The development of the Artificial Intelligence (AI)-based severity inspection model for ankylosing spondylitis (AS) could support health professionals to rapidly assess the severity of the disease, enhance proficiency, and reduce the demands of human resources. This paper aims to develop an AI-based severity inspection model for AS using patients' X-ray images and modified Stoke Ankylosing Spondylitis Spinal Score (mSASSS). The numerical simulation with AI is developed following the progress of data preprocessing, building and testing the model, and then the model. The training data is preprocessed by inviting three experts to check the X-ray images of 222 patients following the Gold Standard. The model is then developed through two stages, including keypoint detection and mSASSS evaluation. The two-stage AI-based severity inspection model for AS was developed to automatically detect spine points and evaluate mSASSS scores. At last, the data obtained from the developed model was compared with those from experts' assessment to analyse the accuracy of the model. The study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. The spine point detection at the first stage achieved 1.57 micrometres in mean error distance with the ground truth, and the second stage of the classification network can reach 0.81 in mean accuracy. The model can correctly identify 97.4% patches belonging to mSASSS score 3, while those belonging to score 0 can still be classified into scores 1 or 2. The automatic severity inspection model for AS developed in this paper is accurate and can support health professionals in rapidly assessing the severity of AS, enhancing assessment proficiency, and reducing the demands of human resources.

Retrospective study. The study aimed to develop and authenticated a deep learning model based on X-ray images to accurately diagnose fresh thoracolumbar vertebral compression fractures. In clinical practice, diagnosing fresh vertebral compression fractures often requires MRI. However, due to the scarcity of MRI resources and the high time and economic costs involved, some patients may not receive timely diagnosis and treatment. Using a deep learning model combined with X-rays for diagnostic assistance could potentially serve as an alternative to MRI. In this study, the main collection included X-ray images suspected of thoracolumbar vertebral compression fractures from the municipal shared database between December 2012 and February 2024. Deep learning models were constructed using frameworks of EfficientNet, MobileNet, and MnasNet, respectively. We conducted a preliminary evaluation of the deep learning model using the validation set. The diagnostic performance of the models was evaluated using metrics such as AUC value, accuracy, sensitivity, specificity, F1 score, precision, and ROC curve. Finally, the deep learning models were compared with evaluations from two spine surgeons of different experience levels on the control set. This study included a total of 3025 lateral X-ray images from 2224 patients. The data set was divided into a training set of 2388 cases, a validation set of 482 cases, and a control set of 155 cases. In the validation set, the three groups of DL models had accuracies of 83.0%, 82.4%, and 82.2%, respectively. The AUC values were 0.861, 0.852, and 0.865, respectively. In the control set, the accuracies of the three groups of DL models were 78.1%, 78.1%, and 80.7%, respectively, all higher than spinal surgeons and significantly higher than junior spine surgeon. This study developed deep learning models for detecting fresh vertebral compression fractures, demonstrating high accuracy.

Deep learning models often struggle to maintain generalizability in medical imaging, particularly under domain-fracture scenarios where distribution shifts arise from varying imaging techniques, acquisition protocols, patient populations, demographics, and equipment. In practice, each hospital may need to train distinct models - differing in learning task, width, and depth - to match local data. For example, one hospital may use Euclidean architectures such as MLPs and CNNs for tabular or grid-like image data, while another may require non-Euclidean architectures such as graph neural networks (GNNs) for irregular data like brain connectomes. How to train such heterogeneous models coherently across datasets, while enhancing each model's generalizability, remains an open problem. We propose unified learning, a new paradigm that encodes each model into a graph representation, enabling unification in a shared graph learning space. A GNN then guides optimization of these unified models. By decoupling parameters of individual models and controlling them through a unified GNN (uGNN), our method supports parameter sharing and knowledge transfer across varying architectures (MLPs, CNNs, GNNs) and distributions, improving generalizability. Evaluations on MorphoMNIST and two MedMNIST benchmarks - PneumoniaMNIST and BreastMNIST - show that unified learning boosts performance when models are trained on unique distributions and tested on mixed ones, demonstrating strong robustness to unseen data with large distribution shifts. Code and benchmarks: https://github.com/basiralab/uGNN

