We aimed to prospectively investigate whether bladder volume measured using deep learning artificial intelligence (AI) algorithms (AI-BV) is more accurate than that measured using conventional methods (C-BV) if using a portable ultrasound bladder scanner (PUBS).

Patients who underwent filling cystometry because of lower urinary tract symptoms between January 2021 and July 2022 were enrolled. Every time the bladder was filled serially with normal saline from 0 mL to maximum cystometric capacity in 50 mL increments, C-BV was measured using PUBS. Ultrasound images obtained during this process were manually annotated to define the bladder contour, which was used to build a deep learning AI model. The true bladder volume (T-BV) for each bladder volume range was compared with C-BV and AI-BV for analysis.

We enrolled 250 patients (213 men and 37 women), and a deep learning AI model was established using 1912 bladder images. There was a significant difference between C-BV (205.5 ± 170.8 mL) and T-BV (190.5 ± 165.7 mL) (p = 0.001), but no significant difference between AI-BV (197.0 ± 161.1 mL) and T-BV (190.5 ± 165.7 mL) (p = 0.081). In bladder volume ranges of 101-150, 151-200, and 201-300 mL, there were significant differences in the percentage of volume differences between [C-BV and T-BV] and [AI-BV and T-BV] (p < 0.05), but no significant difference if converted to absolute values (p > 0.05). C-BV (R2 = 0.91, p < 0.001) and AI-BV (R2 = 0.90, p < 0.001) were highly correlated with T-BV. The mean difference between AI-BV and T-BV (6.5 ± 50.4) was significantly smaller than that between C-BV and T-BV (15.0 ± 50.9) (p = 0.001).

Following image pre-processing, deep learning AI-BV more accurately estimated true BV than conventional methods in this selected cohort on internal validation. Determination of the clinical relevance of these findings and performance in external cohorts requires further study.

The clinical trial was conducted using an approved product for its approved indication, so approval from the Ministry of Food and Drug Safety (MFDS) was not required. Therefore, there is no clinical trial registration number.

Radiologists increasingly use artificial intelligence (AI) to enhance diagnostic accuracy and optimize workflows. However, many lack the technical skills to effectively apply machine learning (ML) and deep learning (DL) algorithms, limiting the accessibility of these methods to radiology researchers who could otherwise benefit from them. Large language models (LLMs), such as GPT-4o, may serve as virtual advisors, offering tailored algorithm recommendations for specific research needs. This study evaluates GPT-4o's effectiveness as a recommender system to enhance radiologists' understanding and implementation of AI in research.

GPT-4o was used to recommend ML and DL algorithms based on specific details provided by researchers, including dataset characteristics, modality types, data sizes, and research objectives. The model acted as a virtual advisor, guiding researchers in selecting the most appropriate models for their studies.

The study systematically evaluated GPT-4o's recommendations for clarity, task alignment, model diversity, and baseline selection. Responses were graded to assess the model's ability to meet the needs of radiology researchers.

GPT-4o effectively recommended appropriate ML and DL algorithms for various radiology tasks, including segmentation, classification, and regression in medical imaging. The model suggested a diverse range of established and innovative algorithms, such as U-Net, Random Forest, Attention U-Net, and EfficientNet, aligning well with accepted practices.

GPT-4o shows promise as a valuable tool for radiologists and early career researchers by providing clear and relevant AI and ML algorithm recommendations. Its ability to bridge the knowledge gap in AI implementation could democratize access to advanced technologies, fostering innovation and improving radiology research quality. Further studies should explore integrating LLMs into routine workflows and their role in ongoing professional development.

Recent developments in artificial intelligence (AI) may significantly alter physiological research and healthcare delivery. Whereas AI applications in medicine have historically been trained for specific tasks, recent technological advances have produced models trained on more diverse datasets with much higher parameter counts. These new, "foundation" models raise the possibility that more flexible AI tools can be applied to a wider set of healthcare tasks than in the past. This review describes how these newer models differ from conventional task-specific AI, which relies heavily on focused datasets and narrow, specific applications. By examining the integration of AI into diagnostic tools, personalized treatment strategies, biomedical research, and healthcare administration, we highlight how these newer models are revolutionizing predictive healthcare analytics and operational workflows. In addition, we address ethical and practical considerations associated with the use of foundation models by highlighting emerging trends, calling for changes to existing guidelines, and emphasizing the importance of aligning AI with clinical goals to ensure its responsible and effective use.

This research aimed to create a dataset of Iliac CTV scans for automated May-Thurner syndrome (MTS) detection using deep learning and radiomics. In addition, it sought to establish an automated segmentation model for Iliac Vein CTV scans and construct a radiomic signature for MTS diagnosis.

We collected a dataset of 490 cases meeting specific inclusion and exclusion criteria, anonymized to comply with HIPAA regulations. Iliac Vein CTV scans were prepared with contrast agent administration, followed by image acquisition and evaluation. A deep learning-based segmentation model, UPerNet, was employed using 10-fold cross-validation. Radiomic features were extracted from the scans and used to construct a diagnostic radiomic signature. Statistical analysis, including Dice values and ROC analysis, was conducted to evaluate segmentation and diagnostic performance.

The dataset consisted of 201 positive cases of MTS and 289 negative cases. The UPerNet segmentation model exhibited remarkable accuracy in identifying MTS regions. A Dice coefficient of 0.925 (95% confidence interval: 0.875-0.961) was observed, indicating the precision and reliability of our segmentation model. Radiomic analysis produced a diagnostic radiomic signature with significant clinical potential. ROC analysis demonstrated promising results, underscoring the efficacy of the developed model in distinguishing MTS cases. The radiomic signature demonstrated strong diagnostic capabilities for MTS. Within the training dataset, it attained a notable area under the curve (AUC) of 0.891, with a 95% confidence interval ranging from 0.825 to 0.956, showcasing its effectiveness. This diagnostic capability extended to the validation dataset, where the AUC remained strong at 0.892 (95% confidence interval: 0.793-0.991). These results highlight the accuracy of our segmentation model and the diagnostic value of our radiomic signature in identifying MTS cases.

This study presents a comprehensive approach to automate MTS detection from Iliac CTV scans, combining deep learning and radiomics. The results suggest the potential clinical utility of the developed model in diagnosing MTS, offering a non-invasive and efficient alternative to traditional methods.

