Next Article in Journal
Anatomical Variants of the Origin of the Coronary Arteries: A Systematic Review and Meta-Analysis of Prevalence
Next Article in Special Issue
Early Detection of Heart Failure with Autonomous AI-Based Model Using Chest Radiographs: A Multicenter Study
Previous Article in Journal
Combined Rotating Ultra-High-Resolution Spectral Domain OCT and Scheimpflug Imaging for In Vivo Corneal Optical Biopsy
Previous Article in Special Issue
Evaluation of the Performance of an Artificial Intelligence (AI) Algorithm in Detecting Thoracic Pathologies on Chest Radiographs
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

Evolving and Novel Applications of Artificial Intelligence in Thoracic Imaging

1
Department of Radiology, The Ohio State University College of Medicine, Columbus, OH 43210, USA
2
Division of Vascular and Interventional Radiology, Department of Radiology, The Ohio State University Wexner Medical Center, Columbus, OH 43210, USA
*
Author to whom correspondence should be addressed.
Diagnostics 2024, 14(13), 1456; https://doi.org/10.3390/diagnostics14131456
Submission received: 30 May 2024 / Revised: 1 July 2024 / Accepted: 6 July 2024 / Published: 8 July 2024
(This article belongs to the Special Issue Applications of Artificial Intelligence in Thoracic Imaging)

Abstract

:
The advent of artificial intelligence (AI) is revolutionizing medicine, particularly radiology. With the development of newer models, AI applications are demonstrating improved performance and versatile utility in the clinical setting. Thoracic imaging is an area of profound interest, given the prevalence of chest imaging and the significant health implications of thoracic diseases. This review aims to highlight the promising applications of AI within thoracic imaging. It examines the role of AI, including its contributions to improving diagnostic evaluation and interpretation, enhancing workflow, and aiding in invasive procedures. Next, it further highlights the current challenges and limitations faced by AI, such as the necessity of ‘big data’, ethical and legal considerations, and bias in representation. Lastly, it explores the potential directions for the application of AI in thoracic radiology.

1. Introduction

Radiology is playing an increasingly pivotal role in healthcare, with the utilization of medical imaging consistently rising each year [1,2]. In the field of thoracic radiology, chest radiography (CXR) remains one of the most frequently performed imaging studies in medicine [3]. Moreover, the expanding implementation of national lung cancer screening programs is expected to substantially increase the utilization of chest computed tomography (CT) [4]. This escalating demand is placing a significant burden on the workload of radiologists.
Artificial intelligence (AI) has emerged as a promising solution to address many of the current challenges confronting radiology. AI has undergone substantial advancements, from machine learning (ML), which enables algorithms to learn from data without explicit programming, to the more recent deep learning (DL)—a subcategory of machine learning that utilizes layers of neural networks [5]. These techniques are now beginning to influence all parts of modern society, including the fields of medicine and radiology. With the clinical significance of thoracic diseases and the availability of abundant data, AI-based tools are actively being developed for thoracic imaging applications [3]. The United States Food and Drug Administration (FDA) has cleared more than 80 AI products for various indications within thoracic radiology alone [6]. As these devices demonstrate their efficacy beyond controlled test settings and transition into clinical environments, understanding how AI is currently being applied has become crucial. This review aims to provide an overview of the significant ways in which AI is reshaping the practice of thoracic radiology, while also discussing limitations and exploring the future directions.

2. Applications of Artificial Intelligence in Thoracic Imaging

Summary of AI applications in thoracic imaging is presented in Figure 1.

2.1. Worklist Prioritization

AI was first used in thoracic radiology to detect CXRs with abnormalities and to prioritize them at the top of the reading worklist [7]. AI-based triage has shown promise in improving radiologist turnaround times, facilitating more timely diagnoses and management of urgent cases. Nam et al. developed a DL algorithm capable of accurately detecting 10 common abnormalities on CXR, with an area under the receiver operating characteristic curve (AUROC) ranging from 0.895 to 1.00 during external validation [8]. Prioritizing the picture archiving and communication system (PACS) worklist based on the algorithm’s results significantly reduced the time-to-report for critical and urgent cases compared to scenarios without the algorithm: 640.5 versus 3371.0 s and 1840.3 versus 2127.1 s, respectively (p < 0.01). Additionally, the interpretation time per CXR in AI-aided readings was significantly shorter compared to conventional reading sessions: 20.5 versus 23.5 s. Similarly, Baltruschat et al. introduced an AI-based worklist prioritization tool, which notably decreased the report turnaround time for all critical findings to 35–50 min, down from 80 min when employing a standard first-in, first-out processing approach [9]. In a simulation study, Annarumma et al. also demonstrated that a triaging system utilizing a DL algorithm reduced the median delay for critical examinations from 7.2 h to 43 min [10].
In addition to CXR, AI is also being explored for use as a triage tool in CT scans. Huang et al. developed a DL-based model capable of detecting pulmonary embolism in computed tomography pulmonary angiography (CTPA) scans, achieving an AUROC of 0.84 on the internal test set and 0.85 on an external dataset [11]. Furthermore, Hata et al. created a DL-based algorithm that identifies aortic dissection on non-contrast-enhanced CT images, achieving an area under the curve (AUC) of 0.940, with accuracy, sensitivity, and specificity comparable to those of radiologists [12]. As AI models begin to integrate into clinical setting as prioritization tools for various conditions, they will assist in identifying clinically important cases, thereby facilitating prompt diagnoses and treatments [13].

2.2. Diagnostic Accuracy

2.2.1. Chest Radiography

A major application of AI in radiology is to enhance diagnostic accuracy in medical imaging. As a result, a significant portion of research of AI in thoracic imaging focuses on this topic [7]. While conventional computer-aided detection (CAD) has been used for decades, recent advancements in DL have significantly enhanced the capabilities of AI-based tools [14].
An effective AI model for lung nodules must not only accurately detect them, as they can be indicative of lung malignancies, but also minimize false positive results. Nam et al. developed a DL-based algorithm for detecting malignant pulmonary nodules on CXR [15]. The model achieved an AUROC between 0.92–0.99, outperforming 17 out of 18 physicians, while also exhibiting false positives rates ranging 0.02 to 0.34 per image. Additionally, all physicians showed improved nodule detection with the assistance of the DL algorithm. In a more recent study, Nam et al. demonstrated that the utilization of AI-based CAD significantly enhanced the detection of actionable (0.59% versus 0.25%, p = 0.008) and malignant nodules (0.15% versus 0.0%, p = 0.008) compared to not using AI in real-world health checkup participants [16].
In addition to detecting lung nodules, AI-based algorithms have also been developed for identifying other pulmonary abnormalities. Rajpurkar et al. developed a DL algorithm that was able to detect 14 different pathologies in CXR [17]. Compared to radiologists, the model achieved comparable level of performance on 11 pathologies, while radiologists achieved higher AUC on cardiomegaly, emphysema, and hiatal hernia. Similarly, Hwang et al. devised a DL algorithm capable of discerning between normal and abnormal CXR images, specifically targeting four major thoracic diseases [18]. The algorithm achieved a median AUC of 0.979, statistically outperforming non-radiology physicians, radiologists, and thoracic radiologists. Furthermore, the study demonstrated that physicians’ performance improved with the assistance of the DL algorithm. The development of newer deep learning AI algorithms may further enhance the accuracy of AI models in detecting pulmonary pathologies [19]. However, one limitation of DL-based CAD is that many studies have relied on isolated findings on CXRs, which may not accurately represent the multiple abnormalities that can present in real-world images [7].
The emergency department (ED) is an area where AI-based tools can provide crucial value in enhancing patient care. Hwang et al. demonstrated that a commercialized DL algorithm achieved an AUC of 0.95 in interpreting patient CXRs in the ED, showing higher sensitivity but lower specificity compared to radiology residents [20]. For the residents, the use of the algorithm resulted in an improvement in sensitivity from 65.6% to 73.4% (p = 0.003), albeit with a slight decrease in specificity from 98.1% to 94.3% (p < 0.001). Additionally, Huang et al. developed an AI model that generated reports based on CXRs from the ED and compared them with reports from a radiologist and teleradiologist [21]. When evaluated by six ED physicians using a Likert score, the reports generated by the AI model showed similar clinical accuracy and textual quality to that of radiologist reports, while demonstrating improved textual quality compared to teleradiologist reports.
In addition to pulmonary diseases, CXR can also detect non-pulmonary conditions. Pyrros et al. developed a DL model that identified patients with type 2 diabetes mellitus by integrating data from both CXR and electronic health records [22]. By analyzing adiposity features, the model achieved a diabetes detection AUC of 0.84 at a prevalence of 16%. Furthermore, the diagnostic utility of CXR in cardiovascular diseases such as heart failure, valvular heart disease, pulmonary hypertension, coronary artery disease, and atrial fibrillation is actively being studied [23,24].
Beyond improving diagnostic accuracy, studies are demonstrating that AI-based tools, when used as a second reader, also have the potential to enhance reader confidence and reduce reading time [25,26]. For instance, Ahn et al. reported a 10% reduction in reporting time for chest radiographs when using a commercially available AI engine compared to without AI (36.9 versus 40.8 s) [25]. Moreover, the implementation of AI can aid in efficiency by triaging and excluding normal chest X-rays, thereby reducing the workload for radiologists. In a stimulation study, Yoo et al. found that commercial AI triaging, which sorts and removes normal CXRs, led to a 40% decrease in workload [27]. Similarly, in an external evaluation, Plesner et al. found that a commercially available AI could autonomously report 28% of normal posteroanterior CXRs with a sensitivity of greater than 99% for detecting any abnormalities [28].

2.2.2. Computed Tomography

The National Lung Screening Trial (NLST) and the Nelson trial have demonstrated the mortality-reducing benefits of low-dose chest CT screening for lung cancer in high-risk patients [29,30]. Consequently, the US Preventative Services Task Force (USPSTF) recommends annual screening for lung cancer with low-dose CT in adults who are at high risk [31].
DL algorithms have demonstrated high accuracy in detecting lung nodules, ranging from 82.2% to 97.6%, with corresponding AUC values of 0.87 to 0.98 [32]. Moreover, a systematic review by Ewals et al. demonstrated that with AI assistance, radiologists showed higher sensitivity and AUC in detecting lung nodules, although with slightly lower specificity [33]. Additionally, not limited to AI tools that directly identify abnormalities, AI-based image reconstruction model has shown to increase nodule detection rates, improve inter-reader agreement, and reduce reading time [34].
Furthermore, AI tools are capable of not only detecting lung nodules but also classifying them as benign or malignant. Ciompi et al. developed a DL system that classified nodules into four different types with an accuracy comparable to that of six physicians (69.6% versus 72.9%, respectively) [35]. Similarly, Ardila et al. achieved an AUC of 0.944 in predicting the risk of lung cancer with their DL algorithm [36]. The model, when no prior CT image was available, outperformed radiologists with an absolute reduction of 11% in false positives and 5% in false negatives. Recent advancements in AI models provide opportunities to expedite the diagnosis of malignant nodules, reduce unnecessary workup for benign nodules, and minimize inter-grader variability [37]. Moreover, Adams et al. demonstrated that integrating the AI risk score and Lung CT Reporting and Data System (Lung-RADS) in lung nodule management resulted in fewer follow-ups and reduced healthcare costs [38]. Additionally, a DL model by Mikhael et al. could predict lung cancer risk over the next 6 years with an AUROC of 0.75–0.81 based solely on a single low-dose CT scan, offering opportunities for early intervention and personalized patient care [39].
In addition to detecting and classifying lung nodules, AI tools can also aid in diagnosing other pulmonary diseases on CT scans. As previously described, DL-based AI models demonstrate high sensitivity and negative predictive value in detecting PE on CTPA scans [11,40]. Walsh et al. developed a DL algorithm for classifying fibrotic lung disease on CT images using existing diagnostic guidelines [41]. The model achieved an accuracy of 76.4% and 73.3% on different test datasets, outperforming 66% of thoracic radiologists. Similarly, Christe et al. developed a ML-based CAD system that classified idiopathic pulmonary fibrosis with accuracy comparable to two radiologists [42]. Furthermore, a DL tool developed by González et al. could detect chronic obstructive pulmonary disease (COPD) in CT scans with a C statistic of 0.856 [43]. The tool achieved accuracy rates of 51.1% and 29.4% for staging COPD in two different participant groups, as well as predicting the risk for acute respiratory disease events and mortality. Expanding beyond pulmonary pathologies, AI algorithms have been studied to assess cardiovascular health by evaluating coronary calcium score and predicting cardiovascular disease risk [44].

