Fractures in children are common in emergency care, and accurate diagnosis is crucial to avoid complications affecting skeletal development. Limited access to pediatric radiology specialists emphasizes the potential of artificial intelligence (AI)-based diagnostic tools. This study evaluates the performance of the AI software BoneView® for detecting fractures of the upper extremity in children aged 2-18 years. A retrospective analysis was conducted using radiographic data from 826 pediatric patients presenting to the university's pediatric emergency department. Independent assessments by two experienced pediatric radiologists served as reference standard. The diagnostic accuracy of the AI tool compared to the reference standard was evaluated and performance parameters, e.g., sensitivity, specificity, positive and negative predictive values were calculated. The AI tool achieved an overall sensitivity of 89% and specificity of 91% for detecting fractures of the upper extremities. Significantly poorer performance compared to the reference standard was observed for the shoulder, elbow, hand, and fingers, while no significant difference was found for the wrist, clavicle, upper arm, and forearm. The software performed best for wrist fractures (sensitivity: 96%; specificity: 94%) and worst for elbow fractures (sensitivity: 87%; specificity: 65%). The software assessed provides diagnostic support in pediatric emergency radiology. While its overall performance is robust, limitations in specific anatomical regions underscore the need for further training of the underlying algorithms. The results suggest that AI can complement clinical expertise but should not replace radiological assessment. Question There is no comprehensive analysis of an AI-based tool for the diagnosis of pediatric fractures focusing on the upper extremities. Findings The AI-based software demonstrated solid overall diagnostic accuracy in the detection of upper limb fractures in children, with performance differing by anatomical region. Clinical relevance AI-based fracture detection can support pediatric emergency radiology, especially where expert interpretation is limited. However, further algorithm training is needed for certain anatomical regions and for detecting associated findings such as joint effusions to maximize clinical benefit.

We evaluated the diagnostic performance of two AI software programs (BoneView and RBfracture) in assisting non-specialist radiologists (NSRs) in detecting scaphoid fractures using conventional wrist radiographs (X-rays). We retrospectively analyzed 724 radiographs from 264 patients with wrist trauma. Patients were classified into two groups: Group 1 included cases with a definitive diagnosis by a specialist radiologist (SR) based on X-rays (either scaphoid fracture or not), while Group 2 comprised indeterminate cases for the SRs requiring a CT scan for a final diagnosis. Indeterminate cases were defined as negative or doubtful X-rays in patients with persistent clinical symptoms. The X-rays were evaluated by AI and two NSRs, independently and in combination. We compared their diagnostic performances using sensitivity, specificity, area under the curve (AUC), and Cohen's kappa for diagnostic agreement. Group 1 included 174 patients, with 80 cases (45.97%) of scaphoid fractures. Group 2 had 90 patients, of which 44 with uncertain diagnoses and 46 negative cases with persistent symptoms. Scaphoid fractures were identified in 51 patients (56.67%) in Group 2 after further CT imaging. In Group 1, AI performed similarly to NSRs (AUC: BoneView 0.83, RBfracture 0.84, NSR1 0.88, NSR2 0.90), without significant contribution of AI to the performance of NSRs. In Group 2, performances were lower (AUC: BoneView 0.62, RBfracture 0.65, NSR1 0.46, NSR2 0.63), but AI assistance significantly improved NSR performance (NSR2 + BoneView AUC = 0.75, p = 0.003; NSR2 + RBfracture AUC = 0.72, p = 0.030). Diagnostic agreement between NSR1 with AI support and SR was moderate (kappa = 0.576), and substantial for NSR2 (kappa = 0.712). AI tools may effectively assist NSRs, especially in complex scaphoid fracture cases.