Accurate segmentation of cerebral arteries on computed tomography angiography (CTA) images is essential for the diagnosis and management of cerebrovascular diseases, including ischemic stroke. This study implemented a deep learning-based U-Net++ model for cerebral artery segmentation in CTA images, focusing on optimizing pruning levels by analyzing the trade-off between segmentation performance and computational cost. Dual-energy CTA and direct subtraction CTA datasets were utilized to segment the internal carotid and vertebral arteries in close proximity to the bone. We implemented four pruning levels (L1-L4) in the U-Net++ model and evaluated the segmentation performance using accuracy, intersection over union, F1-score, boundary F1-score, and Hausdorff distance. Statistical analyses were conducted to assess the significance of segmentation performance differences across pruning levels. In addition, we measured training and inference times to evaluate the trade-off between segmentation performance and computational efficiency. Applying deep supervision improved segmentation performance across all factors. While the L4 pruning level achieved the highest segmentation performance, L3 significantly reduced training and inference times (by an average of 51.56 % and 22.62 %, respectively), while incurring only a small decrease in segmentation performance (7.08 %) compared to L4. These results suggest that L3 achieves an optimal balance between performance and computational cost. This study demonstrates that pruning levels in U-Net++ models can be optimized to reduce computational cost while maintaining effective segmentation performance. By simplifying deep learning models, this approach can improve the efficiency of cerebrovascular segmentation, contributing to faster and more accurate diagnoses in clinical settings.

Distinctive facial phenotypes serve as crucial diagnostic markers for many rare genetic diseases. Although AI-driven image recognition achieves high diagnostic accuracy, it often fails to explain its predictions. In this study, we present the Facial phenotype-Gene-Disease Dataset (FGDD), an explainable dataset collected from 509 research publications. It contains 1,147 data records encompassing 197 disease-causing genes, 437 facial phenotypes, and 211 disease entities, with 689 records having disease labels. Each data record represents a patient group and includes demographic information, variation information, and phenotype information. Baseline and explainability validations conducted on FGDD confirmed the dataset's effectiveness. FGDD supports the training of diagnostic models for rare genetic diseases while delivering explainable results, and provides a foundation for exploring intricate connections between genes, diseases, and facial phenotypes.

The measurement of tidal volumes via respiratory-induced surface movements of the upper body has been an objective in medical diagnostics for decades, but a real breakthrough has not yet been achieved. The improvement of measurement technology through new, improved sensor systems and the use of artificial intelligence have given this field of research a new dynamic in recent years and opened up new possibilities. Based on the measurement from a motion capture system, the respiration-induced surface motions of 16 test subjects were examined, and specific motion parameters were calculated. Subsequently, linear regression and a tailored convolutional neural network (CNN) were used to determine tidal volumes from an optimal set of motion parameters. The results showed that the linear regression approach, after individual calibration, could be used in clinical applications for 13/16 subjects (mean absolute error < 150 mL), while the CNN approach achieved this accuracy in 5/16 subjects. Here, the individual subject-specific calibration provides significant advantages for the linear regression approach compared to the CNN, which does not require calibration. A larger dataset may allow for greater confidence in the outcomes of the CNN approach. A CNN model trained on a larger dataset would improve performance and may enable clinical use. However, the database of 16 subjects only allows for low-risk use in home care or sports. The CNN approach can currently be used to monitor respiration in home care or competitive sports, while it has the potential to be used in clinical applications if based on a larger dataset that could be gradually built up. Thus, a CNN could provide tidal volumes, the missing parameter in vital signs monitoring, without calibration.

Deep learning (DL) models for predicting obstructive coronary artery disease (CAD) using myocardial perfusion imaging (MPI) have shown potential for enhancing diagnostic accuracy. However, their ability to maintain consistent performance across institutions and demographics remains uncertain. This study aimed to investigate the generalizability and potential biases of an in-house MPI DL model between two hospital-based cohorts.

We retrospectively included patients from two medical centers in Taiwan who underwent stress/redistribution thallium-201 MPI followed by invasive coronary angiography within 90 days as the reference standard. A polar map-free 3D DL model trained on 928 MPI images from one center to predict obstructive CAD was tested on internal (933 images) and external (3234 images from the other center) validation sets. Diagnostic performance, assessed using area under receiver operating characteristic curves (AUCs), was compared between the internal and external cohorts, demographic groups, and with the performance of stress total perfusion deficit (TPD).

The model showed significantly lower performance in the external cohort compared to the internal cohort in both patient-based (AUC: 0.713 vs. 0.813) and vessel-based (AUC: 0.733 vs. 0.782) analyses, but still outperformed stress TPD (all p < 0.001). The performance was lower in patients who underwent treadmill stress MPI in the internal cohort and in patients over 70 years old in the external cohort.

This study demonstrated adequate performance but also limitations in the generalizability of the DL-based MPI model, along with biases related to stress type and patient age. Thorough validation is essential before the clinical implementation of DL MPI models.

Advancements in digital and precision medicine have fostered the rapid development of artificial intelligence (AI) applications, including machine learning, artificial neural networks (ANN), and deep learning, within the field of dentistry, particularly in imaging diagnosis and treatment. This review examines the progress of AI across various domains of dentistry, focusing on its role in enhancing diagnostics and optimizing treatment for oral diseases such as endodontic disease, periodontal disease, oral implantology, orthodontics, prosthodontic treatment, and oral and maxillofacial surgery. Additionally, it discusses the emerging opportunities and challenges associated with these technologies. The findings indicate that AI can be effectively utilized in numerous aspects of oral healthcare, including prevention, early screening, accurate diagnosis, treatment plan design assistance, treatment execution, follow-up monitoring, and prognosis assessment. However, notable challenges persist, including issues related to inaccurate data annotation, limited capability for fine-grained feature expression, a lack of universally applicable models, potential biases in learning algorithms, and legal risks pertaining to medical malpractice and data privacy breaches. Looking forward, future research is expected to concentrate on overcoming these challenges to enhance the accuracy and applicability of AI in diagnosing and treating oral diseases. This review aims to provide a comprehensive overview of the current state of AI in dentistry and to identify pathways for its effective integration into clinical practice.

The intricate imaging structures, artifacts, and noise present in ultrasound images and videos pose significant challenges for accurate segmentation. Deep learning has recently emerged as a prominent field, playing a crucial role in medical image processing. This paper reviews ultrasound image and video segmentation methods based on deep learning techniques, summarizing the latest developments in this field, such as diffusion and segment anything models as well as classical methods. These methods are classified into four main categories based on the characteristics of the segmentation methods. Each category is outlined and evaluated in the corresponding section. We provide a comprehensive overview of deep learning-based ultrasound image segmentation methods, evaluation metrics, and common ultrasound datasets, hoping to explain the advantages and disadvantages of each method, summarize its achievements, and discuss challenges and future trends.

Given the significance and potential risks associated with Ocular Surface Squamous Neoplasia (OSSN) and the importance of its differentiation from other conditions, we aimed to develop a Deep Learning (DL) model differentiating OSSN from pterygium (PTG) using slit photographs.

A dataset comprising slit photographs of 162 patients including 77 images of OSSN and 85 images of PTG was assembled. After manual segmentation of the images, a Python-based transfer learning approach utilizing the EfficientNet B7 network was employed for automated image segmentation. GoogleNet, a pre-trained neural network was used to categorize the images into OSSN or PTG. To evaluate the performance of our DL model, K-Fold 10 Cross Validation was implemented, and various performance metrics were measured.