Dark-field radiography is a novel X-ray imaging modality that enables complementary diagnostic information by visualizing the microstructural properties of lung tissue. Implemented via a Talbot-Lau interferometer integrated into a conventional X-ray system, it allows simultaneous acquisition of perfectly temporally and spatially registered attenuation-based conventional and dark-field radiographs. Recent clinical studies have demonstrated that dark-field radiography outperforms conventional radiography in diagnosing and staging pulmonary diseases. However, the polychromatic nature of medical X-ray sources leads to beam-hardening, which introduces structured artifacts in the dark-field radiographs, particularly from osseous structures. This so-called beam-hardening-induced dark-field signal is an artificial dark-field signal and causes undesired cross-talk between attenuation and dark-field channels. This work presents a segmentation-based beam-hardening correction method using deep learning to segment ribs and clavicles. Attenuation contribution masks derived from dual-layer detector computed tomography data, decomposed into aluminum and water, were used to refine the material distribution estimation. The method was evaluated both qualitatively and quantitatively on clinical data from healthy subjects and patients with chronic obstructive pulmonary disease and COVID-19. The proposed approach reduces bone-induced artifacts and improves the homogeneity of the lung dark-field signal, supporting more reliable visual and quantitative assessment in clinical dark-field chest radiography.

Hand and foot digital radiography (DR) is an indispensable tool in medical imaging, with varying diagnostic requirements necessitating different hand and foot positionings. Accurate positioning is crucial for obtaining diagnostically valuable images. Furthermore, adjusting exposure parameters such as exposure area based on patient conditions helps minimize the likelihood of image retakes. We propose a personalized positioning and exposure assistant capable of automatically recognizing hand and foot positionings and recommending appropriate exposure parameters to achieve these objectives. The assistant comprises three modules: (1) Progressive Iterative Hand-Foot Tracker (PIHFT) to iteratively locate hands or feet in RGB images, providing the foundation for accurate pose estimation; (2) Multi-Task Shared Pose Estimation Transformer (MTSPET), a Transformer-based model that encompasses hand and foot estimation branches with similar network architectures, sharing a common backbone. MTSPET outperformed MediaPipe in the hand pose estimation task and successfully transferred this capability to the foot pose estimation task; (3) Domain Expertise-embedded Positioning and Exposure Assistant (DEPEA), which combines the key-point coordinates of hands and feet with specific positioning and exposure parameter requirements, capable of checking patient positioning and inferring exposure areas and Regions of Interest (ROIs) of Digital Automatic Exposure Control (DAEC). Additionally, two datasets were collected and used to train MTSPET. A preliminary clinical trial showed strong agreement between PPEA's outputs and manual annotations, indicating the system's effectiveness in typical clinical scenarios. The contributions of this study lay the foundation for personalized, patient-specific imaging strategies, ultimately enhancing diagnostic outcomes and minimizing the risk of errors in clinical settings.

The rapid growth of multimodal medical imaging data presents significant storage and transmission challenges, particularly in resource-constrained clinical settings. We propose NEURAL, a novel framework that addresses this by using semantics-guided data compression. Our approach repurposes cross-attention scores between the image and its radiological report from a fine-tuned generative vision-language model to structurally prune chest X-rays, preserving only diagnostically critical regions. This process transforms the image into a highly compressed, graph representation. This unified graph-based representation fuses the pruned visual graph with a knowledge graph derived from the clinical report, creating a universal data structure that simplifies downstream modeling. Validated on the MIMIC-CXR and CheXpert Plus dataset for pneumonia detection, NEURAL achieves a 93.4-97.7\% reduction in image data size while maintaining a high diagnostic performance of 0.88-0.95 AUC, outperforming other baseline models that use uncompressed data. By creating a persistent, task-agnostic data asset, NEURAL resolves the trade-off between data size and clinical utility, enabling efficient workflows and teleradiology without sacrificing performance. Our NEURAL code is available at https://github.com/basiralab/NEURAL.

Total-body dual X-ray absorptiometry (TBDXA) imaging is a relatively low-cost whole-body imaging modality, widely used for body composition assessment. We develop and validate a deep learning method for automatic fiducial point placement on TBDXA scans using 1,683 manually-annotated TBDXA scans. The method achieves 99.5% percentage correct keypoints in an external testing dataset. To demonstrate the value for shape and appearance modeling (SAM), our method is used to place keypoints on 35,928 scans for five different TBDXA imaging modes, then associations with health markers are tested in two cohorts not used for SAM model generation using two-sample Kolmogorov-Smirnov tests. SAM feature distributions associated with health biomarkers are shown to corroborate existing evidence and generate new hypotheses on body composition and shape's relationship to various frailty, metabolic, inflammation, and cardiometabolic health markers. Evaluation scripts, model weights, automatic point file generation code, and triangulation files are available at https://github.com/hawaii-ai/dxa-pointplacement.