2.3. Quantitative Analysis

Quantitative analysis has emerged as a promising approach for identifying biomarkers associated with various pulmonary pathologies [45,46]. However, its widespread adoption in routine clinical practice has been hindered by the time-consuming nature of manual tasks [3]. Fortunately, advancements in AI technologies have enabled automatic segmentation of chest CT images with high accuracy and efficiency. For instance, Chassagnon et al. developed a DL algorithm capable of automatically assessing the extent of systemic sclerosis-related interstitial lung disease (ILD) on chest CT images [47]. The disease extent contouring by the algorithm was comparable to that of radiologists, with a median Dice similarity coefficient (DSC) ranging from 0.74 to 0.75, compared to a median DSC of 0.68 to 0.71 observed between radiologists. Additionally, the disease extent calculated by the algorithm correlated with the results of pulmonary function testing (PFT). Hasenstab et al. developed an automated DL algorithm for COPD staging by quantifying emphysema and air trapping [48]. The algorithm’s staging of COPD severity was matched spirometry-based Global Initiative for Chronic Obstructive Lung Disease (GOLD) staging, achieving an AUC of 0.86–0.96. Moreover, it successfully predicted disease progression (odds ratio, 1.50–2.67; p < 0.001) and mortality (hazard ratio, 2.23; p < 0.001). In addition to ILD and COPD, AI-based quantification analysis in segmentation has been studied for pulmonary lobes, airways, and lung nodules [49,50,51,52]. Furthermore, AI models can be employed to efficiently measure bronchial parameters, including in patients with cystic fibrosis [53,54]. Looking ahead, fully automated AI systems may play an essential role in preventive medicine and ‘opportunistic CT screening’ for patients [55].
The variability in patients’ characteristics, scanner types, acquisition techniques, and reconstruction protocols presents a challenge to the implementation of AI in clinical settings [14,56]. These variations impede the reproducibility and generalizability of quantitative image analysis, including radiomics [57]. Recent studies have focused on how AI models can reconstruct images to standardize quantitative analysis. Lee et al. developed a convolutional neural network (CNN)-based DL model capable of converting one CT kernel to another [58]. The converted images demonstrated a reduction in root mean square error (RMSE) compared to original images, with a mean reduction of 65.7%. Similar CT conversions models have also been developed for pulmonary nodules and ILD [59,60]. With the advent of newer models, the reduction in variability between images will facilitate AI research in quantitative imaging.

2.4. Radiomics

Radiomics is a field in radiology that aims to extract quantitative features from an image using algorithms [61]. Many of these features are invisible to the human eye and have been investigated as potential biomarkers, especially in thoracic oncology. Beig et al. utilized perinodular and intranodular radiomic features on noncontrast lung CT scans to differentiate between non-small cell lung cancer adenocarcinomas and benign granulomas [62]. Combining radiomic features in the analysis led to an increase in the AUROC from 0.75 to 0.80. Hawkins et al. found that radiomics applied to low-dose CT images can predict which nodules will become cancerous within 1 to 2 years [63]. The best-performing model was able to predict the emergence of cancer with accuracies of 80% (AUC 0.83) and 79% (AUC 0.75) at 1 and 2 years, respectively.
Beyond detecting lung cancer, radiomics are demonstrating their utility in characterizing and stratifying lung cancers for prognosis [64]. Fan et al. identified radiomic features that could discriminate between invasive adenocarcinoma and non-invasive lesions [65]. The radiomic signature achieved an accuracy ranging from 86.3% to 90.8% in the primary and validation cohorts. Similarly, Cherezov et al. demonstrated that a combination of heterogeneity and radiomic features discriminated between cancerous and non-cancerous nodules, achieving an AUROC of 0.85 and an accuracy of 81.64% [66]. Moreover, their study found that the analysis could also distinguish long- and short-term survivors of lung adenocarcinoma with an AUROC of 0.9 and an accuracy of 85%. Additionally, Wu et al. demonstrated how radiomics can be used as a non-invasive means to determine lung cancer histology by differentiating adenocarcinoma and squamous cell carcinoma, achieving an AUC of 0.72 [67].
Recent studies have investigated the association between radiomics and genomics, termed radiogenomics, to identify specific mutations and guiding treatment decisions [68]. Jia et al. demonstrated the identification of epidermal growth factor receptor (EGFR) mutation in lung adenocarcinoma using radiomic features, achieving an AUC of 0.802 [69]. Incorporating sex and smoking history further improved the model’s performance, with sensitivity and specificity reaching 60.6% and 85.1%, respectively. Rizzo et al. identified specific qualitative CT associations corresponding to various mutations in non-small cell lung cancer (NSCLC), including EGFR, Kirsten rat sarcoma viral oncogene homolog (K-RAS), and anaplastic lymphoma kinase gene (ALK) mutations [70]. Additionally, Trebeschi et al. demonstrated the potential of radiomics features as biomarkers for predicting the response to immunotherapy in NSCLC, achieving an AUC of 0.83. Their study found that the radiomic biomarkers were involved in cell cycle progression and mitosis, offering insights into the mechanism for the preferential response. By integrating genomic and phenotypic information found in imaging data, AI radiomics will expand our knowledge of tumor biology and promote precision medicine [71].

2.5. Radiation and Contrast Reduction

Image reconstruction techniques have been studied with the goal of improving image quality while reducing the radiation dose in CT scans. Iterative reconstruction (IR) algorithms have been introduced to replace conventional filtered back projection, with the goal of reducing image noise and optimizing radiation dose [72,73]. However, IR has its limitations, including over-smoothing and the production of altered image textures [14,73]. Recently, DL-based image reconstruction methods have demonstrated improvements in image denoising, texture improvement, and artifact reduction compared to conventional models at lower doses [74,75,76,77]. AI image reconstruction algorithms offer the promise of ultra-low-dose CT (0.07 mSv or 0.14 mSv) and potentially even photon-counting CT technology in the future [37,78,79]. Moreover, DL techniques can generate synthetic CT images from MRI data and enhance image quality by reducing metal artifacts [80,81]. Furthermore, DL-based grid-like algorithm has shown improvements in both image quality and dose reduction in CXR [82].
Iodinated contrast media, commonly used in CT imaging for diagnostic purposes, carry risks of allergic reactions and nephrotoxicity [83]. Recent advancements in AI now enable the creation of synthetic post-contrast images from non-contrast or low-dose post-contrast CT scans [84]. For instance, Choi et al. utilized a DL model to generate synthetic contrast-enhanced CT images from non-contrast chest CT scans [85]. These synthetic images significantly improved the detection of mediastinal lymph nodes, demonstrating a higher contrast-to-noise ratio (6.15 versus 0.74, p < 0.001) and enhanced lesion conspicuity (p ≤  0.001) compared to non-contrast CT scans. Additionally, Chun et al. produced a synthetic contrast-enhanced CT from non-contrast CT scans using DL, facilitating the delineation of cardiac substructures following breast radiation therapy [86].

2.6. Operational Efficiency

Not limited to diagnostic applications, AI can also be applied in operational tasks within radiology. From AI-assisted clinical decision support for imaging orders to improving automated speech recognition for report generation, AI has the capacity to aid in various aspects of the radiology workflow [87,88]. For instance, Chong et al. decreased the no-show rate for outpatient MRI from 19.3% to 15.9% by using an ML-based model to identify patients at high-risk of not showing up and implementing a telephone reminder system [89].
One notable AI application in thoracic imaging is its use in the follow-up of lung nodules. Pannu et al. showed that the rate of appropriate follow-up for incidental lung nodules remains low, at only 43.2%, representing a missed opportunity for early lung cancer detection [90]. To address this issue, natural language processing (NLP) is being explored as a means to efficiently identify reports mentioning pulmonary nodules and flag them for appropriate follow-up [91]. Desai et al. implemented an NLP algorithm that not only increased the rate of follow-ups completed on time from 64.5% to 84.3% but also achieved 87% agreement between primary care providers and radiologists on the follow-up plan [92]. Similarly, Abbasi et al. developed an AI model for identifying radiology reports containing recommendations for additional imaging, with the external validation set demonstrating 99% accuracy and an F1 score of 95.2% [93]. As the use of AI in non-interpretative areas continues to expand and more tools are integrated, AI is anticipated to play an increasing role in the radiology workflow.

2.7. COVID-19

The COVID-19 pandemic accelerated research into AI applications in thoracic imaging, which can be crucial for detecting COVID-19, assessing disease severity, and predicting prognosis [94]. While reverse transcriptase-polymerase chain reaction (RT-PCR) remains the gold standard for diagnosing COVID-19, challenges such as kit availability, reliability, and turnaround time persist [95,96]. In a systematic review, Guidgar et al. found that AI applied to various chest imaging modalities, including CXR, CT, and ultrasound, aids in the early diagnosis of COVID-19 [97]. Additionally, AI applications in chest CT facilitate the quantification and monitoring of COVID-19 progression [98,99]. Furthermore, AI in chest imaging provides prognostic information, including predicting the development of acute respiratory distress syndrome (ARDS), the need for oxygen therapy or intubation, and hospital stay duration or adverse outcomes [100,101,102,103,104]. In a meta-analysis conducted by Jia et al., AI models were found to distinguish COVID-19 from other pneumonias with a pooled AUC, sensitivity, and specificity of 0.96, 0.92, and 0.91, respectively [105].
Nevertheless, there remain limitations in the practical application of AI tools within the current clinical workflow. A systematic review by Roberts et al. revealed methodological flaws or a risk of underlying bias in machine learning models for detecting and prognosticating COVID-19 [106]. Additionally, these models require validation through prospective real-time studies. Sun et al. evaluated the real-time performance of an interpretable AI model for detecting COVID-19 on CXR, finding that the model had an AUC of 0.70 and underperformed compared to radiologists [107]. Furthermore, it is essential to address issues regarding the generalizability and applicability of AI models [108]. Continued research into AI for COVID-19 will be crucial in establishing the groundwork for preparing and managing potential resurgences of COVID-19 or other public health outbreaks in the future.

2.8. Tuberculosis Screening and Triage

Although tuberculosis (TB) is relatively rare in the United States, it was the leading global cause of death from a single infectious pathogen before COVID-19. In 2019, an estimated 10 million people developed active TB, resulting in 1.4 million deaths, with developing countries in Africa and Southeast Asia being disproportionately affected [109]. The World Health Organization (WHO) endorses the use of CXR both as a triage and as a screening tool for individuals at high risk of TB [110]. However, despite the high sensitivity and specificity of CXR for detecting pulmonary TB, its utility is limited by the availability of trained radiologists, particularly in countries with high prevalence [14,111].
AI for CAD has emerged as a promising solution to overcome existing limitations. While traditional machine learning techniques utilizing various features, such as texture analysis, for CAD were developed earlier, the advent of DL has significantly enhanced TB detection [3,112]. Hwang et al. developed a DL algorithm that achieved classification and localization performance with an AUROC ranging from 0.973 to 1.000, outperforming physicians, including thoracic radiologists, whose AUROC ranged from 0.746 to 0.971 [113]. Further studies have demonstrated that DL-based algorithms can now assess the disease activity, monitor post-treatment changes, and accurately identify active TB [114,115]. Additionally, Sethanan et al. developed a web-based application capable of classifying TB based on drug resistance from a single CXR [116]. In response to these advancements, the WHO, in 2020, recommended the use of CAD programs as an alternative to human readers for the screening and triage of TB [111]. Even when not used as stand-alone tools, AI may reduce reading time and improve accuracy when used as a second reader [117,118]. Although DL-based algorithms are currently being validated in real-world settings and there are concerns about increased false positive rates, they hold the potential to become a crucial tool in the global effort to eliminate TB [7,117,119].