There was a statistically significant difference in mean age between the OSSN (63.23 ± 13.74 years) and PTG groups (47.18 ± 11.53) (P-value =.000). Furthermore, 84.41% of patients in the OSSN group and 80.00% of the patients in the PTG group were male. Our classification model, trained on automatically segmented images, demonstrated reliable performance measures in distinguishing OSSN from PTG, with an Area Under Curve (AUC) of 98%, sensitivity, F1 score, and accuracy of 94%, and a Matthews Correlation Coefficient (MCC) of 88%.

This study presents a novel DL model that effectively segments and classifies OSSN from PTG images with a relatively high accuracy. In addition to its clinical use, this model can be potentially used as a telemedicine application.

Machine learning, a common artificial intelligence technology in medical image analysis, enables computers to learn statistical patterns from pairs of data and annotated labels. Supervised learning in machine learning allows the computer to predict how a specific anatomic structure should be segmented in new patients. This study aimed to develop and validate a deep learning algorithm that automatically creates 3-dimensional surface models of human teeth from a cone-beam computed tomography scan.

A multiresolution dataset, including 216 × 272 × 272, 512 × 512 × 512, and 576 × 768 × 768. Ground truth labels for teeth segmentation were generated. Random partitioning was applied to allocate 140 patients to the training set, 40 to the validation set, and 30 scans for testing and model performance evaluation. Different evaluation metrics were used for assessment.

Our teeth identification model has achieved an accuracy of 87.92% ± 4.43% on the test set. The general (binary) teeth segmentation model achieved a notably higher accuracy, segmenting the teeth with 93.16% ± 1.18%.

The success of our model not only validates the efficacy of using artificial intelligence for dental imaging analysis but also sets a promising foundation for future advancements in automated and precise dental segmentation techniques.

Fractures are one of the most common reasons of admission to emergency department affecting individuals of all ages and regions worldwide that can be misdiagnosed during radiologic examination. Accurate and timely diagnosis of fracture is crucial for patients, and artificial intelligence that uses algorithms to imitate human intelligence to aid or enhance human performs is a promising solution to address this issue. In the last few years, numerous commercially available algorithms have been developed to enhance radiology practice and a large number of studies apply artificial intelligence to fracture detection. Recent contributions in literature have described numerous advantages showing how artificial intelligence performs better than doctors who have less experience in interpreting musculoskeletal X-rays, and assisting radiologists increases diagnostic accuracy and sensitivity, improves efficiency, and reduces interpretation time. Furthermore, algorithms perform better when they are trained with big data on a wide range of fracture patterns and variants and can provide standardized fracture identification across different radiologist, thanks to the structured report. In this review article, we discuss the use of artificial intelligence in fracture identification and its benefits and disadvantages. We also discuss its current potential impact on the field of radiology and radiomics.

The purpose of this study was to create and test a deep learning algorithm that could identify and distinguish between membranes attached to optic disc [OD; retinal detachment (RD)/posterior vitreous detachment (PVD)] based on ocular ultrasonography (USG).

We obtained a database of B-scan ultrasonography from a high-volume imaging center. A transformer-based Vision Transformer (ViT) model was employed, pre-trained on ImageNet21K, to classify ultrasound B-scan images into healthy, RD, and PVD. Images were pre-processed using Hugging Face's AutoImage Processor for standardization. Labels were mapped to numerical values, and the dataset was split into training and validation (505 samples), and testing (212 samples) subsets to evaluate model performance. Alternate methods, such as ensemble strategies and object detection pipelines, were explored but showed limited improvement in classification accuracy.

The AI model demonstrated high classification performance, achieving an accuracy of 98.21% for PVD, 97.22% for RD, and 95.83% for normal cases. Sensitivity was 98.21% for PVD, 96.55% for RD, and 92.86% for normal cases, while specificity reached 95.16%, 100%, and 95.42%, respectively. Despite the overall strong performance, some misclassification occurred, with seven instances of RD being incorrectly labeled as PVD.

We developed a transformer-based deep learning algorithm for ocular ultrasonography that accurately identifies membranes attached to the optic disc, distinguishing between RD (97.22% accuracy) and PVD (98.21% accuracy). Despite seven misclassifications, our model demonstrates robust performance and enhances diagnostic efficiency in high-volume imaging settings, thereby facilitating timely referrals and ultimately improving patient outcomes in urgent care scenarios. Overall, this promising innovation shows potential for clinical adoption.

Novel non-invasive machine learning algorithms may improve accuracy and reduce the need for biopsy when diagnosing idiopathic pulmonary fibrosis (IPF). We conducted a cost-effectiveness analysis of diagnostic strategies for IPF.

We developed a decision analytic model to evaluate diagnostic strategies for IPF in the United States. To assess the full spectrum of costs and benefits, we compared four interventions: a machine learning diagnostic algorithm, a genomic classifier, a biopsy-all strategy, and a treat-all strategy. The analysis was conducted from the health sector perspective with a lifetime horizon. The primary outcome measures were costs, Quality-Adjusted Life-Years (QALYs) gained, and Incremental Cost-Effectiveness Ratios (ICERs) based on the average of 10,000 probabilistic runs of the model.

Compared to a biopsy-all strategy the machine learning algorithm and genomic classifer reduced diagnostic-related costs by $14,876 and $3,884, respectively. Use of the machine learning algorithm consistently reduced diagnostic costs. When including downstream treatment costs and benefits of anti-fibrotic treatment, the machine learning algorithm had an ICER of $331,069 per QALY gained compared to the biopsy-all strategy. The genomic classifier had a higher ICER of $390,043 per QALY gained, while the treat-all strategy had the highest ICER of $3,245,403 per QALY gained. Results were sensitive to changes in various input parameters including IPF treatment costs, sensitivity and specificity of novel screening tools, and the rate of additional diagnostics following inconclusive results. High treatment costs were found to drive overall cost regardless of the diagnostic method. As treatment costs lowered, the supplemental diagnostic tools became increasingly cost-effective.

Novel tools for diagnosing IPF reduced diagnostic costs, while overall incremental cost-effectiveness ratios were high due to treatment costs. New IPF diagnosis approaches may become more favourable with lower-cost treatments for IPF.

Oral cavity cancer is associated with high morbidity and mortality, particularly with advanced stage diagnosis. Oral cavity cancer, typically squamous cell carcinoma (OSCC), is often preceded by oral potentially malignant disorders (OPMDs), which comprise eleven disorders with variable risks for malignant transformation. While OPMDs are clinical diagnoses, conventional oral exam followed by biopsy and histopathological analysis is the gold standard for diagnosis of OSCC. There is vast heterogeneity in the clinical presentation of OPMDs, with possible visual similarities to early-stage OSCC or even to various benign oral mucosal abnormalities. The diagnostic challenge of OSCC/OPMDs is compounded in the non-specialist or primary care setting. There has been significant research interest in technology to assist in the diagnosis of OSCC/OPMDs. Artificial intelligence (AI), which enables machine performance of human tasks, has already shown promise in several domains of medical diagnostics. Computer vision, the field of AI dedicated to the analysis of visual data, has over the past decade been applied to clinical photographs for the diagnosis of OSCC/OPMDs. Various methodological concerns and limitations may be encountered in the literature on OSCC/OPMD image analysis. This narrative review delineates the current landscape of AI clinical photograph analysis in the diagnosis of OSCC/OPMDs and navigates the limitations, methodological issues, and clinical workflow implications of this field, providing context for future research considerations.