Explainable Artificial Intelligence (XAI) encompasses a broad spectrum of methods that aim to enhance the transparency of deep learning models, with Class Activation Mapping (CAM) methods widely used for visual interpretability. However, systematic evaluations of these methods in veterinary radiography remain scarce. This study presents a comparative analysis of eleven CAM methods, including GradCAM, XGradCAM, ScoreCAM, and EigenCAM, on a dataset of 7362 canine and feline X-ray images. A ResNet18 model was chosen based on the specificity of the dataset and preliminary results where it outperformed other models. Quantitative and qualitative evaluations were performed to determine how well each CAM method produced interpretable heatmaps relevant to clinical decision-making. Among the techniques evaluated, EigenGradCAM achieved the highest mean score and standard deviation (SD) of 2.571 (SD = 1.256), closely followed by EigenCAM at 2.519 (SD = 1.228) and GradCAM++ at 2.512 (SD = 1.277), with methods such as FullGrad and XGradCAM achieving worst scores of 2.000 (SD = 1.300) and 1.858 (SD = 1.198) respectively. Despite variations in saliency visualization, no single method universally improved veterinarians' diagnostic confidence. While certain CAM methods provide better visual cues for some pathologies, they generally offered limited explainability and didn't substantially improve veterinarians' diagnostic confidence.

In this work, we present GazeLT, a human visual attention integration-disintegration approach for long-tailed disease classification. A radiologist's eye gaze has distinct patterns that capture both fine-grained and coarser level disease related information. While interpreting an image, a radiologist's attention varies throughout the duration; it is critical to incorporate this into a deep learning framework to improve automated image interpretation. Another important aspect of visual attention is that apart from looking at major/obvious disease patterns, experts also look at minor/incidental findings (few of these constituting long-tailed classes) during the course of image interpretation. GazeLT harnesses the temporal aspect of the visual search process, via an integration and disintegration mechanism, to improve long-tailed disease classification. We show the efficacy of GazeLT on two publicly available datasets for long-tailed disease classification, namely the NIH-CXR-LT (n=89237) and the MIMIC-CXR-LT (n=111898) datasets. GazeLT outperforms the best long-tailed loss by 4.1% and the visual attention-based baseline by 21.7% in average accuracy metrics for these datasets. Our code is available at https://github.com/lordmoinak1/gazelt.

The aim of this study was to evaluate GPT-4o's multimodal reasoning ability to review panoramic radiograph (PR) and verify its radiologic findings, while exploring the role of prompt engineering in enhancing its performance. The study included 230 PRs from West China Hospital of Stomatology in 2024, which were interpreted to generate the PR findings. A total of 300 instances of interpretation errors, were manually inserted into the PR findings. The ablation study was conducted to assess whether GPT-4o can perform reasoning on PR under a zero-shot prompt. Prompt engineering was employed to enhance the reasoning capabilities of GPT-4o in identifying interpretation errors with PRs. The prompt strategies included chain-of-thought, self-consistency, in-context learning, multimodal in-context learning, and their systematic integration into a meta-prompt. Recall, accuracy, and F1 score were employed to evaluate the outputs. Subsequently, the localization capability of GPT-4o and its influence on reasoning capability were evaluated. In the ablation study, GPT-4o's recall increased significantly from 2.67 to 43.33% upon acquiring PRs (P < 0.001). GPT-4o with the meta prompt demonstrated improvements in recall (43.33% vs. 52.67%, P = 0.022), accuracy (39.95% vs. 68.75%, P < 0.001), and F1 score (0.42 vs. 0.60, P < 0.001) compared to the zero-shot prompt and other prompt strategies. The localization accuracy of GPT-4o was 45.67% (137 out of 300, 95% CI: 40.00 to 51.34). A significant correlation was observed between its localization accuracy and reasoning capability under the meta prompt (φ coefficient = 0.33, p < 0.001). The model's recall increased by 5.49% (P = 0.031) by providing accurate localization cues within the meta prompt. GPT-4o demonstrated a certain degree of multimodal capability for PR, with performance enhancement through prompt engineering. Nevertheless, its performance remains inadequate for clinical requirements. Future efforts will be necessary to identify additional factors influencing the model's reasoning capability or to develop more advanced models. Evaluating GPT-4o's capability to interpret and reason through PRs and exploring potential methods to enhance its performance before clinical application in assisting radiological assessments.