2.9. Perioperative Care and Interventional Radiology

Imaging plays an essential role in medical procedures and perioperative tasks. Recent developments have increasingly highlighted the utility of AI in preoperative evaluations, extending beyond its established applications such as diagnosing, prognosing, and staging lung cancer [48]. One notable application of AI is its role in preoperative risk assessment. Given the high morbidity rates associated with thoracic surgeries, accurately evaluating individual patient risks before the procedure is crucial [120]. ML-based algorithms are demonstrating greater accuracy in assessing associated risks compared to conventional risk indexes [121,122,123]. Additionally, Chang et al. incorporated an ML model to predict the feasibility of immediate weaning from the ventilator for patients following lung resection [124]. Furthermore, Li et al. created a DL-based 3D reconstruction system capable of visualizing pulmonary anatomic structures, including segmental bronchi, arteries, and veins, with high accuracy [125]. Toggweiler et al. developed a fully automated DL model for planning transcatheter aortic valve replacement (TAVR) using CT imaging, achieving correlation coefficients of 0.95–0.97 for annular perimeter/area and ascending aorta [126]. As AI models continue to improve in accuracy and become more individualized, they will significantly aid in decision-making and planning for wide range of medical interventions.
Furthermore, AI offers various utilities directly related to interventional radiology procedures. Too et al. conducted a retrospective study on a DL model capable of generating needle trajectories for transthoracic lung biopsy [127]. Among the proposed trajectories generated by the software, 85.3% were found to be feasible, with 82% matching the actual path. Similarly, Kisting et al. developed an ML-based algorithm to evaluate pathways for percutaneous lung biopsies [128]. Their model had a correlation of 0.86 with physicians and demonstrated high sensitivity and specificity for safe and unsafe pathways (94.4% and 82.4%, respectively).
Image fusion, which integrates pre-procedural 3D anatomic data with real-time 2D fluoroscopy, offers improved precision during biopsies and ablations [129]. This technique involves incorporating AI model into virtual and augmented reality platforms for automatic landmark recognition, enhancing localization accuracy and efficiency [130]. Moreover, DL-based methods are being applied to generate digital subtraction angiography (DSA), enabling vessel extraction from a single live image without the use of a mask image, thus mitigating image artifacts [131]. With some studies already showcasing the feasibility of autonomous robotic surgeries, there is potential for further integration of AI to augment various procedures in the future [132,133].
Figure 2 illustrates how AI can contribute to the radiology workflow.

3. Limitations of AI

An overview of the limitations and future directions for AI applications is described in Figure 3.
Table 1 presents an overview of the benefits and limitations of AI applications in thoracic radiology.

3.1. AI in the Real World

While AI algorithms have proven their efficacy in controlled test environments, achieving comparable performance in the real-world settings is paramount. The data used for training often exhibit a narrow distribution of variables and higher disease prevalence, potentially leading to ‘covariate shift’ or ‘probability shift’ [134]. Studies have found thoracic AI model to underperform in real-world practice and to have limited applicability [135,136,137]. Keane and Topol describe the concept of the ‘AI Chasm’, emphasizing the importance of developing an algorithm that will generalize across various populations and imaging studies rather than one that works well on a small dataset from a specific population [138]. One study found that the majority of studies on AI algorithms for diagnosis lacked external validation and design features necessary for testing in real-world scenarios [139]. Moreover, selection bias can manifest with any data set that fails to accurately represent the whole population [140]. Seyyed-Kalantari et al. found that AI algorithms trained using three large public CXR datasets displayed bias by underdiagnosing patients in underserved populations [141].
A primary cause and potential solution to this issue lie in the need for ‘big data’. Research on AI algorithms, particularly deep learning, requires access to high-quality, labeled, and publicly available data, which can be exceedingly challenging. Most data are not accessible for research, and even if they are accessible, the data is often not curated, organized, anonymized, annotated, or linked to the ‘ground truth’ [142]. Additionally, high-dimensional data are required to avoid ‘overfitting’, a phenomenon in which the algorithm fails to adequately perform on new data or different samples by learning the noise in addition to the actual data [143].
Numerous efforts are underway to overcome these obstacles, such as the de-identification process for protected health information (PHI) [144]. Natural language processing techniques are commonly employed to efficiently extract labels from radiology reports, but the generated labels may be inaccurate and prone to bias [145]. Innovative data-gathering methods that preserve privacy, such as federated learning, a method to transfer multiple decentralized data, or patient-mediated data sharing, where the patient contribute their own data for research, are actively being researched [142,146]. For instance, by utilizing federated learning, Dayan et al. developed a model that demonstrated a 16% improvement in average AUC across 20 participating sites and a 38% increase in generalizability for predicting outcomes in patients with COVID-19 [146]. Preprocessing techniques such as cross-population training and testing have been studied to mitigate biases such as sampling and confounding bias in training data [147,148]. Multicenter and international collaboration will be essential for the widespread application of AI in healthcare.

3.2. Explainability

One issue with AI algorithms, especially deep learning, is the explainability issue, or the ‘black box’ phenomenon, which describes the lack of transparency in how the algorithm makes decisions. The lack of understanding of how AI algorithms work and process data prevents acceptance and trust in the system [149]. Furthermore, the implementation of AI in this manner could significantly impact the training of new radiologists, as they might overly rely on the AI algorithm without meaningful interactions to develop their image interpretation skills [143]. While techniques such as saliency maps and feature visualization have been developed to address this concern, these methods may be insufficient for complex tasks [150]. Continued efforts to develop explainable AI models are needed to help gain approval from radiologists and integrate them into clinical practice.

3.3. Workflow Integration

Translating research findings into daily clinical practice is a crucial step in the application of AI. For AI assistance to enhance efficiency, seamless integration with minimal disruption is necessary. Additionally, the AI algorithm must display high performance in the targeted clinical cohort. Gaube et al. demonstrated that diagnostic accuracy significantly worsened when physicians received inaccurate advice [151].
Integrating the AI algorithm with PACS and electronic health records (EMR) is a necessary step to enhance operational efficiency. Moreover, AI models that can incorporate clinical context will improve decision-making. Huang et al. found that combining imaging with clinical data from the EMR led to increased accuracy and AUROC compared to single-modality models [152].
Most importantly, successful integration of medical AI relies heavily on support from radiologists. A study found that radiologists rated advice as lower quality when it came from an AI system [151]. Efforts to increase physician understanding of AI and its limitation are pivotal for further AI advancements [150]. Given the capacity of AI for self-learning, establishing a feedback system will be crucial for tailoring AI models to meet the diverse needs of all stakeholders. Demonstrating the clinical utility of AI through more clinical trials and observational studies is imperative, as improved diagnostic accuracy may not always translate to better patient outcomes [14,153]. Topol highlights the idea that metrics such as ROC or AUC are not necessarily indicative of clinical utility and do not demonstrate clinical efficacy [154]. High-quality evidence that determines the clinical impact of AI in healthcare, particularly focusing on the benefits for patients, is needed.
Furthermore, discussing reimbursement for AI model is vital for its widespread adoption [68]. Reimbursing AI technology separately should be considered if the application demonstrates clinical utility, such as improved diagnostic performance or providing new diagnostic information [155]. Reimbursement frameworks that incorporate all stakeholders, as well as propose the sustainable implementation of AI with high value, are being studied [155,156].

3.4. Legal and Ethical Concerns

One consideration with AI in radiology is data ethics. It is essential to address all aspects of data, including its collection and usage, data ownership, and protection [157]. Ongoing legislative efforts aimed at safeguarding patient privacy while facilitating innovative AI research will be pivotal [158].
Another issue is determining liability when utilizing AI model. It is essential to clarify responsibility for adverse outcomes, whether it lies with the manufacturer of the device, the hospital or department procuring the tool, or the physician utilizing the AI model [140]. Ambiguities persist, for instance in scenarios where AI triage model fails to detect a critical finding, resulting in an adverse event for the patient [143].
Implementing AI systems demands significant resources and expertise, which smaller hospitals and resource-limited countries may lack the capacity to accommodate [140]. Efforts to mitigate disparities in access to AI and its potential benefits are crucial. Additionally, resource-poor populations may be more vulnerable to automation bias, relying heavily on AI programs that could result in errors of omission and commission [157]. Equally vital to the development and implementation of AI systems is the promotion of safe and ethical practices for AI in healthcare.

4. Future Directions

While AI holds promise as a potential revolution in radiology, it remains in its early stages of development and requires ongoing validation. However, AI will continue to evolve with the advancements in data collection, algorithm development, and computational power. Initially focusing on non-diagnostic tasks such as worklist prioritization and operational workflow, AI will demonstrate its utility in the clinical setting. Regarding diagnosis, AI will greatly increase efficiency in detecting common diseases, while complex tasks will likely still remain the responsibility of radiologists [159].
It is crucial to recognize that the role of AI is not to replace radiologists but to assist them. Currently, the majority of AI algorithms are trained to excel at a single specific task. However, in the future, systems will have the capability to integrate multiple functions, such as image reconstruction, detection, and report generation, to significantly enhance lung cancer screening [160].
Large language models (LLMs), such as the ‘generative pre-trained transformer’ (GPT)-based models developed by OpenAI, are specifically designed to comprehend and generate text that closely resembles human language [161]. GPT models, based on DL principles, are gaining significant attention for their ability to generate output by learning from textual instructions, often requiring minimal training data [162]. Research has already been conducted on using GPT models to extract information from thoracic imaging reports [163,164]. In a study by Fink et al., GPT-4, the fourth version of GPT models, demonstrated high performance in extracting lesion parameters (98.6%), identifying metastatic disease (98.1%), and generating labels for oncologic progression (F1 score of 0.96) from CT reports on lung cancer [164]. Adams et al. employed GPT-4 as a tool to convert free-text radiology reports into structured templates for standardization [165]. Furthermore, Zaki et al. demonstrated the potential utility of LLMs in predicting the appropriate imaging study for a specific clinical presentation [161]. Applications of LLMs in healthcare can benefit both physicians and patients and span across multiple domains, including interventional radiology [162].
AI can significantly impact areas most affected by the shortage of radiologists, serving as a bridge to address resource limitation and barriers to health equity [166]. With AI technologies, healthcare providers can extend radiological services to underserved regions and populations, helping to reduce disparities in access to care.

5. Conclusions

Undoubtedly, AI will increasingly shape the future of radiology, necessitating a comprehensive understanding of its impact and limitations to adequately prepare for what lies ahead. Already, AI is influencing multiple aspects of thoracic radiology, ranging from detection to diagnosis and prognosis. It facilitates techniques such as quantitative imaging and radiomics, improving the characterization and understanding of pulmonary diseases. Additionally, AI streamlines operational workflow, enhances imaging safety, and enables faster diagnosis of critical cases. Furthermore, AI plays a critical role in global TB and COVID-19 management, while also contributing significantly to procedural planning and risk stratification. Despite challenges that remain in implementing AI in clinical practice, ongoing advancements promise continued improvement. The field of AI in thoracic radiology presents exciting opportunities revolutionize radiology practice and enhance patient outcomes worldwide.