The necessity for explainability of artificial intelligence technologies in medical applications has been widely discussed and heavily debated within the literature. This paper comprises a systematized review of the arguments supporting and opposing this purported necessity. Both sides of the debate within the literature are quoted to synthesize discourse on common recurring themes and subsequently critically analyze and respond to it. While the use of autonomous black box algorithms is compellingly discouraged, the same cannot be said for the whole of medical artificial intelligence technologies that lack explainability. We contribute novel comparisons of unexplainable clinical artificial intelligence tools, diagnosis of idiopathy, and diagnoses by exclusion, to analyze implications on patient autonomy and informed consent. Applying a novel approach using comparisons with clinical practice guidelines, we contest the claim that lack of explainability compromises clinician due diligence and undermines epistemological responsibility. We find it problematic that many arguments in favour of the practical, ethical, or legal necessity of clinical artificial intelligence explainability conflate the use of unexplainable AI with automated decision making, or equate the use of clinical artificial intelligence with the exclusive use of clinical artificial intelligence.

The hypothesis of changing network layers to increase the accuracy of dose distribution prediction, instead of expanding their dimensions, which requires complex calculations, has been considered in our study.

A total of 137 prostate cancer patients treated with the tomotherapy technique were categorized as 80% training and validating as well as 20% testing for the nested UNet and UNet architectures. Mean absolute error (MAE) was used to measure the dosimetry indices of dose-volume histograms (DVHs), and geometry indices, including the structural similarity index measure (SSIM), dice similarity coefficient (DSC), and Jaccard similarity coefficient (JSC), were used to evaluate the isodose volume (IV) similarity prediction. To verify a statistically significant difference, the two-way statistical Wilcoxon test was used at a level of 0.05 (p < 0.05).

Use of a nested UNet architecture reduced the predicted dose MAE in DVH indices. The MAE for planning target volume (PTV), bladder, rectum, and right and left femur were D98% = 1.11 ± 0.90; D98% = 2.27 ± 2.85, Dmean = 0.84 ± 0.62; D98% = 1.47 ± 12.02, Dmean = 0.77 ± 1.59; D2% = 0.65 ± 0.70, Dmean = 0.96 ± 2.82; and D2% = 1.18 ± 6.65, Dmean = 0.44 ± 1.13, respectively. Additionally, the greatest geometric similarity was observed in the mean SSIM for UNet and nested UNet (0.91 vs. 0.94, respectively).

The nested UNet network can be considered a suitable network due to its ability to improve the accuracy of dose distribution prediction compared to the UNet network in an acceptable time.

The rapid advancement of technology has brought significant changes to various fields, including medical imaging (MI). This discussion paper explores the integration of computing technologies (e.g. Python and MATLAB), digital image processing (e.g. image enhancement, segmentation and three-dimensional reconstruction) and artificial intelligence (AI) into the undergraduate MI curriculum. By examining current educational practices, gaps and limitations that hinder the development of future-ready MI professionals are identified. A comprehensive curriculum framework is proposed, incorporating essential computational skills, advanced image processing techniques and state-of-the-art AI tools, such as large language models like ChatGPT. The proposed curriculum framework aims to improve the quality of MI education significantly and better equip students for future professional practice and challenges while enhancing diagnostic accuracy, improving workflow efficiency and preparing students for the evolving demands of the MI field.

Machine Learning (ML) has become an essential tool for analyzing biomedical data, facilitating the prediction of treatment outcomes and patient survival. However, the effectiveness of ML models heavily relies on both the choice of algorithms and the quality of the input data. In this study, we aimed to develop a novel predictive model to estimate individual survival for patients diagnosed with glioblastoma (GBM), focusing on key variables such as O6-Methylguanine-DNA Methyltransferase (MGMT) methylation status, age, and sex.

To identify the optimal approach, we utilized retrospective data from 218 patients treated at our brain tumor center. The performance of the ML models was evaluated within repeated tenfold regression. The pipeline comprised five regression estimators, including both linear and non-linear algorithms. Permutation feature importance highlighted the feature with the most significant impact on the model. Statistical significance was assessed using a permutation test procedure.

The best machine learning algorithm achieved a mean absolute error (MAE) of 12.65 (SD = ± 2.18) and an explained variance (EV) of 7% (SD = ± 1.8%) with p < 0.001. Linear algorithms led to more accurate predictions than non-linear estimators. Feature importance testing indicated that age and positive MGMT-methylation influenced the predictions the most.

In summary, here we provide a novel approach allowing to predict GBM patient's survival in months solely based on key parameters such as age, sex and MGMT-methylation status and underscores MGMT-methylation status as key prognostic factor for GBM patients survival.

Artificial intelligence (AI) transforms image data analysis across many biomedical fields, such as cell biology, radiology, pathology, cancer biology, and immunology, with object detection, image feature extraction, classification, and segmentation applications. Advancements in deep learning (DL) research have been a critical factor in advancing computer techniques for biomedical image analysis and data mining. A significant improvement in the accuracy of cell detection and segmentation algorithms has been achieved as a result of the emergence of open-source software and innovative deep neural network architectures. Automated cell segmentation now enables the extraction of quantifiable cellular and spatial features from microscope images of cells and tissues, providing critical insights into cellular organization in various diseases. This review aims to examine the latest AI and DL techniques for cell analysis and data mining in microscopy images, aid the biologists who have less background knowledge in AI and machine learning (ML), and incorporate the ML models into microscopy focus images.

Cardiovascular diseases and cancer are the leading causes of morbidity and mortality in modern society. Expanding cancer therapies that have improved prognosis may also be associated with cardiotoxicity, and extended life span after survivorship is associated with the increasing prevalence of cardiovascular disease. As such, the field of cardio-oncology has been rapidly expanding, with an aim to identify cardiotoxicity and cardiac disease early in a patient who is receiving treatment for cancer or is in survivorship. Artificial intelligence is revolutionizing modern medicine with its ability to identify cardiac disease early. This article comprehensively reviews applications of artificial intelligence specifically applied to electrocardiograms, echocardiography, cardiac magnetic resonance imaging, and nuclear imaging to predict cardiac toxicity in the setting of cancer therapies, with a view to reduce early complications and cardiac side effects from cancer therapies such as chemotherapy, radiation therapy, or immunotherapy.

Postencephalitic epilepsy (PEE) is a severe neurological complication following encephalitis. Early identification of individuals at high risk for PEE is important for timely intervention.

To develop a large self-supervised vision foundation model using a big dataset of multi-contrast head MRI scans, followed by fine-tuning with MRI data and follow-up outcomes from patients with PEE to develop a PEE association model.

Retrospective.

