rigid, affine and deformable transformations in three stages. The network was trained unsupervised with Mutual Information and Gradient L2 loss. We compared the results of our proposed multistage network with a rigid-affine-deformable registration with the classical registration method NiftyReg as a baseline and a multistage network, which combines affine and deformable transformation, as a benchmark. To evaluate the performance of the proposed multistage network, we used four three-dimensional multimodal in vivo datasets: three renal MR datasets consisting of T1-weighted and T2-weighted MR scans and one liver dataset containing CT and T1-weighted MR scans. Experimental results showed that combining rigid, affine and deformable transformations in a multistage network leads to registration results with a high structural similarity, overlap of the corresponding structures (Dice: 76.7 ± 12.5, 61.1 ± 14.0, 64.8 ± 16.2, 68.1 ± 24.6 for the four datasets) and a low level of image folding (|J| ≤ 0: less than or equal to 1.1%), resulting in a medical plausible registration result. Full article

Brain tumors pose significant health challenges worldwide, with glioblastoma being one of the most aggressive forms. The accurate determination of the O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status is crucial for personalized treatment strategies. However, traditional methods are labor-intensive and time-consuming. This paper proposes a novel multi-modal approach, BTDNet, that leverages multi-parametric MRI scans, including FLAIR, T1w, T1wCE, and T2 3D volumes, to predict the MGMT promoter methylation status. BTDNet’s main contribution involves addressing two main challenges: the variable volume lengths (i.e., each volume consists of a different number of slices) and the volume-level annotations (i.e., the whole 3D volume is annotated and not the independent slices that it consists of). BTDNet consists of four components: (i) data augmentation (which performs geometric transformations, convex combinations of data pairs, and test-time data augmentation); (ii) 3D analysis (which performs global analysis through a CNN-RNN); (iii) routing (which contains a mask layer that handles variable input feature lengths); and (iv) modality fusion (which effectively enhances data representation, reduces ambiguities, and mitigates data scarcity). The proposed method outperformed state-of-the-art methods in the RSNA-ASNR-MICCAI BraTS 2021 Challenge by at least 3.3% in terms of the F1 score, offering a promising avenue for enhancing brain tumor diagnosis and treatment. Full article

Background: The Sella Turcica is a critical structure from an orthodontic perspective, and its morphological characteristics can help in understanding craniofacial deformities. However, accurately extracting Sella Turcica shapes can be challenging due to the indistinct edges and indefinite boundaries present in X-ray images. This study aimed to develop and validate an automated Sella Morphology Network (SellaMorph-Net), a hybrid deep learning pipeline for segmenting Sella Turcica structure and extracting different morphological types; Methods: The SellaMorph-Net model proposed in this study combined attention-gating and recurrent residual convolutional layers (AGM and RrCL) to enhance the encoder’s abilities. The model’s output was then passed through a squeeze-and-excitation (SE) module to improve the network’s robustness. In addition, dropout layers were added to the end of each convolution block to prevent overfitting. A Zero-shot classifier was employed for multiple classifications, and the model’s output layer used five colour codes to represent different morphological types. The model’s performance was evaluated using various quantitative metrics, such as global accuracy and mean pixel-wise Intersection over Union (IoU) and dice coefficient, based on qualitative results; Results: The study collected 1653 radiographic images and categorised them into four classes based on the predefined shape of Sella Turcica. These classes were further divided into three subgroups based on the complexity of the Sella structures. The proposed SellaMorph-Net model achieved a global accuracy of 97.570, mean pixel-wise IoU scores of 0.7129, and a dice coefficient of 0.7324, significantly outperforming the VGG-19 and InceptionV3 models. The publicly available IEEE ISBI 2015 challenge dataset and our dataset were used to evaluate the test performance between the state-of-the-art and proposed models. The proposed model provided higher testing results, which were 0.7314 IoU and 0.7768 dice for our dataset and 0.7864 IoU and 0.8313 dice for the challenge dataset; Conclusions: The proposed hybrid SellaMorph-Net model provides an accurate and reliable pipeline for detecting morphological types of Sella Turcica using full lateral cephalometric images. Future work will focus on further improvement and utilisation of the developed model as a prognostic tool for predicting anomalies related to Sella structures. Full article

To prevent pedicle screw implant failure, a diagnostic technique that allows surgeons to evaluate implant stability easily, quickly, and quantitatively in clinical orthopedic situations is required. This study aimed to predict the insertion torque equivalent to laboratory-level evaluation accuracy. This serves as an index of the implant stability of pedicle screws placed in cadaveric bone, which relies on laser resonance frequency analyses (L-RFA) when irradiating with two types of lasers. The machine learning analysis was optimized using a dataset with artificial bone as teaching data. In this analysis, many explanatory variables extracted from the laser-induced vibration spectra obtained during an analysis/RFA evaluation were predicted by selecting important variables using the least absolute shrinkage and selection operator and performing a non-linear approximation using support vector regression. It was found that combining both artificial and cadaveric bone data with the bone densities as teaching data dramatically improved the determination coefficient from R² = −0.144 to R² = 0.858 as the prediction accuracy and reduced the influence of differences between artificial and cadaveric bones. This technology will contribute to the development of preventive diagnostic technologies that can be used during surgery, which is necessary in order to further advance treatment technologies. Full article

Alzheimer’s is a neurodegenerative disorder affecting the central nervous system and cognitive processes, explicitly impairing detailed mental analysis. Throughout this condition, the affected individual’s cognitive abilities to process and analyze information gradually deteriorate, resulting in mental decline. In recent years, there has been a notable increase in endeavors aimed at identifying Alzheimer’s disease and addressing its progression. Research studies have demonstrated the significant involvement of genetic factors, stress, and nutrition in developing this condition. The utilization of computer-aided analysis models based on machine learning and artificial intelligence has the potential to significantly enhance the exploration of various neuroimaging methods and non-image biomarkers. This study conducts a comparative assessment of more than 80 publications that have been published since 2017. Alzheimer’s disease detection is facilitated by utilizing fundamental machine learning architectures such as support vector machines, decision trees, and ensemble models. Furthermore, around 50 papers that utilized a specific architectural or design approach concerning Alzheimer’s disease were examined. The body of literature under consideration has been categorized and elucidated through the utilization of data-related, methodology-related, and medical-fostering components to illustrate the underlying challenges. The conclusion section of our study encompasses a discussion of prospective avenues for further investigation and furnishes recommendations for future research activities on the diagnosis of Alzheimer’s disease. Full article

The stiffness of tumor cells has a significant influence on invasion and metastasis strategies. In this study, we developed a novel detection method, called laser resonance frequency analysis (L-RFA), for soft tissue tumors in physical oncology. In addition, we evaluated the characteristics of the laser-induced photo-thermal elastic wave (LIPTEW) obtained by L-RFA using agarose gels with different stiffnesses to simulate soft tissues. The LIPTEW diagnosis based on the audible wave range indicated a great potential too, which allows for the measurement of the stiffness of single cells while maintaining organ geometry. In particular, we observed vibrations with high spatial resolution of less than one-tenth of the laser irradiation spot size. From the obtained results, our proposed machine learning method achieved high accuracy and precision, with coefficient of determination R² = 0.950. The characterization of the LIPTEW on the L-RFA to predict single cell stiffness could be a milestone for future studies on physical oncology, soft-tissue tumor stiffness diagnoses, and medical imaging technologies. Full article