Author Contributions

Conceptualization, J.Y.C. and M.S.M.; literature search and data curation, J.Y.C. and M.S.M.; writing—original draft preparation, J.Y.C.; writing—review and editing, J.Y.C. and M.S.M.; supervision, M.S.M. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

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

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Kwee, T.C.; Almaghrabi, M.T.; Kwee, R.M. Diagnostic radiology and its future: What do clinicians need and think? Eur. Radiol. 2023, 33, 9401–9410. [Google Scholar] [CrossRef] [PubMed]
  2. Smith-Bindman, R.; Kwan, M.L.; Marlow, E.C.; Theis, M.K.; Bolch, W.; Cheng, S.Y.; Bowles, E.J.A.; Duncan, J.R.; Greenlee, R.T.; Kushi, L.H.; et al. Trends in Use of Medical Imaging in US Health Care Systems and in Ontario, Canada, 2000–2016. JAMA 2019, 322, 843–856. [Google Scholar] [CrossRef]
  3. Kim, Y.; Park, J.Y.; Hwang, E.J.; Lee, S.M.; Park, C.M. Applications of artificial intelligence in the thorax: A narrative review focusing on thoracic radiology. J. Thorac. Dis. 2021, 13, 6943–6962. [Google Scholar] [CrossRef] [PubMed]
  4. McLoud, T.C.; Little, B.P. Thoracic Radiology: Recent Developments and Future Trends. Radiology 2023, 306, e223121. [Google Scholar] [CrossRef] [PubMed]
  5. Busnatu, Ș.; Niculescu, A.G.; Bolocan, A.; Petrescu, G.E.D.; Păduraru, D.N.; Năstasă, I.; Lupușoru, M.; Geantă, M.; Andronic, O.; Grumezescu, A.M.; et al. Clinical Applications of Artificial Intelligence-An Updated Overview. J. Clin. Med. 2022, 11, 2265. [Google Scholar] [CrossRef] [PubMed]
  6. American College of Radiology Data Science Institute. Available online: https://aicentral.acrdsi.org/All-Ai-products#f:subspeciality=[Chest] (accessed on 3 May 2024).
  7. Chassagnon, G.; Vakalopoulou, M.; Paragios, N.; Revel, M.P. Artificial intelligence applications for thoracic imaging. Eur. J. Radiol. 2020, 123, 108774. [Google Scholar] [CrossRef] [PubMed]
  8. Nam, J.G.; Kim, M.; Park, J.; Hwang, E.J.; Lee, J.H.; Hong, J.H.; Goo, J.M.; Park, C.M. Development and validation of a deep learning algorithm detecting 10 common abnormalities on chest radiographs. Eur. Respir. J. 2021, 57, 2003061. [Google Scholar] [CrossRef] [PubMed]
  9. Baltruschat, I.; Steinmeister, L.; Nickisch, H.; Saalbach, A.; Grass, M.; Adam, G.; Knopp, T.; Ittrich, H. Smart chest X-ray worklist prioritization using artificial intelligence: A clinical workflow simulation. Eur. Radiol. 2021, 31, 3837–3845. [Google Scholar] [CrossRef] [PubMed]
  10. Annarumma, M.; Withey, S.J.; Bakewell, R.J.; Pesce, E.; Goh, V.; Montana, G. Automated Triaging of Adult Chest Radiographs with Deep Artificial Neural Networks. Radiology 2019, 291, 196–202. [Google Scholar] [CrossRef]
  11. Huang, S.C.; Kothari, T.; Banerjee, I.; Chute, C.; Ball, R.L.; Borus, N.; Huang, A.; Patel, B.N.; Rajpurkar, P.; Irvin, J.; et al. PENet-a scalable deep-learning model for automated diagnosis of pulmonary embolism using volumetric CT imaging. NPJ Digit. Med. 2020, 3, 61. [Google Scholar] [CrossRef]
  12. Hata, A.; Yanagawa, M.; Yamagata, K.; Suzuki, Y.; Kido, S.; Kawata, A.; Doi, S.; Yoshida, Y.; Miyata, T.; Tsubamoto, M.; et al. Deep learning algorithm for detection of aortic dissection on non-contrast-enhanced CT. Eur. Radiol. 2021, 31, 1151–1159. [Google Scholar] [CrossRef] [PubMed]
  13. Glessgen, C.G.; Boulougouri, M.; Vallée, J.P.; Noble, S.; Platon, A.; Poletti, P.A.; Paul, J.F.; Deux, J.F. Artificial intelligence-based opportunistic detection of coronary artery stenosis on aortic computed tomography angiography in emergency department patients with acute chest pain. Eur. Heart J. Open 2023, 3, oead088. [Google Scholar] [CrossRef]
  14. Hwang, E.J.; Park, C.M. Clinical Implementation of Deep Learning in Thoracic Radiology: Potential Applications and Challenges. Korean J. Radiol. 2020, 21, 511–525. [Google Scholar] [CrossRef] [PubMed]
  15. Nam, J.G.; Park, S.; Hwang, E.J.; Lee, J.H.; Jin, K.N.; Lim, K.Y.; Vu, T.H.; Sohn, J.H.; Hwang, S.; Goo, J.M.; et al. Development and Validation of Deep Learning-based Automatic Detection Algorithm for Malignant Pulmonary Nodules on Chest Radiographs. Radiology 2019, 290, 218–228. [Google Scholar] [CrossRef] [PubMed]
  16. Nam, J.G.; Hwang, E.J.; Kim, J.; Park, N.; Lee, E.H.; Kim, H.J.; Nam, M.; Lee, J.H.; Park, C.M.; Goo, J.M. AI Improves Nodule Detection on Chest Radiographs in a Health Screening Population: A Randomized Controlled Trial. Radiology 2023, 307, e221894. [Google Scholar] [CrossRef]
  17. Rajpurkar, P.; Irvin, J.; Ball, R.L.; Zhu, K.; Yang, B.; Mehta, H.; Duan, T.; Ding, D.; Bagul, A.; Langlotz, C.P.; et al. Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists. PLoS Med. 2018, 15, e1002686. [Google Scholar] [CrossRef] [PubMed]
  18. Hwang, E.J.; Park, S.; Jin, K.N.; Kim, J.I.; Choi, S.Y.; Lee, J.H.; Goo, J.M.; Aum, J.; Yim, J.J.; Cohen, J.G.; et al. Development and Validation of a Deep Learning-Based Automated Detection Algorithm for Major Thoracic Diseases on Chest Radiographs. JAMA Netw. Open 2019, 2, e191095. [Google Scholar] [CrossRef] [PubMed]
  19. Alkassar, S.; Abdullah, M.A.; Jebur, B.A.; Abdul-Majeed, G.H.; Wei, B.; Woo, W.L. Automated diagnosis of childhood pneumonia in chest radiographs using modified densely residual bottleneck-layer features. Appl. Sci. 2021, 11, 11461. [Google Scholar] [CrossRef]
  20. Hwang, E.J.; Nam, J.G.; Lim, W.H.; Park, S.J.; Jeong, Y.S.; Kang, J.H.; Hong, E.K.; Kim, T.M.; Goo, J.M.; Park, S.; et al. Deep Learning for Chest Radiograph Diagnosis in the Emergency Department. Radiology 2019, 293, 573–580. [Google Scholar] [CrossRef]
  21. Huang, J.; Neill, L.; Wittbrodt, M.; Melnick, D.; Klug, M.; Thompson, M.; Bailitz, J.; Loftus, T.; Malik, S.; Phull, A.; et al. Generative Artificial Intelligence for Chest Radiograph Interpretation in the Emergency Department. JAMA Netw. Open 2023, 6, e2336100. [Google Scholar] [CrossRef]
  22. Pyrros, A.; Borstelmann, S.M.; Mantravadi, R.; Zaiman, Z.; Thomas, K.; Price, B.; Greenstein, E.; Siddiqui, N.; Willis, M.; Shulhan, I.; et al. Opportunistic detection of type 2 diabetes using deep learning from frontal chest radiographs. Nat. Commun. 2023, 14, 4039. [Google Scholar] [CrossRef] [PubMed]
  23. Farina, J.M.; Pereyra, M.; Mahmoud, A.K.; Scalia, I.G.; Abbas, M.T.; Chao, C.J.; Barry, T.; Ayoub, C.; Banerjee, I.; Arsanjani, R. Artificial Intelligence-Based Prediction of Cardiovascular Diseases from Chest Radiography. J. Imaging 2023, 9, 236. [Google Scholar] [CrossRef] [PubMed]
  24. Ueda, D.; Matsumoto, T.; Ehara, S.; Yamamoto, A.; Walston, S.L.; Ito, A.; Shimono, T.; Shiba, M.; Takeshita, T.; Fukuda, D.; et al. Artificial intelligence-based model to classify cardiac functions from chest radiographs: A multi-institutional, retrospective model development and validation study. Lancet Digit. Health 2023, 5, e525–e533. [Google Scholar] [CrossRef]
  25. Ahn, J.S.; Ebrahimian, S.; McDermott, S.; Lee, S.; Naccarato, L.; Di Capua, J.F.; Wu, M.Y.; Zhang, E.W.; Muse, V.; Miller, B.; et al. Association of Artificial Intelligence-Aided Chest Radiograph Interpretation With Reader Performance and Efficiency. JAMA Netw. Open 2022, 5, e2229289. [Google Scholar] [CrossRef]
  26. Hsu, H.H.; Ko, K.H.; Chou, Y.C.; Wu, Y.C.; Chiu, S.H.; Chang, C.K.; Chang, W.C. Performance and reading time of lung nodule identification on multidetector CT with or without an artificial intelligence-powered computer-aided detection system. Clin. Radiol. 2021, 76, 626.e23–626.e32. [Google Scholar] [CrossRef] [PubMed]
  27. Yoo, H.; Kim, E.Y.; Kim, H.; Choi, Y.R.; Kim, M.Y.; Hwang, S.H.; Kim, Y.J.; Cho, Y.J.; Jin, K.N. Artificial Intelligence-Based Identification of Normal Chest Radiographs: A Simulation Study in a Multicenter Health Screening Cohort. Korean J. Radiol. 2022, 23, 1009–1018. [Google Scholar] [CrossRef] [PubMed]
  28. Plesner, L.L.; Müller, F.C.; Nybing, J.D.; Laustrup, L.C.; Rasmussen, F.; Nielsen, O.W.; Boesen, M.; Andersen, M.B. Autonomous Chest Radiograph Reporting Using AI: Estimation of Clinical Impact. Radiology 2023, 307, e222268. [Google Scholar] [CrossRef] [PubMed]
  29. Aberle, D.R.; Adams, A.M.; Berg, C.D.; Black, W.C.; Clapp, J.D.; Fagerstrom, R.M.; Gareen, I.F.; Gatsonis, C.; Marcus, P.M.; Sicks, J.D. Reduced lung-cancer mortality with low-dose computed tomographic screening. N. Engl. J. Med. 2011, 365, 395–409. [Google Scholar] [CrossRef]
  30. de Koning, H.J.; van der Aalst, C.M.; de Jong, P.A.; Scholten, E.T.; Nackaerts, K.; Heuvelmans, M.A.; Lammers, J.J.; Weenink, C.; Yousaf-Khan, U.; Horeweg, N.; et al. Reduced Lung-Cancer Mortality with Volume CT Screening in a Randomized Trial. N. Engl. J. Med. 2020, 382, 503–513. [Google Scholar] [CrossRef]
  31. Krist, A.H.; Davidson, K.W.; Mangione, C.M.; Barry, M.J.; Cabana, M.; Caughey, A.B.; Davis, E.M.; Donahue, K.E.; Doubeni, C.A.; Kubik, M.; et al. Screening for Lung Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2021, 325, 962–970. [Google Scholar] [CrossRef]
  32. Pehrson, L.M.; Nielsen, M.B.; Ammitzbøl Lauridsen, C. Automatic Pulmonary Nodule Detection Applying Deep Learning or Machine Learning Algorithms to the LIDC-IDRI Database: A Systematic Review. Diagnostics 2019, 9, 29. [Google Scholar] [CrossRef]
  33. Ewals, L.J.S.; van der Wulp, K.; van den Borne, B.; Pluyter, J.R.; Jacobs, I.; Mavroeidis, D.; van der Sommen, F.; Nederend, J. The Effects of Artificial Intelligence Assistance on the Radiologists’ Assessment of Lung Nodules on CT Scans: A Systematic Review. J. Clin. Med. 2023, 12, 3536. [Google Scholar] [CrossRef] [PubMed]