Fifty-seven thousand six hundred twenty-one contrast-enhanced head MRI scans from 34,871 patients for foundation model construction, and head MRI scans from 144 patients with encephalitis (64 PEE, 80 N-PEE) for the PEE association model.

1.5-T, 3-T, T1-weighted imaging, T2-weighted imaging, fluid attenuated inversion recovery, T1-weighted contrast-enhanced imaging.

The foundation model was developed using self-supervised learning and cross-contrast context recovery. Patients with encephalitis were monitored for a median of 3.7 years (range 0.7-7.5 years), with epilepsy diagnosed according to International League Against Epilepsy. Occlusion sensitivity mapping highlighted brain regions involved in PEE classifications. Model performance was compared with DenseNet without pre-trained weights.

Performance was assessed via confusion matrices, accuracy, sensitivity, specificity, precision, F1 score, and area under the receiver operating characteristic curve (AUC). The DeLong test evaluated AUC between the two models (P < 0.05 for statistical significance).

The PEE association model achieved accuracy, sensitivity, specificity, precision, F1 score, and AUC of 79.3% (95% CI: 0.71-0.92), 92.3% (95% CI: 0.80-1.00), 68.8% (95% CI: 0.55-0.87), 70.6% (95% CI: 0.61-0.90), 80.0% (95% CI: 0.71-0.93), and 81.0% (95% CI: 0.68-0.92), respectively. A significant AUC improvement was found compared to DenseNet (Delong test, P = 0.03). The association model focused on brain regions affected by encephalitis.

Using extensive unlabeled data via self-supervised learning addressed the limitations of supervised tasks with limited data. The fine-tuned foundation model outperformed DenseNet, which was trained exclusively on task data.

This research develops a model to assess the occurrence epilepsy after encephalitis, a severe brain inflammation condition. By using over 57,000 brain scans, the study trains a computer program to recognize patterns in brain images. The model analyzes whole-brain scans to identify areas commonly affected by the disease, such as the temporal and frontal lobes. It was tested on data from patients with encephalitis and showed better performance than older methods. The model can assess the risk of secondary epilepsy in patients with encephalitis, allowing doctors to intervene early and improve treatment outcomes for those affected by this condition.

4 TECHNICAL EFFICACY: Stage 1.

Challenges persist in achieving automatic and efficient inflammation quantification using dynamic contrast-enhanced (DCE) MRI in rheumatoid arthritis (RA) patients.

To investigate an automatic artificial intelligence (AI) approach and an optimized dynamic MRI protocol for quantifying disease activity in RA in whole hands while excluding arterial pixels.

Retrospective.

Twelve RA patients underwent DCE-MRI with 27 phases for creating the AI model and tested on images with a variable number of phases from 35 RA patients.

3.0 T/DCE T1-weighted gradient echo sequence (mDixon, water image).

The model was trained with various DCE-MRI time-intensity number of phases. Evaluations were conducted for similarity between AI segmentation and manual outlining in 51 ROIs with synovitis. The relationship between synovial volume via AI segmentation with rheumatoid arthritis magnetic resonance imaging scoring (RAMRIS) across whole hands was then evaluated. The reference standard was determined by an experienced musculoskeletal radiologist.

Area under the curve (AUC) of receiver operating characteristic (ROC), Dice and Spearman's rank correlation coefficients, and interclass correlation coefficient (ICC). A P-value <0.05 was considered statistically significant.

A minimum of 15 phases (acquisition time at least 2.5 minutes) was found to be necessary. AUC ranged from 0.941 ± 0.009 to 0.965 ± 0.009. The Dice score was 0.557-0.615. Spearman's correlation coefficients between the AI model and ground truth were 0.884-0.927 and 0.736-0.831, for joint ROIs and whole hands, respectively. The Spearman's correlation coefficient for the additional test set between the model trained with 15 phases and RAMRIS was 0.768.

The AI-based classification model effectively identified synovitis pixels while excluding arteries. The optimal performance was achieved with at least 15 phases, providing a quantitative assessment of inflammatory activity in RA while minimizing acquisition time.

3 TECHNICAL EFFICACY: Stage 2.

To explore the predictive value of traditional machine learning (ML) and deep learning (DL) algorithms based on computed tomography pulmonary angiography (CTPA) images for short-term adverse outcomes in patients with acute pulmonary embolism (APE). This retrospective study enrolled 132 patients with APE confirmed by CTPA. Thrombus segmentation and texture feature extraction was performed using 3D-Slicer software. The least absolute shrinkage and selection operator (LASSO) algorithm was used for feature dimensionality reduction and selection, with optimal λ values determined using leave-one-fold cross-validation to identify texture features with non-zero coefficients. ML models (logistic regression, random forest, decision tree, support vector machine) and DL models (ResNet 50 and Vgg 19) were used to construct the prediction models. Model performance was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC). The cohort included 84 patients in the good prognosis group and 48 patients in the poor prognosis group. Univariate and multivariate logistic regression analyses showed that diabetes, RV/LV ≥ 1.0, and Qanadli index form independent risk factors predicting poor prognosis in patients with APE(P < 0.05). A total of 750 texture features were extracted, with 4 key features identified through screening. There was a weak positive correlation between texture features and clinical parameters. ROC curves analysis demonstrated AUC values of 0.85 (0.78-0.92), 0.76 (0.67-0.84), and 0.89 (0.83-0.95) for the clinical, texture feature, and combined models, respectively. In the ML models, the random forest model achieved the highest AUC (0.85), and the support vector machine model achieved the lowest AUC (0.62). And the AUCs for the DL models (ResNet 50 and Vgg 19) were 0.91 (95%CI: 0.90-0.92) and 0.94(95%CI: 0.93-0.95), respectively. Vgg 19 model demonstrated exceptional precision (0.93), recall (0.76), specificity (0.95) and F1 score (0.84). Both ML and DL models based on thrombus texture features from CTPA images demonstrated higher predictive efficacy for short-term adverse outcomes in patients with APE, especially the random forest and Vgg 19 models, potentially assisting clinical management in timely interventions to improve patient prognosis.

The future of artificial intelligence (AI) safety is expected to include bias mitigation methods from development to application. The complexity and integration of these methods could grow in conjunction with advances in AI and human-AI interactions. Numerous methods are being proposed to mitigate bias, but without a structured way to compare their strengths and weaknesses. In this work, we present two approaches to systematically amplify subgroup performance bias. These approaches allow for the evaluation and comparison of the effectiveness of bias mitigation methods on AI models by varying the degrees of bias, and can be applied to any classification model. We used these approaches to compare four off-the-shelf bias mitigation methods. Both amplification approaches promote the development of learning shortcuts in which the model forms associations between patient attributes and AI output. We demonstrate these approaches in a case study, evaluating bias in the determination of COVID status from chest x-rays. The maximum achieved increase in performance bias, measured as a difference in predicted prevalence, was 72% and 32% for bias between subgroups related to patient sex and race, respectively. These changes in predicted prevalence were not accompanied by substantial changes in the differences in subgroup area under the receiver operating characteristic curves, indicating that the increased bias is due to the formation of learning shortcuts, not a difference in ability to distinguish positive and negative patients between subgroups.