Accurate age estimation is essential for assessing pediatric developmental stages and for forensics. Conventionally, pediatric age is clinically estimated by bone age through wrist X-rays. However, recent advances in deep learning enable other radiological modalities to serve as a promising complement. This study aims to explore the effectiveness of deep learning for pediatric age estimation using chest X-rays. We developed a ResNet-based deep neural network model enhanced with Coordinate Attention mechanism to predict pediatric age from chest X-rays. A dataset comprising 128,008 images was retrospectively collected from two large tertiary hospitals in Shanghai. Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) were employed as main evaluation metrics across age groups. Further analysis was conducted using Spearman correlation and heatmap visualizations. The model achieved an MAE of 5.86 months for males and 5.80 months for females on the internal validation set. On the external test set, the MAE was 7.40 months for males and 7.29 months for females. The Spearman correlation coefficient was above 0.98, indicating a strong positive correlation between the predicted and true age. Heatmap analysis revealed the deep learning model mainly focused on the spine, mediastinum, heart and great vessels, with additional attention given to surrounding bones. We successfully constructed a large dataset of pediatric chest X-rays and developed a neural network model integrated with Coordinate Attention for age prediction. Experiments demonstrated the model's robustness and proved that chest X-rays can be effectively utilized for accurate pediatric age estimation. By integrating pediatric chest X-rays with age data using deep learning, we can provide more support for predicting children's age, thereby aiding in the screening of abnormal growth and development in children. This study explores whether deep learning could leverage chest X-rays for pediatric age prediction. Trained on over 120,000 images, the model shows high accuracy on internal and external validation sets. This method provides a potential complement for traditional bone age assessment and could reduce radiation exposure.

Adolescent idiopathic scoliosis (AIS) affects a significant portion of the adolescent population, leading to severe spinal deformities if untreated. Diagnosis, surgical planning, and assessment of outcomes are determined primarily by the Cobb angle on anteroposterior spinal radiographs. Screening for scoliosis enables early interventions and improved outcomes. However, screenings are often conducted through school entities where a trained radiologist may not be available to accurately interpret the imaging results. In this study, we developed an artificial intelligence tool utilizing a keypoint region-based convolutional neural network (KR-CNN) for automated thoracic Cobb angle (TCA) measurement. The KR-CNN was trained on 609 whole-spine radiographs of AIS patients and validated using our institutional AIS registry, which included 83 patients who underwent posterior spinal fusion with both preoperative and postoperative anteroposterior X-ray images. The KR-CNN model demonstrated superior performance metrics, including a mean absolute error (MAE) of 2.22, mean squared error (MSE) of 9.1, symmetric mean absolute percentage error (SMAPE) of 4.29, and intraclass correlation coefficient (ICC) of 0.98, outperforming existing methods. This method will enable fast and accurate screening for AIS and assessment of postoperative outcomes and provides a development framework for further automation and validation of spinopelvic measurements.

The Philippines' high tuberculosis (TB) burden calls for effective point-of-care screening. Systematic TB case finding using chest X-ray (CXR) with computer-aided detection powered by deep learning-based artificial intelligence (AI-CAD) provided this opportunity. We aimed to comprehensively review AI-CAD's real-life performance in the local context to support refining its integration into the country's programmatic TB elimination efforts. Retrospective cross-sectional data analysis was done on case-finding activities conducted in four regions of the Philippines between May 2021 and March 2024. Individuals 15 years and older with complete CXR and molecular World Health Organization-recommended rapid diagnostic (mWRD) test results were included. TB presumptive was detected either by CXR or TB signs and symptoms and/or official radiologist readings. The overall diagnostic accuracy of CXR with AI-CAD, stratified by different factors, was assessed using a fixed abnormality threshold and mWRD as the standard reference. Given the imbalanced dataset, we evaluated both precision-recall (PRC) and receiver operating characteristic (ROC) plots. Due to limited verification of CAD-negative individuals, we used "pseudo-sensitivity" and "pseudo-specificity" to reflect estimates based on partial testing. We identified potential factors that may affect performance metrics. Using a 0.5 abnormality threshold in analyzing 5740 individuals, the AI-CAD model showed high pseudo-sensitivity at 95.6% (95% CI, 95.1-96.1) but low pseudo-specificity at 28.1% (26.9-29.2) and positive predictive value (PPV) at 18.4% (16.4-20.4). The area under the operating characteristic curve was 0.820, whereas the area under the precision-recall curve was 0.489. Pseudo-sensitivity was higher among males, younger individuals, and newly diagnosed TB. Threshold analysis revealed trade-offs, as increasing the threshold score to 0.68 saved more mWRD tests (42%) but led to an increase in missed cases (10%). Threshold adjustments affected PPV, tests saved, and case detection differently across settings. Scaling up AI-CAD use in TB screening to improve TB elimination efforts could be beneficial. There is a need to calibrate threshold scores based on resource availability, prevalence, and program goals. ROC and PRC plots, which specify PPV, could serve as valuable metrics for capturing the best estimate of model performance and cost-benefit ratios within the context-specific implementation of resource-limited settings.