  34. Martini, K.; Blüthgen, C.; Eberhard, M.; Schönenberger, A.L.N.; De Martini, I.; Huber, F.A.; Barth, B.K.; Euler, A.; Frauenfelder, T. Impact of Vessel Suppressed-CT on Diagnostic Accuracy in Detection of Pulmonary Metastasis and Reading Time. Acad. Radiol. 2021, 28, 988–994. [Google Scholar] [CrossRef] [PubMed]
  35. Ciompi, F.; Chung, K.; van Riel, S.J.; Setio, A.A.A.; Gerke, P.K.; Jacobs, C.; Scholten, E.T.; Schaefer-Prokop, C.; Wille, M.M.W.; Marchianò, A.; et al. Towards automatic pulmonary nodule management in lung cancer screening with deep learning. Sci. Rep. 2017, 7, 46479. [Google Scholar] [CrossRef] [PubMed]
  36. Ardila, D.; Kiraly, A.P.; Bharadwaj, S.; Choi, B.; Reicher, J.J.; Peng, L.; Tse, D.; Etemadi, M.; Ye, W.; Corrado, G.; et al. End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nat. Med. 2019, 25, 954–961. [Google Scholar] [CrossRef] [PubMed]
  37. Adams, S.J.; Stone, E.; Baldwin, D.R.; Vliegenthart, R.; Lee, P.; Fintelmann, F.J. Lung cancer screening. Lancet 2023, 401, 390–408. [Google Scholar] [CrossRef] [PubMed]
  38. Adams, S.J.; Mondal, P.; Penz, E.; Tyan, C.C.; Lim, H.; Babyn, P. Development and Cost Analysis of a Lung Nodule Management Strategy Combining Artificial Intelligence and Lung-RADS for Baseline Lung Cancer Screening. J. Am. Coll. Radiol. 2021, 18, 741–751. [Google Scholar] [CrossRef] [PubMed]
  39. Mikhael, P.G.; Wohlwend, J.; Yala, A.; Karstens, L.; Xiang, J.; Takigami, A.K.; Bourgouin, P.P.; Chan, P.; Mrah, S.; Amayri, W.; et al. Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography. J. Clin. Oncol. 2023, 41, 2191–2200. [Google Scholar] [CrossRef] [PubMed]
  40. Cheikh, A.B.; Gorincour, G.; Nivet, H.; May, J.; Seux, M.; Calame, P.; Thomson, V.; Delabrousse, E.; Crombé, A. How artificial intelligence improves radiological interpretation in suspected pulmonary embolism. Eur. Radiol. 2022, 32, 5831–5842. [Google Scholar] [CrossRef]
  41. Walsh, S.L.F.; Calandriello, L.; Silva, M.; Sverzellati, N. Deep learning for classifying fibrotic lung disease on high-resolution computed tomography: A case-cohort study. Lancet Respir. Med. 2018, 6, 837–845. [Google Scholar] [CrossRef]
  42. Christe, A.; Peters, A.A.; Drakopoulos, D.; Heverhagen, J.T.; Geiser, T.; Stathopoulou, T.; Christodoulidis, S.; Anthimopoulos, M.; Mougiakakou, S.G.; Ebner, L. Computer-Aided Diagnosis of Pulmonary Fibrosis Using Deep Learning and CT Images. Investig. Radiol. 2019, 54, 627–632. [Google Scholar] [CrossRef] [PubMed]
  43. González, G.; Ash, S.Y.; Vegas-Sánchez-Ferrero, G.; Onieva Onieva, J.; Rahaghi, F.N.; Ross, J.C.; Díaz, A.; San José Estépar, R.; Washko, G.R. Disease Staging and Prognosis in Smokers Using Deep Learning in Chest Computed Tomography. Am. J. Respir. Crit. Care Med. 2018, 197, 193–203. [Google Scholar] [CrossRef] [PubMed]
  44. Stemmer, A.; Shadmi, R.; Bregman-Amitai, O.; Chettrit, D.; Blagev, D.; Orlovsky, M.; Deutsch, L.; Elnekave, E. Using machine learning algorithms to review computed tomography scans and assess risk for cardiovascular disease: Retrospective analysis from the National Lung Screening Trial (NLST). PLoS ONE 2020, 15, e0236021. [Google Scholar] [CrossRef] [PubMed]
  45. Nambu, A.; Zach, J.; Schroeder, J.; Jin, G.; Kim, S.S.; Kim, Y.I.; Schnell, C.; Bowler, R.; Lynch, D.A. Quantitative computed tomography measurements to evaluate airway disease in chronic obstructive pulmonary disease: Relationship to physiological measurements, clinical index and visual assessment of airway disease. Eur. J. Radiol. 2016, 85, 2144–2151. [Google Scholar] [CrossRef]
  46. Bartholmai, B.J.; Raghunath, S.; Karwoski, R.A.; Moua, T.; Rajagopalan, S.; Maldonado, F.; Decker, P.A.; Robb, R.A. Quantitative computed tomography imaging of interstitial lung diseases. J. Thorac. Imaging 2013, 28, 298–307. [Google Scholar] [CrossRef] [PubMed]
  47. Chassagnon, G.; Vakalopoulou, M.; Régent, A.; Zacharaki, E.I.; Aviram, G.; Martin, C.; Marini, R.; Bus, N.; Jerjir, N.; Mekinian, A.; et al. Deep Learning-based Approach for Automated Assessment of Interstitial Lung Disease in Systemic Sclerosis on CT Images. Radiol. Artif. Intell. 2020, 2, e190006. [Google Scholar] [CrossRef] [PubMed]
  48. Hasenstab, K.A.; Yuan, N.; Retson, T.; Conrad, D.J.; Kligerman, S.; Lynch, D.A.; Hsiao, A. Automated CT Staging of Chronic Obstructive Pulmonary Disease Severity for Predicting Disease Progression and Mortality with a Deep Learning Convolutional Neural Network. Radiol. Cardiothorac. Imaging 2021, 3, e200477. [Google Scholar] [CrossRef] [PubMed]
  49. Park, J.; Yun, J.; Kim, N.; Park, B.; Cho, Y.; Park, H.J.; Song, M.; Lee, M.; Seo, J.B. Fully Automated Lung Lobe Segmentation in Volumetric Chest CT with 3D U-Net: Validation with Intra- and Extra-Datasets. J. Digit. Imaging 2020, 33, 221–230. [Google Scholar] [CrossRef] [PubMed]
  50. Garcia-Uceda, A.; Selvan, R.; Saghir, Z.; Tiddens, H.; de Bruijne, M. Automatic airway segmentation from computed tomography using robust and efficient 3-D convolutional neural networks. Sci. Rep. 2021, 11, 16001. [Google Scholar] [CrossRef]
  51. Dong, X.; Xu, S.; Liu, Y.; Wang, A.; Saripan, M.I.; Li, L.; Zhang, X.; Lu, L. Multi-view secondary input collaborative deep learning for lung nodule 3D segmentation. Cancer Imaging 2020, 20, 53. [Google Scholar] [CrossRef]
  52. de Margerie-Mellon, C.; Chassagnon, G. Artificial intelligence: A critical review of applications for lung nodule and lung cancer. Diagn Interv. Imaging 2023, 104, 11–17. [Google Scholar] [CrossRef]
  53. Dudurych, I.; Garcia-Uceda, A.; Petersen, J.; Du, Y.; Vliegenthart, R.; de Bruijne, M. Reproducibility of a combined artificial intelligence and optimal-surface graph-cut method to automate bronchial parameter extraction. Eur. Radiol. 2023, 33, 6718–6725. [Google Scholar] [CrossRef]
  54. Lv, Q.; Gallardo-Estrella, L.; Andrinopoulou, E.R.; Chen, Y.; Charbonnier, J.P.; Sandvik, R.M.; Caudri, D.; Nielsen, K.G.; de Bruijne, M.; Ciet, P.; et al. Automatic analysis of bronchus-artery dimensions to diagnose and monitor airways disease in cystic fibrosis. Thorax 2023, 79, 13–22. [Google Scholar] [CrossRef] [PubMed]
  55. Pickhardt, P.J. Value-added Opportunistic CT Screening: State of the Art. Radiology 2022, 303, 241–254. [Google Scholar] [CrossRef]
  56. Lee, S.B.; Cho, Y.J.; Hong, Y.; Jeong, D.; Lee, J.; Kim, S.H.; Lee, S.; Choi, Y.H. Deep Learning-Based Image Conversion Improves the Reproducibility of Computed Tomography Radiomics Features: A Phantom Study. Investig. Radiol. 2022, 57, 308–317. [Google Scholar] [CrossRef]
  57. Chung, C.; Kalpathy-Cramer, J.; Knopp, M.V.; Jaffray, D.A. In the Era of Deep Learning, Why Reconstruct an Image at All? J. Am. Coll. Radiol. 2021, 18, 170–173. [Google Scholar] [CrossRef]
  58. Lee, S.M.; Lee, J.G.; Lee, G.; Choe, J.; Do, K.H.; Kim, N.; Seo, J.B. CT Image Conversion among Different Reconstruction Kernels without a Sinogram by Using a Convolutional Neural Network. Korean J. Radiol. 2019, 20, 295–303. [Google Scholar] [CrossRef]
  59. Choe, J.; Lee, S.M.; Do, K.H.; Lee, G.; Lee, J.G.; Lee, S.M.; Seo, J.B. Deep Learning-based Image Conversion of CT Reconstruction Kernels Improves Radiomics Reproducibility for Pulmonary Nodules or Masses. Radiology 2019, 292, 365–373. [Google Scholar] [CrossRef] [PubMed]
  60. Ahn, Y.; Lee, S.M.; Nam, Y.; Lee, H.; Choe, J.; Do, K.H.; Seo, J.B. Deep Learning-Based CT Reconstruction Kernel Conversion in the Quantification of Interstitial Lung Disease: Effect on Reproducibility. Acad. Radiol. 2024, 31, 693–705. [Google Scholar] [CrossRef]
  61. Gillies, R.J.; Kinahan, P.E.; Hricak, H. Radiomics: Images Are More than Pictures, They Are Data. Radiology 2016, 278, 563–577. [Google Scholar] [CrossRef]
  62. Beig, N.; Khorrami, M.; Alilou, M.; Prasanna, P.; Braman, N.; Orooji, M.; Rakshit, S.; Bera, K.; Rajiah, P.; Ginsberg, J.; et al. Perinodular and Intranodular Radiomic Features on Lung CT Images Distinguish Adenocarcinomas from Granulomas. Radiology 2019, 290, 783–792. [Google Scholar] [CrossRef]
  63. Hawkins, S.; Wang, H.; Liu, Y.; Garcia, A.; Stringfield, O.; Krewer, H.; Li, Q.; Cherezov, D.; Gatenby, R.A.; Balagurunathan, Y.; et al. Predicting Malignant Nodules from Screening CT Scans. J. Thorac. Oncol. 2016, 11, 2120–2128. [Google Scholar] [CrossRef]
  64. Chassagnon, G.; De Margerie-Mellon, C.; Vakalopoulou, M.; Marini, R.; Hoang-Thi, T.N.; Revel, M.P.; Soyer, P. Artificial intelligence in lung cancer: Current applications and perspectives. Jpn. J. Radiol. 2023, 41, 235–244. [Google Scholar] [CrossRef] [PubMed]
  65. Fan, L.; Fang, M.; Li, Z.; Tu, W.; Wang, S.; Chen, W.; Tian, J.; Dong, D.; Liu, S. Radiomics signature: A biomarker for the preoperative discrimination of lung invasive adenocarcinoma manifesting as a ground-glass nodule. Eur. Radiol. 2019, 29, 889–897. [Google Scholar] [CrossRef]
  66. Cherezov, D.; Goldgof, D.; Hall, L.; Gillies, R.; Schabath, M.; Müller, H.; Depeursinge, A. Revealing Tumor Habitats from Texture Heterogeneity Analysis for Classification of Lung Cancer Malignancy and Aggressiveness. Sci. Rep. 2019, 9, 4500. [Google Scholar] [CrossRef]
  67. Wu, W.; Parmar, C.; Grossmann, P.; Quackenbush, J.; Lambin, P.; Bussink, J.; Mak, R.; Aerts, H.J. Exploratory Study to Identify Radiomics Classifiers for Lung Cancer Histology. Front. Oncol. 2016, 6, 71. [Google Scholar] [CrossRef]
  68. Vliegenthart, R.; Fouras, A.; Jacobs, C.; Papanikolaou, N. Innovations in thoracic imaging: CT, radiomics, AI and x-ray velocimetry. Respirology 2022, 27, 818–833. [Google Scholar] [CrossRef]
  69. Jia, T.Y.; Xiong, J.F.; Li, X.Y.; Yu, W.; Xu, Z.Y.; Cai, X.W.; Ma, J.C.; Ren, Y.C.; Larsson, R.; Zhang, J.; et al. Identifying EGFR mutations in lung adenocarcinoma by noninvasive imaging using radiomics features and random forest modeling. Eur. Radiol. 2019, 29, 4742–4750. [Google Scholar] [CrossRef] [PubMed]