Vertebral compression fractures (VCFs) are the most common type of osteoporotic fractures, yet they are often clinically silent and undiagnosed. Chest frontal radiographs (CFRs) are frequently used in clinical practice and a portion of VCFs can be detected through this technology. This study aimed to develop an automatic artificial intelligence (AI) tool using deep learning (DL) model for the opportunistic screening of VCFs from CFRs. The datasets were collected from four medical centers, comprising 19,145 vertebrae (T6-T12) from 2735 patients. Patients from Center 1, 2 and 3 were divided into the training and internal testing datasets in an 8:2 ratio (n = 2361, with 16,527 vertebrae). Patients from Center 4 were used as the external test dataset (n = 374, with 2618 vertebrae). Model performance was assessed using sensitivity, specificity, accuracy and the area under the curve (AUC). A reader study with five clinicians of different experience levels was conducted with and without AI assistance. In the internal testing dataset, the model achieved a sensitivity of 83.0 % and an AUC of 0.930 at the fracture level. In the external testing dataset, the model demonstrated a sensitivity of 78.4 % and an AUC of 0.942 at the fracture level. The model's sensitivity outperformed that of five clinicians with different levels of experience. Notably, AI assistance significantly improved sensitivity at the patient level for both junior clinicians (from 56.1 % without AI to 81.6 % with AI) and senior clinicians (from 65.0 % to 85.6 %). In conclusion, the automatic AI tool significantly increases clinicians' sensitivity in diagnosing fractures on CFRs, showing great potential for the opportunistic screening of VCFs.

Vector-borne diseases (VBDs) represent a critical global public health concern, with approximately 80% of the world's population at risk of one or more VBD. Manual disease vector identification is time-consuming and expert-dependent, hindering disease control efforts. Deep learning (DL), widely used in image, text, and audio tasks, offers automation potential for disease vector identification. This paper explores the substantial potential of combining DL with disease vector identification. Our aim is to comprehensively summarize the current status of DL in disease vector identification, covering data collection, data preprocessing, model construction, evaluation methods, and applications in identification spanning from species classification to object detection and breeding site identification. We also discuss the challenges and possible prospects for DL in disease vector identification for further research. © 2024 Society of Chemical Industry.

Artificial intelligence is based on algorithms that enable machines to perform tasks and activities that generally require human intelligence, and its use offers innovative solutions in various fields. Machine learning, a subset of artificial intelligence, concentrates on empowering computers to learn and enhance from data autonomously; this narrative review seeks to elucidate the utilization of artificial intelligence in fostering physical activity, training, exercise, and health outcomes, addressing a significant gap in the comprehension of practical applications.

Only Randomized Controlled Trials (RCTs) published in English were included. Inclusion criteria: all RCTs that use artificial intelligence to program, supervise, manage, or assist physical activity, training, exercise, or health programs. Only studies published from January 1, 2014, were considered. Exclusion criteria: all the studies that used robot-assisted, robot-supported, or robotic training were excluded.

A total of 1772 studies were identified. After the first stage, where the duplicates were removed, 1,004 articles were screened by title and abstract. A total of 24 studies were identified, and finally, after a full-text review, 15 studies were identified as meeting all eligibility criteria for inclusion. The findings suggest that artificial intelligence holds promise in promoting physical activity across diverse populations, including children, adolescents, adults, older adult, and individuals with disabilities.

Our research found that artificial intelligence, machine learning and deep learning techniques were used: (a) as part of applications to generate automatic messages and be able to communicate with users; (b) as a predictive approach and for gesture and posture recognition; (c) as a control system; (d) as data collector; and (e) as a guided trainer.

Breast cancer has reached the highest incidence rate worldwide among all malignancies since 2020. Breast imaging plays a significant role in early diagnosis and intervention to improve the outcome of breast cancer patients. In the past decade, deep learning has shown remarkable progress in breast cancer imaging analysis, holding great promise in interpreting the rich information and complex context of breast imaging modalities. Considering the rapid improvement in deep learning technology and the increasing severity of breast cancer, it is critical to summarize past progress and identify future challenges to be addressed. This paper provides an extensive review of deep learning-based breast cancer imaging research, covering studies on mammograms, ultrasound, magnetic resonance imaging, and digital pathology images over the past decade. The major deep learning methods and applications on imaging-based screening, diagnosis, treatment response prediction, and prognosis are elaborated and discussed. Drawn from the findings of this survey, we present a comprehensive discussion of the challenges and potential avenues for future research in deep learning-based breast cancer imaging.

Triple-negative breast cancer (TNBC) is a unique breast cancer subtype characterized by the lack of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression in tumor cells. TNBC represents about 15% to 20% of all breast cancers and is aggressive and highly malignant. Currently, TNBC diagnosis primarily depends on pathological examination, while treatment efficacy is assessed through imaging, biomarker detection, pathological evaluation, and clinical symptom improvement. Among these, biomarker detection and pathological assessments are invasive, time-intensive procedures that may be difficult for patients with severe comorbidities and high complication risks. Thus, there is an urgent need for new, supportive tools in TNBC diagnosis and treatment. Deep learning and radiomics techniques represent advanced machine learning methodologies and are also emerging outcomes in the medical-engineering field in recent years. They are extensions of conventional imaging diagnostic methods and have demonstrated tremendous potential in image segmentation, reconstruction, recognition, and classification. These techniques hold certain application prospects for the diagnosis of TNBC, assessment of treatment response, and long-term prognosis prediction. This article reviews recent progress in the application of deep learning, ultrasound, MRI, and radiomics for TNBC diagnosis and treatment, based on research from both domestic and international scholars.

Cholangiocarcinoma (CCA) is a malignant biliary system tumor and the second most common primary hepatic neoplasm, following hepatocellular carcinoma. CCA still has an extremely high unfavorable prognosis, regardless of type and location, and complete surgical resection remains the only curative therapeutic option; however, due to the underhanded onset and rapid progression of CCA, most patients present with advanced stages at first diagnosis, with only 30 to 60% of CCA patients eligible for surgery. Recent innovations in medical imaging combined with the use of radiomics and artificial intelligence (AI) can lead to improvements in the early detection, characterization, and pre-treatment staging of these tumors, guiding clinicians to make personalized therapeutic strategies. The aim of this review is to provide an overview of how radiological features of CCA can be analyzed through radiomics and with the help of AI for many different purposes, such as differential diagnosis, the prediction of lymph node metastasis, the defining of prognostic groups, and the prediction of early recurrence. The combination of radiomics with AI has immense potential. Still, its effectiveness in practice is yet to be validated by prospective multicentric studies that would allow for the development of standardized radiomics models.

To construct a prediction model based on deep learning (DL) and radiomics features of diffusion weighted imaging (DWI), and clinical variables for evaluating TP53 mutations in endometrial cancer (EC).