The COVID-19 pandemic emphasised necessity for prompt, precise diagnostics, secure data storage, and robust privacy protection in healthcare. Existing diagnostic systems often suffer from limited transparency, inadequate performance, and challenges in ensuring data security and privacy. The research proposes a novel privacy-preserving diagnostic framework, Heterogeneous Convolutional-recurrent attention Transfer learning based ResNeXt with Modified Greater Cane Rat optimisation (HCTR-MGR), that integrates deep learning, Explainable Artificial Intelligence (XAI), and blockchain technology. The HCTR model combines convolutional layers for spatial feature extraction, recurrent layers for capturing spatial dependencies, and attention mechanisms to highlight diagnostically significant regions. A ResNeXt-based transfer learning backbone enhances performance, while the MGR algorithm improves robustness and convergence. A trust-based permissioned blockchain stores encrypted patient metadata to ensure data security and integrity and eliminates centralised vulnerabilities. The framework also incorporates SHAP and LIME for interpretable predictions. Experimental evaluation on two benchmark chest X-ray datasets demonstrates superior diagnostic performance, achieving 98-99% accuracy, 97-98% precision, 95-97% recall, 99% specificity, and 95-98% F1-score, offering a 2-6% improvement over conventional models such as ResNet, SARS-Net, and PneuNet. These results underscore the framework's potential for scalable, secure, and clinically trustworthy deployment in real-world healthcare systems.

Achieving robust performance and fairness across diverse patient populations remains a challenge in developing clinically deployable deep learning models for diagnostic imaging. Synthetic data generation has emerged as a promising strategy to address limitations in dataset scale and diversity. We introduce RoentGen-v2, a text-to-image diffusion model for chest radiographs that enables fine-grained control over both radiographic findings and patient demographic attributes, including sex, age, and race/ethnicity. RoentGen-v2 is the first model to generate clinically plausible images with demographic conditioning, facilitating the creation of a large, demographically balanced synthetic dataset comprising over 565,000 images. We use this large synthetic dataset to evaluate optimal training pipelines for downstream disease classification models. In contrast to prior work that combines real and synthetic data naively, we propose an improved training strategy that leverages synthetic data for supervised pretraining, followed by fine-tuning on real data. Through extensive evaluation on over 137,000 chest radiographs from five institutions, we demonstrate that synthetic pretraining consistently improves model performance, generalization to out-of-distribution settings, and fairness across demographic subgroups. Across datasets, synthetic pretraining led to a 6.5% accuracy increase in the performance of downstream classification models, compared to a modest 2.7% increase when naively combining real and synthetic data. We observe this performance improvement simultaneously with the reduction of the underdiagnosis fairness gap by 19.3%. These results highlight the potential of synthetic imaging to advance equitable and generalizable medical deep learning under real-world data constraints. We open source our code, trained models, and synthetic dataset at https://github.com/StanfordMIMI/RoentGen-v2 .