  70. Rizzo, S.; Petrella, F.; Buscarino, V.; De Maria, F.; Raimondi, S.; Barberis, M.; Fumagalli, C.; Spitaleri, G.; Rampinelli, C.; De Marinis, F.; et al. CT Radiogenomic Characterization of EGFR, K-RAS, and ALK Mutations in Non-Small Cell Lung Cancer. Eur. Radiol. 2016, 26, 32–42. [Google Scholar] [CrossRef] [PubMed]
  71. Lee, G.; Park, H.; Bak, S.H.; Lee, H.Y. Radiomics in Lung Cancer from Basic to Advanced: Current Status and Future Directions. Korean J. Radiol. 2020, 21, 159–171. [Google Scholar] [CrossRef]
  72. Willemink, M.J.; Noël, P.B. The evolution of image reconstruction for CT-from filtered back projection to artificial intelligence. Eur Radiol 2019, 29, 2185–2195. [Google Scholar] [CrossRef]
  73. Padole, A.; Ali Khawaja, R.D.; Kalra, M.K.; Singh, S. CT radiation dose and iterative reconstruction techniques. AJR Am. J. Roentgenol. 2015, 204, W384-392. [Google Scholar] [CrossRef]
  74. Singh, R.; Wu, W.; Wang, G.; Kalra, M.K. Artificial intelligence in image reconstruction: The change is here. Phys. Med. 2020, 79, 113–125. [Google Scholar] [CrossRef]
  75. Wolterink, J.M.; Leiner, T.; Viergever, M.A.; Isgum, I. Generative Adversarial Networks for Noise Reduction in Low-Dose CT. IEEE Trans. Med. Imaging 2017, 36, 2536–2545. [Google Scholar] [CrossRef] [PubMed]
  76. Kang, E.; Min, J.; Ye, J.C. A deep convolutional neural network using directional wavelets for low-dose X-ray CT reconstruction. Med. Phys. 2017, 44, e360–e375. [Google Scholar] [CrossRef]
  77. Chandran, M.O.; Pendem, S.; P, S.P.; Chacko, C.; Priyanka; Kadavigere, R. Influence of deep learning image reconstruction algorithm for reducing radiation dose and image noise compared to iterative reconstruction and filtered back projection for head and chest computed tomography examinations: A systematic review. F1000Research 2024, 13, 274. [Google Scholar] [CrossRef]
  78. Jiang, B.; Li, N.; Shi, X.; Zhang, S.; Li, J.; de Bock, G.H.; Vliegenthart, R.; Xie, X. Deep Learning Reconstruction Shows Better Lung Nodule Detection for Ultra-Low-Dose Chest CT. Radiology 2022, 303, 202–212. [Google Scholar] [CrossRef]
  79. Nakamura, Y.; Higaki, T.; Kondo, S.; Kawashita, I.; Takahashi, I.; Awai, K. An introduction to photon-counting detector CT (PCD CT) for radiologists. Jpn. J. Radiol. 2023, 41, 266–282. [Google Scholar] [CrossRef] [PubMed]
  80. Longuefosse, A.; Raoult, J.; Benlala, I.; Denis de Senneville, B.; Benkert, T.; Macey, J.; Bui, S.; Berger, P.; Ferretti, G.; Gaubert, J.Y.; et al. Generating High-Resolution Synthetic CT from Lung MRI with Ultrashort Echo Times: Initial Evaluation in Cystic Fibrosis. Radiology 2023, 308, e230052. [Google Scholar] [CrossRef]
  81. Zhang, Y.; Yu, H. Convolutional Neural Network Based Metal Artifact Reduction in X-Ray Computed Tomography. IEEE Trans. Med. Imaging 2018, 37, 1370–1381. [Google Scholar] [CrossRef]
  82. Ahn, S.Y.; Chae, K.J.; Goo, J.M. The Potential Role of Grid-Like Software in Bedside Chest Radiography in Improving Image Quality and Dose Reduction: An Observer Preference Study. Korean J. Radiol. 2018, 19, 526–533. [Google Scholar] [CrossRef]
  83. Faucon, A.L.; Bobrie, G.; Clément, O. Nephrotoxicity of iodinated contrast media: From pathophysiology to prevention strategies. Eur. J. Radiol. 2019, 116, 231–241. [Google Scholar] [CrossRef]
  84. Pasquini, L.; Napolitano, A.; Pignatelli, M.; Tagliente, E.; Parrillo, C.; Nasta, F.; Romano, A.; Bozzao, A.; Di Napoli, A. Synthetic Post-Contrast Imaging through Artificial Intelligence: Clinical Applications of Virtual and Augmented Contrast Media. Pharmaceutics 2022, 14, 2378. [Google Scholar] [CrossRef] [PubMed]
  85. Choi, J.W.; Cho, Y.J.; Ha, J.Y.; Lee, S.B.; Lee, S.; Choi, Y.H.; Cheon, J.E.; Kim, W.S. Generating synthetic contrast enhancement from non-contrast chest computed tomography using a generative adversarial network. Sci. Rep. 2021, 11, 20403. [Google Scholar] [CrossRef]
  86. Chun, J.; Chang, J.S.; Oh, C.; Park, I.; Choi, M.S.; Hong, C.S.; Kim, H.; Yang, G.; Moon, J.Y.; Chung, S.Y.; et al. Synthetic contrast-enhanced computed tomography generation using a deep convolutional neural network for cardiac substructure delineation in breast cancer radiation therapy: A feasibility study. Radiat. Oncol. 2022, 17, 83. [Google Scholar] [CrossRef]
  87. Ranschaert, E.; Topff, L.; Pianykh, O. Optimization of Radiology Workflow with Artificial Intelligence. Radiol. Clin. N. Am. 2021, 59, 955–966. [Google Scholar] [CrossRef]
  88. Kapoor, N.; Lacson, R.; Khorasani, R. Workflow Applications of Artificial Intelligence in Radiology and an Overview of Available Tools. J. Am. Coll. Radiol. 2020, 17, 1363–1370. [Google Scholar] [CrossRef]
  89. Chong, L.R.; Tsai, K.T.; Lee, L.L.; Foo, S.G.; Chang, P.C. Artificial Intelligence Predictive Analytics in the Management of Outpatient MRI Appointment No-Shows. AJR Am. J. Roentgenol. 2020, 215, 1155–1162. [Google Scholar] [CrossRef]
  90. Pannu, J.K.; Venious, G.; Gallagher, R.; Shaver, A.; Cloyes, R.; Josan, E.; Donnelly, E.; King, M.; Knopp, M.; Merritt, R.; et al. P1.03-01 Do We Follow Incidental Lung Nodules Appropriately? A Retrospective Study. J. Thorac. Oncol. 2022, 17, S100. [Google Scholar] [CrossRef]
  91. Liu, J.A.; Yang, I.Y.; Tsai, E.B. Artificial Intelligence (AI) for Lung Nodules, From the AJR Special Series on AI Applications. AJR Am. J. Roentgenol. 2022, 219, 703–712. [Google Scholar] [CrossRef]
  92. Desai, S.; Kapoor, N.; Hammer, M.M.; Levie, A.; Sivashanker, K.; Lacson, R.; Khorasani, R. RADAR: A Closed-Loop Quality Improvement Initiative Leveraging A Safety Net Model for Incidental Pulmonary Nodule Management. Jt. Comm. J. Qual Patient Saf. 2021, 47, 275–281. [Google Scholar] [CrossRef] [PubMed]
  93. Abbasi, N.; Lacson, R.; Kapoor, N.; Licaros, A.; Guenette, J.P.; Burk, K.S.; Hammer, M.; Desai, S.; Eappen, S.; Saini, S.; et al. Development and External Validation of an Artificial Intelligence Model for Identifying Radiology Reports Containing Recommendations for Additional Imaging. AJR Am. J. Roentgenol. 2023, 221, 377–385. [Google Scholar] [CrossRef]
  94. Yanagawa, M.; Ito, R.; Nozaki, T.; Fujioka, T.; Yamada, A.; Fujita, S.; Kamagata, K.; Fushimi, Y.; Tsuboyama, T.; Matsui, Y.; et al. New trend in artificial intelligence-based assistive technology for thoracic imaging. Radiol. Med. 2023, 128, 1236–1249. [Google Scholar] [CrossRef] [PubMed]
  95. Ebrahimzadeh, S.; Islam, N.; Dawit, H.; Salameh, J.P.; Kazi, S.; Fabiano, N.; Treanor, L.; Absi, M.; Ahmad, F.; Rooprai, P.; et al. Thoracic imaging tests for the diagnosis of COVID-19. Cochrane Database Syst. Rev. 2022, 5, Cd013639. [Google Scholar] [CrossRef] [PubMed]
  96. Kriza, C.; Amenta, V.; Zenié, A.; Panidis, D.; Chassaigne, H.; Urbán, P.; Holzwarth, U.; Sauer, A.V.; Reina, V.; Griesinger, C.B. Artificial intelligence for imaging-based COVID-19 detection: Systematic review comparing added value of AI versus human readers. Eur. J. Radiol. 2021, 145, 110028. [Google Scholar] [CrossRef] [PubMed]
  97. Gudigar, A.; Raghavendra, U.; Nayak, S.; Ooi, C.P.; Chan, W.Y.; Gangavarapu, M.R.; Dharmik, C.; Samanth, J.; Kadri, N.A.; Hasikin, K.; et al. Role of Artificial Intelligence in COVID-19 Detection. Sensors 2021, 21, 8045. [Google Scholar] [CrossRef]
  98. Lessmann, N.; Sánchez, C.I.; Beenen, L.; Boulogne, L.H.; Brink, M.; Calli, E.; Charbonnier, J.P.; Dofferhoff, T.; van Everdingen, W.M.; Gerke, P.K.; et al. Automated Assessment of COVID-19 Reporting and Data System and Chest CT Severity Scores in Patients Suspected of Having COVID-19 Using Artificial Intelligence. Radiology 2021, 298, E18-e28. [Google Scholar] [CrossRef]
  99. Pu, J.; Leader, J.K.; Bandos, A.; Ke, S.; Wang, J.; Shi, J.; Du, P.; Guo, Y.; Wenzel, S.E.; Fuhrman, C.R.; et al. Automated quantification of COVID-19 severity and progression using chest CT images. Eur. Radiol. 2021, 31, 436–446. [Google Scholar] [CrossRef] [PubMed]
  100. Lee, S.; Summers, R.M. Clinical Artificial Intelligence Applications in Radiology: Chest and Abdomen. Radiol. Clin. N. Am. 2021, 59, 987–1002. [Google Scholar] [CrossRef]
  101. Wang, Y.; Chen, Y.; Wei, Y.; Li, M.; Zhang, Y.; Zhang, N.; Zhao, S.; Zeng, H.; Deng, W.; Huang, Z.; et al. Quantitative analysis of chest CT imaging findings with the risk of ARDS in COVID-19 patients: A preliminary study. Ann. Transl. Med. 2020, 8, 594. [Google Scholar] [CrossRef]
  102. Lanza, E.; Muglia, R.; Bolengo, I.; Santonocito, O.G.; Lisi, C.; Angelotti, G.; Morandini, P.; Savevski, V.; Politi, L.S.; Balzarini, L. Quantitative chest CT analysis in COVID-19 to predict the need for oxygenation support and intubation. Eur. Radiol. 2020, 30, 6770–6778. [Google Scholar] [CrossRef]
  103. Yue, H.; Yu, Q.; Liu, C.; Huang, Y.; Jiang, Z.; Shao, C.; Zhang, H.; Ma, B.; Wang, Y.; Xie, G.; et al. Machine learning-based CT radiomics method for predicting hospital stay in patients with pneumonia associated with SARS-CoV-2 infection: A multicenter study. Ann. Transl. Med. 2020, 8, 859. [Google Scholar] [CrossRef]
  104. Mushtaq, J.; Pennella, R.; Lavalle, S.; Colarieti, A.; Steidler, S.; Martinenghi, C.M.A.; Palumbo, D.; Esposito, A.; Rovere-Querini, P.; Tresoldi, M.; et al. Initial chest radiographs and artificial intelligence (AI) predict clinical outcomes in COVID-19 patients: Analysis of 697 Italian patients. Eur. Radiol. 2021, 31, 1770–1779. [Google Scholar] [CrossRef]
  105. Jia, L.L.; Zhao, J.X.; Pan, N.N.; Shi, L.Y.; Zhao, L.P.; Tian, J.H.; Huang, G. Artificial intelligence model on chest imaging to diagnose COVID-19 and other pneumonias: A systematic review and meta-analysis. Eur. J. Radiol. Open 2022, 9, 100438. [Google Scholar] [CrossRef]