DWI and clinical data from 155 EC patients were included in this study, consisting of 80 in the training set, 35 in the test set, and 40 in the external validation set. Radiomics features, convolutional neural network-based DL features, and clinical variables were analyzed. Feature selection was performed using Mann-Whitney U test, LASSO regression, and SelectKBest. Prediction models were established by gaussian process (GP) and decision tree (DT) algorithms and evaluated by the area under the receiver operating characteristic curve (AUC), net reclassification index (NRI), calibration curves, and decision curve analysis (DCA).

Compared to the DL (AUCtraining = 0.830, AUCtest = 0.779, and AUCvalidation = 0.711), radiomics (AUCtraining = 0.810, AUCtest = 0.710, and AUCvalidation = 0.839), and clinical (AUCtraining = 0.780, AUCtest = 0.685, and AUCvalidation = 0.695) models, the combined model based on the GP algorithm, which consisted of four DL features, five radiomics features, and two clinical variables, not only demonstrated the highest diagnostic efficacy (AUCtraining = 0.949, AUCtest = 0.877, and AUCvalidation = 0.914) but also led to an improvement in risk reclassification of the TP53 mutation (NIRtraining = 66.38%, 56.98%, and 83.48%, NIRtest = 50.72%, 80.43%, and 89.49%, and NIRvalidation = 64.58%, 87.50%, and 120.83%, respectively). In addition, the combined model exhibited good agreement and clinical utility in calibration curves and DCA analyses, respectively.

A prediction model based on the GP algorithm and consisting of DL and radiomics features of DWI as well as clinical variables can effectively assess TP53 mutation in EC.

The hybrid CNN-transformer structures harness the global contextualization of transformers with the local feature acuity of CNNs, propelling medical image segmentation to the next level. However, the majority of research has focused on the design and composition of hybrid structures, neglecting the data structure, which enhance segmentation performance, optimize resource efficiency, and bolster model generalization and interpretability. In this work, we propose a data-oriented octree inverse hierarchical order aggregation hybrid transformer-CNN (nnU-OctTN), which focuses on delving deeply into the data itself to identify and harness potential. The nnU-OctTN employs the U-Net as a foundational framework, with the node aggregation transformer serving as the encoder. Data features are stored within an octree data structure with each node computed autonomously yet interconnected through a block-to-block local information exchange mechanism. Oriented towards multi-resolution feature data map learning, a cross-fusion module has been designed that associates the encoder and decoder in a staggered vertical and horizontal approach. Inspired by nnUNet, our framework automatically adapts network parameters to the dataset instead of using pre-trained weights for initialization. The nnU-OctTN method was evaluated on the BTCV, ACDC, and BraTS datasets and achieved excellent performance with dice score coefficient (DSC) 86.95, 92.82, and 90.61, respectively, demonstrating its generalizability and effectiveness. Cross-fusion module effectiveness and model scalability are validated through ablation experiments on BTCV and Kidney. Extensive qualitative and quantitative experimental results demonstrate that nnU-OctTN achieves high-quality 3D medical segmentation that has competitive performance against current state-of-the-art methods, providing a promising idea for clinical applications.

Recently, deep learning has become a popular area of research, and has revolutionized the diagnosis and prediction of ocular diseases, especially fundus diseases. This study aimed to conduct a bibliometric analysis of deep learning in the field of ophthalmology to describe international research trends and examine the current research directions.

This cross-sectional bibliometric analysis examined the development of research on deep learning in the field of ophthalmology and its sub-topics from 2015 to 2024. Visualization of similarities (VOS)-viewer was used to analyze and evaluate 3,055 articles. Data from the articles were collected on a specific date (September 11, 2024) and downloaded from the Web of Science Core Collection (WOSCC) in plain-text format.

A total of 3,055 relevant articles on the WOSCC published from 2015 to 2024 were included in this analysis. The first article on the application of deep learning to ophthalmology was published in 2015, and the number of articles on the subject has grown significantly since 2019. China was the most productive country (n=1,187), followed by the United States (n=673). Sun Yat-sen University was the institution with the most publications. Cheng and Bogunovic were the most frequently published authors. The following four different clusters were identified based on a co-occurrence cluster analysis of high-frequency keywords: (I) deep learning for the segmentation and feature extraction of ophthalmic images; (II) deep learning for the automatic detection and classification of ophthalmic images; (III) application of deep learning to ophthalmic imaging techniques; and (IV) deep learning for the diagnosis and management of ophthalmic diseases.

The analysis of fundus images and the clinical application of deep learning techniques have emerged as prominent research areas in the field of ophthalmology. The substantial increase in publications and citations signifies the expanding impact and global collaboration in the application of deep learning research to ophthalmology. By identifying four distinct clusters representing sub-topics in deep learning ophthalmology research, this study contributes to the understanding of current trends and potential future advancements in the field.

To evaluate the performance of machine learning and then deep learning to detect a systemic scleroderma (SSc) landscape from the same set of nailfold capillaroscopy (NC) images from the French prospective multicenter observational study SCLEROCAP.

NC images from the first 100 SCLEROCAP patients were analyzed to assess the performance of machine learning and then deep learning in identifying the SSc landscape, the NC images having previously been independently and consensually labeled by expert clinicians. Images were divided into a training set (70 %) and a validation set (30 %). After features extraction from the NC images, we tested six classifiers (random forests (RF), support vector machine (SVM), logistic regression (LR), light gradient boosting (LGB), extreme gradient boosting (XGB), K-nearest neighbors (KNN)) on the training set with five different combinations of the images. The performance of each classifier was evaluated by the F1 score. In the deep learning section, we tested three pre-trained models from the TIMM library (ResNet-18, DenseNet-121 and VGG-16) on raw NC images after applying image augmentation methods.

With machine learning, performance ranged from 0.60 to 0.73 for each variable, with Hu and Haralick moments being the most discriminating. Performance was highest with the RF, LGB and XGB models (F1 scores: 0.75-0.79). The highest score was obtained by combining all variables and using the LGB model (F1 score: 0.79 ± 0.05, p < 0.01). With deep learning, performance reached a minimum accuracy of 0.87. The best results were obtained with the DenseNet-121 model (accuracy 0.94 ± 0.02, F1 score 0.94 ± 0.02, AUC 0.95 ± 0.03) as compared to ResNet-18 (accuracy 0.87 ± 0.04, F1 score 0.85 ± 0.03, AUC 0.87 ± 0.04) and VGG-16 (accuracy 0.90 ± 0.03, F1 score 0.91 ± 0.02, AUC 0.91 ± 0.04).

By using machine learning and then deep learning on the same set of labeled NC images from the SCLEROCAP study, the highest performances to detect SSc landscape were obtained with deep learning and in particular DenseNet-121. This pre-trained model could therefore be used to automatically interpret NC images in case of suspected SSc. This result nevertheless needs to be confirmed on a larger number of NC images.