Supervised learning techniques have proven their efficacy in many applications with abundant data. However, applying these methods to medical imaging is challenging due to the scarcity of data, given the high acquisition costs and intricate data characteristics of those images, thereby limiting the full potential of deep neural networks. To address the lack of data, augmentation techniques leverage geometry, color, and the synthesis ability of generative models (GMs). Despite previous efforts, gaps in the generation process limit the impact of data augmentation to improve understanding of medical images, e.g., the highly structured nature of some domains, such as X-ray images, is ignored. Current GMs rely solely on the network's capacity to blindly synthesize augmentations that preserve semantic relationships of chest X-ray images, such as anatomical restrictions, representative structures, or structural similarities consistent across datasets. In this paper, we introduce a novel GM that leverages the structural resemblance of medical images by learning a latent graph representation (LGR). We design an end-to-end model to learn (i) a LGR that captures the intrinsic structure of X-ray images and (ii) a graph convolutional network (GCN) that reconstructs the X-ray image from the LGR. We employ adversarial training to guide the generator and discriminator models in learning the distribution of the learned LGR. Using the learned GCN, our approach generates structure-preserving synthetic images by mapping generated LGRs to X-ray. Additionally, we evaluate the learned graph representation for other tasks, such as X-ray image classification and segmentation. Numerical experiments demonstrate the efficacy of our approach, increasing performance up to $3\%$ and $2\%$ for classification and segmentation, respectively.

In recent years, spinal X-ray image segmentation has played a vital role in the computer-aided diagnosis of various adolescent spinal disorders. However, due to the complex morphology of lesions and the fact that most existing methods are tailored to single-disease scenarios, current segmentation networks struggle to balance local detail preservation and global structural understanding across different disease types. As a result, they often suffer from limited accuracy, insufficient robustness, and poor adaptability. To address these challenges, we propose a novel fully automated spinal segmentation network, DCE-UNet, which integrates the local modeling strength of convolutional neural networks (CNNs) with the global contextual awareness of Transformers. The network introduces several architectural and feature fusion innovations. Specifically, a lightweight Transformer module is incorporated in the encoder to model high-level semantic features and enhance global contextual understanding. In the decoder, a Rec-Block module combining residual convolution and channel attention is designed to improve feature reconstruction and multi-scale fusion during the upsampling process. Additionally, the downsampling feature extraction path integrates a novel DC-Block that fuses channel and spatial attention mechanisms, enhancing the network's ability to represent complex lesion structures. Experiments conducted on a self-constructed large-scale multi-disease adolescent spinal X-ray dataset demonstrate that DCE-UNet achieves a Dice score of 91.3%, a mean Intersection over Union (mIoU) of 84.1, and a Hausdorff Distance (HD) of 4.007, outperforming several state-of-the-art comparison networks. Validation on real segmentation tasks further confirms that DCE-UNet delivers consistently superior performance across various lesion regions, highlighting its strong adaptability to multiple pathologies and promising potential for clinical application.

This study comprehensively explores knowledge distillation frameworks for COVID-19 and lung cancer classification using chest X-ray (CXR) images. We employ high-capacity teacher models, including VGG19 and lightweight Vision Transformers (Visformer-S and AutoFormer-V2-T), to guide the training of a compact, hardware-aware student model derived from the OFA-595 supernet. Our approach leverages hybrid supervision, combining ground-truth labels with teacher models' soft targets to balance accuracy and computational efficiency. We validate our models on two benchmark datasets: COVID-QU-Ex and LCS25000, covering multiple classes, including COVID-19, healthy, non-COVID pneumonia, lung, and colon cancer. To interpret the spatial focus of the models, we employ Score-CAM-based visualizations, which provide insight into the reasoning process of both teacher and student networks. The results demonstrate that the distilled student model maintains high classification performance with significantly reduced parameters and inference time, making it an optimal choice in resource-constrained clinical environments. Our work underscores the importance of combining model efficiency with explainability for practical, trustworthy medical AI solutions.