  106. Roberts, M.; Driggs, D.; Thorpe, M.; Gilbey, J.; Yeung, M.; Ursprung, S.; Aviles-Rivero, A.I.; Etmann, C.; McCague, C.; Beer, L.; et al. Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans. Nat. Mach. Intell. 2021, 3, 199–217. [Google Scholar] [CrossRef]
  107. Sun, J.; Peng, L.; Li, T.; Adila, D.; Zaiman, Z.; Melton-Meaux, G.B.; Ingraham, N.E.; Murray, E.; Boley, D.; Switzer, S.; et al. Performance of a Chest Radiograph AI Diagnostic Tool for COVID-19: A Prospective Observational Study. Radiol. Artif. Intell. 2022, 4, e210217. [Google Scholar] [CrossRef] [PubMed]
  108. Buttia, C.; Llanaj, E.; Raeisi-Dehkordi, H.; Kastrati, L.; Amiri, M.; Meçani, R.; Taneri, P.E.; Ochoa, S.A.G.; Raguindin, P.F.; Wehrli, F.; et al. Prognostic models in COVID-19 infection that predict severity: A systematic review. Eur. J. Epidemiol. 2023, 38, 355–372. [Google Scholar] [CrossRef] [PubMed]
  109. Chakaya, J.; Khan, M.; Ntoumi, F.; Aklillu, E.; Fatima, R.; Mwaba, P.; Kapata, N.; Mfinanga, S.; Hasnain, S.E.; Katoto, P.; et al. Global Tuberculosis Report 2020—Reflections on the Global TB burden, treatment and prevention efforts. Int. J. Infect. Dis. 2021, 113 (Suppl. S1), S7–S12. [Google Scholar] [CrossRef] [PubMed]
  110. World Health Organization. Chest Radiography in Tuberculosis Detection: Summary of Current WHO Recommendations and Guidance on Programmatic Approaches; Report No. 9241511508; World Health Organization: Geneva, Switzerland, 2016.
  111. World Health Organization. WHO Guidelines Approved by the Guidelines Review Committee. In WHO Consolidated Guidelines on Tuberculosis: Module 2: Screening—Systematic Screening for Tuberculosis Disease; World Health Organization: Geneva, Switzerland, 2021. [Google Scholar]
  112. Hogeweg, L.; Sánchez, C.I.; Maduskar, P.; Philipsen, R.; Story, A.; Dawson, R.; Theron, G.; Dheda, K.; Peters-Bax, L.; Van Ginneken, B. Automatic detection of tuberculosis in chest radiographs using a combination of textural, focal, and shape abnormality analysis. IEEE Trans. Med. Imaging 2015, 34, 2429–2442. [Google Scholar] [CrossRef]
  113. Hwang, E.J.; Park, S.; Jin, K.N.; Kim, J.I.; Choi, S.Y.; Lee, J.H.; Goo, J.M.; Aum, J.; Yim, J.J.; Park, C.M. Development and Validation of a Deep Learning-based Automatic Detection Algorithm for Active Pulmonary Tuberculosis on Chest Radiographs. Clin. Infect. Dis. 2019, 69, 739–747. [Google Scholar] [CrossRef]
  114. Lee, S.; Yim, J.J.; Kwak, N.; Lee, Y.J.; Lee, J.K.; Lee, J.Y.; Kim, J.S.; Kang, Y.A.; Jeon, D.; Jang, M.J.; et al. Deep Learning to Determine the Activity of Pulmonary Tuberculosis on Chest Radiographs. Radiology 2021, 301, 435–442. [Google Scholar] [CrossRef]
  115. Park, J.; Hwang, E.J.; Lee, J.H.; Hong, W.; Nam, J.G.; Lim, W.H.; Kim, J.H.; Goo, J.M.; Park, C.M. Identification of Active Pulmonary Tuberculosis Among Patients With Positive Interferon-Gamma Release Assay Results: Value of a Deep Learning-based Computer-aided Detection System in Different Scenarios of Implementation. J. Thorac. Imaging 2023, 38, 145–153. [Google Scholar] [CrossRef] [PubMed]
  116. Sethanan, K.; Pitakaso, R.; Srichok, T.; Khonjun, S.; Weerayuth, N.; Prasitpuriprecha, C.; Preeprem, T.; Jantama, S.S.; Gonwirat, S.; Enkvetchakul, P.; et al. Computer-aided diagnosis using embedded ensemble deep learning for multiclass drug-resistant tuberculosis classification. Front. Med. 2023, 10, 1122222. [Google Scholar] [CrossRef] [PubMed]
  117. Hwang, E.J.; Jeong, W.G.; David, P.M.; Arentz, M.; Ruhwald, M.; Yoon, S.H. AI for Detection of Tuberculosis: Implications for Global Health. Radiol. Artif. Intell. 2024, 6, e230327. [Google Scholar] [CrossRef] [PubMed]
  118. Geric, C.; Qin, Z.Z.; Denkinger, C.M.; Kik, S.V.; Marais, B.; Anjos, A.; David, P.M.; Ahmad Khan, F.; Trajman, A. The rise of artificial intelligence reading of chest X-rays for enhanced TB diagnosis and elimination. Int. J. Tuberc. Lung Dis. 2023, 27, 367–372. [Google Scholar] [CrossRef] [PubMed]
  119. Creswell, J.; Vo, L.N.Q.; Qin, Z.Z.; Muyoyeta, M.; Tovar, M.; Wong, E.B.; Ahmed, S.; Vijayan, S.; John, S.; Maniar, R.; et al. Early user perspectives on using computer-aided detection software for interpreting chest X-ray images to enhance access and quality of care for persons with tuberculosis. BMC Glob. Public Health 2023, 1, 30. [Google Scholar] [CrossRef]
  120. Brunelli, A.; Salati, M.; Rocco, G.; Varela, G.; Van Raemdonck, D.; Decaluwe, H.; Falcoz, P.E. European risk models for morbidity (EuroLung1) and mortality (EuroLung2) to predict outcome following anatomic lung resections: An analysis from the European Society of Thoracic Surgeons database. Eur. J. Cardiothorac. Surg. 2017, 51, 490–497. [Google Scholar] [CrossRef]
  121. Bolourani, S.; Wang, P.; Patel, V.M.; Manetta, F.; Lee, P.C. Predicting respiratory failure after pulmonary lobectomy using machine learning techniques. Surgery 2020, 168, 743–752. [Google Scholar] [CrossRef] [PubMed]
  122. Bellini, V.; Valente, M.; Del Rio, P.; Bignami, E. Artificial intelligence in thoracic surgery: A narrative review. J. Thorac. Dis. 2021, 13, 6963–6975. [Google Scholar] [CrossRef]
  123. Etienne, H.; Hamdi, S.; Le Roux, M.; Camuset, J.; Khalife-Hocquemiller, T.; Giol, M.; Debrosse, D.; Assouad, J. Artificial intelligence in thoracic surgery: Past, present, perspective and limits. Eur. Respir. Rev. 2020, 29, 200010. [Google Scholar] [CrossRef]
  124. Chang, Y.J.; Hung, K.C.; Wang, L.K.; Yu, C.H.; Chen, C.K.; Tay, H.T.; Wang, J.J.; Liu, C.F. A Real-Time Artificial Intelligence-Assisted System to Predict Weaning from Ventilator Immediately after Lung Resection Surgery. Int. J. Environ. Res. Public Health 2021, 18, 2713. [Google Scholar] [CrossRef] [PubMed]
  125. Li, X.; Zhang, S.; Luo, X.; Gao, G.; Luo, X.; Wang, S.; Li, S.; Zhao, D.; Wang, Y.; Cui, X.; et al. Accuracy and efficiency of an artificial intelligence-based pulmonary broncho-vascular three-dimensional reconstruction system supporting thoracic surgery: Retrospective and prospective validation study. EBioMedicine 2023, 87, 104422. [Google Scholar] [CrossRef]
  126. Toggweiler, S.; Wyler von Ballmoos, M.C.; Moccetti, F.; Douverny, A.; Wolfrum, M.; Imamoglu, Z.; Mohler, A.; Gülan, U.; Kim, W.K. A fully automated artificial intelligence-driven software for planning of transcatheter aortic valve replacement. Cardiovasc. Revasc. Med. 2024, in press. [Google Scholar] [CrossRef]
  127. Too, C.W.; Fong, K.Y.; Hang, G.; Sato, T.; Nyam, C.Q.; Leong, S.H.; Ng, K.W.; Ng, W.L.; Kawai, T. Artificial Intelligence-Guided Segmentation and Path Planning Software for Transthoracic Lung Biopsy. J. Vasc. Interv. Radiol. 2024, 35, 780–789.e781. [Google Scholar] [CrossRef]
  128. Kisting, M.A.; Hinshaw, J.L.; Toia, G.V.; Ziemlewicz, T.J.; Kisting, A.L.; Lee, F.T., Jr.; Wagner, M.G. Artificial Intelligence-Aided Selection of Needle Pathways: Proof-of-Concept in Percutaneous Lung Biopsies. J. Vasc. Interv. Radiol. 2024, 35, 770–779.e771. [Google Scholar] [CrossRef] [PubMed]
  129. Gurgitano, M.; Angileri, S.A.; Rodà, G.M.; Liguori, A.; Pandolfi, M.; Ierardi, A.M.; Wood, B.J.; Carrafiello, G. Interventional Radiology ex-machina: Impact of Artificial Intelligence on practice. Radiol. Med. 2021, 126, 998–1006. [Google Scholar] [CrossRef]
  130. von Ende, E.; Ryan, S.; Crain, M.A.; Makary, M.S. Artificial Intelligence, Augmented Reality, and Virtual Reality Advances and Applications in Interventional Radiology. Diagnostics 2023, 13, 892. [Google Scholar] [CrossRef] [PubMed]
  131. Gao, Y.; Song, Y.; Yin, X.; Wu, W.; Zhang, L.; Chen, Y.; Shi, W. Deep learning-based digital subtraction angiography image generation. Int. J. Comput. Assist. Radiol. Surg. 2019, 14, 1775–1784. [Google Scholar] [CrossRef]
  132. Shademan, A.; Decker, R.S.; Opfermann, J.D.; Leonard, S.; Krieger, A.; Kim, P.C. Supervised autonomous robotic soft tissue surgery. Sci. Transl. Med. 2016, 8, 337–364. [Google Scholar] [CrossRef]
  133. Darbari, A.; Kumar, K.; Darbari, S.; Patil, P.L. Requirement of artificial intelligence technology awareness for thoracic surgeons. Cardiothorac. Surg. 2021, 29, 13. [Google Scholar] [CrossRef]
  134. Hochhegger, B.; Pasini, R.; Roncally Carvalho, A.; Rodrigues, R.; Altmayer, S.; Kayat Bittencourt, L.; Marchiori, E.; Forghani, R. Artificial Intelligence for Cardiothoracic Imaging: Overview of Current and Emerging Applications. Semin. Roentgenol. 2023, 58, 184–195. [Google Scholar] [CrossRef]
  135. Kim, C.; Yang, Z.; Park, S.H.; Hwang, S.H.; Oh, Y.W.; Kang, E.Y.; Yong, H.S. Multicentre external validation of a commercial artificial intelligence software to analyse chest radiographs in health screening environments with low disease prevalence. Eur. Radiol. 2023, 33, 3501–3509. [Google Scholar] [CrossRef]
  136. Maiter, A.; Hocking, K.; Matthews, S.; Taylor, J.; Sharkey, M.; Metherall, P.; Alabed, S.; Dwivedi, K.; Shahin, Y.; Anderson, E.; et al. Evaluating the performance of artificial intelligence software for lung nodule detection on chest radiographs in a retrospective real-world UK population. BMJ Open 2023, 13, e077348. [Google Scholar] [CrossRef] [PubMed]
  137. Topff, L.; Groot Lipman, K.B.W.; Guffens, F.; Wittenberg, R.; Bartels-Rutten, A.; van Veenendaal, G.; Hess, M.; Lamerigts, K.; Wakkie, J.; Ranschaert, E.; et al. Is the generalizability of a developed artificial intelligence algorithm for COVID-19 on chest CT sufficient for clinical use? Results from the International Consortium for COVID-19 Imaging AI (ICOVAI). Eur. Radiol. 2023, 33, 4249–4258. [Google Scholar] [CrossRef]