Early identification of pancreatitis remains a significant clinical diagnostic challenge that impacts patient outcomes. The evolution of quantitative imaging followed by deep learning models has shown great promise in the non-invasive diagnosis of pancreatitis and its complications. We provide an overview of advancements in diagnostic imaging and quantitative imaging methods along with the evolution of artificial intelligence (AI). In this article, we review the current and future states of methodology and limitations of AI in improving clinical support in the context of early detection and management of pancreatitis.

Pelvic ring disruptions result from blunt injury mechanisms and are potentially lethal mainly due to associated injuries and massive pelvic hemorrhage. The severity of pelvic fractures in trauma victims is frequently assessed by grading the fracture according to the Tile AO/OTA classification in whole-body Computed Tomography (CT) scans. Due to the high volume of whole-body CT scans generated in trauma centers, the overall information content of a single whole-body CT scan and low manual CT reading speed, an automatic approach to Tile classification would provide substantial value, e.g., to prioritize the reading sequence of the trauma radiologists or enable them to focus on other major injuries in multi-trauma patients. In such a high-stakes scenario, an automated method for Tile grading should ideally be transparent such that the symbolic information provided by the method follows the same logic a radiologist or orthopedic surgeon would use to determine the fracture grade. This paper introduces an automated yet interpretable pelvic trauma decision support system to assist radiologists in fracture detection and Tile grading. To achieve interpretability despite processing high-dimensional whole-body CT images, we design a neurosymbolic algorithm that operates similarly to human interpretation of CT scans. The algorithm first detects relevant pelvic fractures on CTs with high specificity using Faster-RCNN. To generate robust fracture detections and associated detection (un)certainties, we perform test-time augmentation of the CT scans to apply fracture detection several times in a self-ensembling approach. The fracture detections are interpreted using a structural causal model based on clinical best practices to infer an initial Tile grade. We apply a Bayesian causal model to recover likely co-occurring fractures that may have been rejected initially due to the highly specific operating point of the detector, resulting in an updated list of detected fractures and corresponding final Tile grade. Our method is transparent in that it provides fracture location and types, as well as information on important counterfactuals that would invalidate the system's recommendation. Our approach achieves an AUC of 0.89/0.74 for translational and rotational instability,which is comparable to radiologist performance. Despite being designed for human-machine teaming, our approach does not compromise on performance compared to previous black-box methods.

Medical diagnostics comprise recognizing patterns in images, tissue slides, and symptoms. Deep learning algorithms (DLs) are well suited to such tasks, but they are black boxes in various ways. To explain DL Computer-Aided Diagnostic (CAD) results and their accuracy to patients, to manage or drive the direction of future medical DLs, to make better decisions with CAD, etc., clinical professionals may benefit from hands-on, under-the-hood lessons about medical DL. For those who already have some high-level knowledge about DL, the next step is to gain a more-fundamental understanding of DLs, which may help illuminate inside the boxes. The objectives of this Continuing Medical Education (CME) article include:Better understanding can come from relatable medical analogies and personally experiencing quick simulations to observe deep learning in action, akin to the way clinicians are trained to perform other tasks. We developed readily-implementable demonstrations and simulation exercises. We framed the exercises using analogies to breast cancer, malignancy and cancer stage as example diagnostic applications. The simulations revealed a nuanced relationship between DL output accuracy and the quantity and nature of the data. The simulation results provided lessons-learned and implications for the clinical world. Although we focused on DLs for diagnosis, they are similar to DLs for treatment (e.g. radiotherapy) so that treatment providers may also benefit from this tutorial.

Hepatocellular carcinoma (HCC) is a prevalent cancer that significantly contributes to mortality globally, primarily due to its late diagnosis. Early detection is crucial yet challenging. This study leverages the potential of deep learning (DL) technologies, employing the You Only Look Once (YOLO) architecture, to enhance the detection of HCC in computed tomography (CT) images, aiming to improve early diagnosis and thereby patient outcomes. We used a dataset of 1290 CT images from 122 patients, segmented according to a standard 70:20:10 split for training, validation, and testing phases. The YOLO-based DL model was trained on these images, with subsequent phases for validation and testing to assess the model's diagnostic capabilities comprehensively. The model exhibited exceptional diagnostic accuracy, with a precision of 0.97216, recall of 0.919, and an overall accuracy of 95.35%, significantly surpassing traditional diagnostic approaches. It achieved a specificity of 95.83% and a sensitivity of 94.74%, evidencing its effectiveness in clinical settings and its potential to reduce the rate of missed diagnoses and unnecessary interventions. The implementation of the YOLO architecture for detecting HCC in CT scans has shown substantial promise, indicating that DL models could soon become a standard tool in oncological diagnostics. As artificial intelligence technology continues to evolve, its integration into healthcare systems is expected to advance the accuracy and efficiency of diagnostics in oncology, enhancing early detection and treatment strategies and potentially improving patient survival rates.

Lumbar spinal stenosis (LSS) is a major cause of pain and disability in older individuals worldwide. Although increasing studies of traditional machine learning (TML) and deep learning (DL) were conducted in the field of diagnosing LSS and gained prominent results, the performance of these models has not been analyzed systematically.

This systematic review and meta-analysis aimed to pool the results and evaluate the heterogeneity of the current studies in using TML or DL models to diagnose LSS, thereby providing more comprehensive information for further clinical application.

This review was performed under the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines using articles extracted from PubMed, Embase databases, and Cochrane Library databases. Studies that evaluated DL or TML algorithms assessment value on diagnosing LSS were included, while those with duplicated or unavailable data were excluded. Quality Assessment of Diagnostic Accuracy Studies 2 was used to estimate the risk of bias in each study. The MIDAS module and the METAPROP module of Stata (StataCorp) were used for data synthesis and statistical analyses.

A total of 12 studies with 15,044 patients reported the assessment value of TML or DL models for diagnosing LSS. The risk of bias assessment yielded 4 studies with high risk of bias, 3 with unclear risk of bias, and 5 with completely low risk of bias. The pooled sensitivity and specificity were 0.84 (95% CI: 0.82-0.86; I2=99.06%) and 0.87 (95% CI 0.84-0.90; I2=98.7%), respectively. The diagnostic odds ratio was 36 (95% CI 26-49), the positive likelihood ratio (LR+) was 6.6 (95% CI 5.1-8.4), and the negative likelihood ratio (LR-) was 0.18 (95% CI 0.16-0.21). The summary receiver operating characteristic curves, the area under the curve of TML or DL models for diagnosing LSS of 0.92 (95% CI 0.89-0.94), indicating a high diagnostic value.

This systematic review and meta-analysis emphasize that despite the generally satisfactory diagnostic performance of artificial intelligence systems in the experimental stage for the diagnosis of LSS, none of them is reliable and practical enough to apply in real clinical practice. Further efforts, including optimization of model balance, widely accepted objective reference standards, multimodal strategy, large dataset for training and testing, external validation, and sufficient and scientific report, should be made to bridge the distance between current TML or DL models and real-life clinical applications in future studies.

PROSPERO CRD42024566535; https://tinyurl.com/msx59x8k.