  138. Keane, P.A.; Topol, E.J. With an eye to AI and autonomous diagnosis. NPJ Digit. Med. 2018, 1, 40. [Google Scholar] [CrossRef]
  139. Kim, D.W.; Jang, H.Y.; Kim, K.W.; Shin, Y.; Park, S.H. Design Characteristics of Studies Reporting the Performance of Artificial Intelligence Algorithms for Diagnostic Analysis of Medical Images: Results from Recently Published Papers. Korean J. Radiol. 2019, 20, 405–410. [Google Scholar] [CrossRef]
  140. Brady, A.P.; Neri, E. Artificial Intelligence in Radiology-Ethical Considerations. Diagnostics 2020, 10, 231. [Google Scholar] [CrossRef]
  141. Seyyed-Kalantari, L.; Zhang, H.; McDermott, M.B.A.; Chen, I.Y.; Ghassemi, M. Underdiagnosis bias of artificial intelligence algorithms applied to chest radiographs in under-served patient populations. Nat. Med. 2021, 27, 2176–2182. [Google Scholar] [CrossRef]
  142. Langlotz, C.P.; Allen, B.; Erickson, B.J.; Kalpathy-Cramer, J.; Bigelow, K.; Cook, T.S.; Flanders, A.E.; Lungren, M.P.; Mendelson, D.S.; Rudie, J.D.; et al. A Roadmap for Foundational Research on Artificial Intelligence in Medical Imaging: From the 2018 NIH/RSNA/ACR/The Academy Workshop. Radiology 2019, 291, 781–791. [Google Scholar] [CrossRef] [PubMed]
  143. Gleeson, F.; Revel, M.P.; Biederer, J.; Larici, A.R.; Martini, K.; Frauenfelder, T.; Screaton, N.; Prosch, H.; Snoeckx, A.; Sverzellati, N.; et al. Implementation of artificial intelligence in thoracic imaging-a what, how, and why guide from the European Society of Thoracic Imaging (ESTI). Eur. Radiol. 2023, 33, 5077–5086. [Google Scholar] [CrossRef] [PubMed]
  144. Shahid, A.; Bazargani, M.H.; Banahan, P.; Mac Namee, B.; Kechadi, T.; Treacy, C.; Regan, G.; MacMahon, P. A Two-Stage De-Identification Process for Privacy-Preserving Medical Image Analysis. Healthcare 2022, 10, 755. [Google Scholar] [CrossRef]
  145. Oakden-Rayner, L. Exploring Large-scale Public Medical Image Datasets. Acad. Radiol. 2020, 27, 106–112. [Google Scholar] [CrossRef]
  146. Dayan, I.; Roth, H.R.; Zhong, A.; Harouni, A.; Gentili, A.; Abidin, A.Z.; Liu, A.; Costa, A.B.; Wood, B.J.; Tsai, C.S.; et al. Federated learning for predicting clinical outcomes in patients with COVID-19. Nat. Med. 2021, 27, 1735–1743. [Google Scholar] [CrossRef]
  147. Banerjee, I.; Bhattacharjee, K.; Burns, J.L.; Trivedi, H.; Purkayastha, S.; Seyyed-Kalantari, L.; Patel, B.N.; Shiradkar, R.; Gichoya, J. “Shortcuts” Causing Bias in Radiology Artificial Intelligence: Causes, Evaluation, and Mitigation. J. Am. Coll. Radiol. 2023, 20, 842–851. [Google Scholar] [CrossRef]
  148. Das, D.; Santosh, K.C.; Pal, U. Cross-Population Train/Test Deep Learning Model: Abnormality Screening in Chest X-Rays. In Proceedings of the 2020 IEEE 33rd International Symposium on Computer-Based Medical Systems (CBMS), Rochester, MN, USA, 28–30 July 2020; pp. 514–519. [Google Scholar]
  149. Cutillo, C.M.; Sharma, K.R.; Foschini, L.; Kundu, S.; Mackintosh, M.; Mandl, K.D. Machine intelligence in healthcare-perspectives on trustworthiness, explainability, usability, and transparency. NPJ Digit. Med. 2020, 3, 47. [Google Scholar] [CrossRef] [PubMed]
  150. Milam, M.E.; Koo, C.W. The current status and future of FDA-approved artificial intelligence tools in chest radiology in the United States. Clin. Radiol. 2023, 78, 115–122. [Google Scholar] [CrossRef]
  151. Gaube, S.; Suresh, H.; Raue, M.; Merritt, A.; Berkowitz, S.J.; Lermer, E.; Coughlin, J.F.; Guttag, J.V.; Colak, E.; Ghassemi, M. Do as AI say: Susceptibility in deployment of clinical decision-aids. NPJ Digit. Med. 2021, 4, 31. [Google Scholar] [CrossRef] [PubMed]
  152. Huang, S.C.; Pareek, A.; Seyyedi, S.; Banerjee, I.; Lungren, M.P. Fusion of medical imaging and electronic health records using deep learning: A systematic review and implementation guidelines. NPJ Digit. Med. 2020, 3, 136. [Google Scholar] [CrossRef] [PubMed]
  153. Park, S.H.; Han, K. Methodologic Guide for Evaluating Clinical Performance and Effect of Artificial Intelligence Technology for Medical Diagnosis and Prediction. Radiology 2018, 286, 800–809. [Google Scholar] [CrossRef] [PubMed]
  154. Topol, E.J. High-performance medicine: The convergence of human and artificial intelligence. Nat. Med. 2019, 25, 44–56. [Google Scholar] [CrossRef]
  155. Lobig, F.; Subramanian, D.; Blankenburg, M.; Sharma, A.; Variyar, A.; Butler, O. To pay or not to pay for artificial intelligence applications in radiology. NPJ Digit. Med. 2023, 6, 117. [Google Scholar] [CrossRef] [PubMed]
  156. Abràmoff, M.D.; Roehrenbeck, C.; Trujillo, S.; Goldstein, J.; Graves, A.S.; Repka, M.X.; Silva Iii, E.Z. A reimbursement framework for artificial intelligence in healthcare. NPJ Digit. Med. 2022, 5, 72. [Google Scholar] [CrossRef]
  157. Geis, J.R.; Brady, A.P.; Wu, C.C.; Spencer, J.; Ranschaert, E.; Jaremko, J.L.; Langer, S.G.; Borondy Kitts, A.; Birch, J.; Shields, W.F.; et al. Ethics of Artificial Intelligence in Radiology: Summary of the Joint European and North American Multisociety Statement. Radiology 2019, 293, 436–440. [Google Scholar] [CrossRef]
  158. Laptev, V.A.; Ershova, I.V.; Feyzrakhmanova, D.R. Medical Applications of Artificial Intelligence (Legal Aspects and Future Prospects). Laws 2022, 11, 3. [Google Scholar] [CrossRef]
  159. Svoboda, E. Artificial intelligence is improving the detection of lung cancer. Nature 2020, 587, S20–S22. [Google Scholar] [CrossRef]
  160. Fromherz, M.R.; Makary, M.S. Artificial intelligence: Advances and new frontiers in medical imaging. Artif. Intell. Med. Imaging 2022, 3, 33–41. [Google Scholar] [CrossRef]
  161. Zaki, H.A.; Aoun, A.; Munshi, S.; Abdel-Megid, H.; Nazario-Johnson, L.; Ahn, S.H. The Application of Large Language Models for Radiologic Decision Making. J. Am. Coll. Radiol. 2024, 21, 1072–1078. [Google Scholar] [CrossRef]
  162. Campbell, W.A.t.; Chick, J.F.B.; Shin, D.; Makary, M.S. Understanding ChatGPT for evidence-based utilization in interventional radiology. Clin. Imaging 2024, 108, 110098. [Google Scholar] [CrossRef] [PubMed]
  163. Savage, C.H.; Park, H.; Kwak, K.; Smith, A.D.; Rothenberg, S.A.; Parekh, V.S.; Doo, F.X.; Yi, P.H. General-Purpose Large Language Models Versus a Domain-Specific Natural Language Processing Tool for Label Extraction From Chest Radiograph Reports. AJR Am. J. Roentgenol. 2024, 222, e2330573. [Google Scholar] [CrossRef]
  164. Fink, M.A.; Bischoff, A.; Fink, C.A.; Moll, M.; Kroschke, J.; Dulz, L.; Heußel, C.P.; Kauczor, H.U.; Weber, T.F. Potential of ChatGPT and GPT-4 for Data Mining of Free-Text CT Reports on Lung Cancer. Radiology 2023, 308, e231362. [Google Scholar] [CrossRef]
  165. Adams, L.C.; Truhn, D.; Busch, F.; Kader, A.; Niehues, S.M.; Makowski, M.R.; Bressem, K.K. Leveraging GPT-4 for Post Hoc Transformation of Free-text Radiology Reports into Structured Reporting: A Multilingual Feasibility Study. Radiology 2023, 307, e230725. [Google Scholar] [CrossRef] [PubMed]
  166. Singh, S.; Elahi, A.; Schweitzer, A.; Adekanmi, A.; Atalabi, O.; Mollura, D.J.; Dako, F. Deploying Artificial Intelligence for Thoracic Imaging Around the World. J. Am. Coll. Radiol. 2023, 20, 859–862. [Google Scholar] [CrossRef]
Figure 1. AI applications in thoracic imaging.
Figure 1. AI applications in thoracic imaging.
Diagnostics 14 01456 g001
Figure 2. Representation of AI workflow in thoracic radiology.
Figure 2. Representation of AI workflow in thoracic radiology.
Diagnostics 14 01456 g002
Figure 3. Limitations and future directions for AI applications in thoracic imaging.
Figure 3. Limitations and future directions for AI applications in thoracic imaging.
Diagnostics 14 01456 g003
Table 1. Benefits and limitations of AI applications in thoracic radiology.
Table 1. Benefits and limitations of AI applications in thoracic radiology.
Applications of AI BenefitsLimitations
Worklist PrioritizationHigh clinical impact
Reduced turnaround time for critical cases
Resource optimization
Liability
Workflow integration
Diagnostic AccuracyImproved diagnostic performance
Reduced reading time
Increased reader confidence
Explainability
Lack of generalization
Bias
Interpretation errors
Quantitative AnalysisEfficiency
Standardization
Data discrepancy
Need for high quality data
RadiomicsPrecision medicine
Prognosis
Thoracic oncology research
Standardization
Workflow integration
Radiation and Contrast ReductionPatient safety
Enhanced image quality
Cost saving
Workflow integration
Algorithm accuracy
Operational EfficiencyResource allocation
Cost saving
Less adverse risk
Workflow integration
COVID-19Early diagnosis and monitoringReal-world validation
Lack of generalization
TB Screening and TriageAddress global health disparities and radiologist shortage
Automated triage
Early detection and treatment
Real-world validation
False-positive rates
Perioperative Care and Interventional RadiologyIncreased precision and accuracy
Patient safety
Risk assessment
Clinical integration
Ethical concerns and liability
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

Chang, J.Y.; Makary, M.S. Evolving and Novel Applications of Artificial Intelligence in Thoracic Imaging. Diagnostics 2024, 14, 1456. https://doi.org/10.3390/diagnostics14131456

AMA Style

Chang JY, Makary MS. Evolving and Novel Applications of Artificial Intelligence in Thoracic Imaging. Diagnostics. 2024; 14(13):1456. https://doi.org/10.3390/diagnostics14131456

Chicago/Turabian Style

Chang, Jin Y., and Mina S. Makary. 2024. "Evolving and Novel Applications of Artificial Intelligence in Thoracic Imaging" Diagnostics 14, no. 13: 1456. https://doi.org/10.3390/diagnostics14131456

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