The epiphyses of the hand and wrist serve as crucial indicators for assessing skeletal maturity in adolescents. This study aimed to develop a deep learning (DL) model for bone age (BA) assessment using hand and wrist X-ray images, addressing the challenge of classifying BA in adolescents. The results of this DL-based classification were then compared and analyzed with those obtained from manual assessment. A retrospective analysis was conducted on 688 hand and wrist X-ray images of adolescents aged 11.00-23.99 years from western China, which were randomly divided into training set, validation set and test set. The BA assessment results were initially analyzed and compared using four DL network models: InceptionV3, InceptionV3 + SE + Sex, InceptionV3 + Bilinear and InceptionV3 + Bilinear. + SE + Sex, to identify the DL model with the best classification performance. Subsequently, the results of the top-performing model were compared with those of manual classification. The study findings revealed that the InceptionV3 + Bilinear + SE + Sex model exhibited the best performance, achieving classification accuracies of 96.15% and 90.48% for the training and test set, respectively. Furthermore, based on the InceptionV3 + Bilinear + SE + Sex model, classification accuracies were calculated for four age groups (< 14.0 years, 14.0 years ≤ age < 16.0 years, 16.0 years ≤ age < 18.0 years, ≥ 18.0 years), with notable accuracies of 100% for the age groups 16.0 years ≤ age < 18.0 years and ≥ 18.0 years. The BA classification, utilizing the feature fusion DL network model, holds significant reference value for determining the age of criminal responsibility of adolescents, particularly at the critical legal age boundaries of 14.0, 16.0, and 18.0 years.

Deep learning has transformed computer vision but relies heavily on large labeled datasets and computational resources. Transfer learning, particularly fine-tuning pretrained models, offers a practical alternative; however, models pretrained on natural image datasets such as ImageNet may fail to capture domain-specific characteristics in medical imaging. This study introduces an unsupervised learning framework that extracts high-value dermatological features instead of relying solely on ImageNet-based pretraining. We employ a Variational Autoencoder (VAE) trained from scratch on a proprietary dermatological dataset, allowing the model to learn a structured and clinically relevant latent space. This self-supervised feature extractor is then compared to an ImageNet-pretrained backbone under identical classification conditions, highlighting the trade-offs between general-purpose and domain-specific pretraining. Our results reveal distinct learning patterns. The self-supervised model achieves a final validation loss of 0.110 (-33.33%), while the ImageNet-pretrained model stagnates at 0.100 (-16.67%), indicating overfitting. Accuracy trends confirm this: the self-supervised model improves from 45% to 65% (+44.44%) with a near-zero overfitting gap, whereas the ImageNet-pretrained model reaches 87% (+50.00%) but plateaus at 75% (+19.05%), with its overfitting gap increasing to +0.060. These findings suggest that while ImageNet pretraining accelerates convergence, it also amplifies overfitting on non-clinically relevant features. In contrast, self-supervised learning achieves steady improvements, stronger generalization, and superior adaptability, underscoring the importance of domain-specific feature extraction in medical imaging.

This research aims to enhance X-ray lung segmentation by addressing boundary distortions in anatomical structures, with the objective of refining segmentation boundaries and improving the morphological shape of segmented objects. The proposed approach combines the K-segment principal curve with Gaussian Process Regression (GPR) to refine segmentation boundaries, evaluated using lung X-ray datasets at varying resolutions. Several state-of-the-art models, including U-Net, SegNet, and TransUnet, were also assessed for comparison. The model employed a custom kernel for GPR, combining Radial Basis Function (RBF) with a cosine similarity term. The effectiveness of the model was evaluated using metrics such as the Dice Coefficient (DC) and Jaccard Index (JC) for segmentation accuracy, along with Average Symmetric Surface Distance (ASSD) and Hausdorff Distance (HD) for boundary alignment. The proposed method achieved superior segmentation performance, particularly at the highest resolution (1024x1024 pixels), with a DC of 95.7% for the left lung and 94.1% for the right lung. Among the different models, TransUnet outperformed others across both the semantic segmentation and boundary refinement stages, showing significant improvements in DC, JC, ASSD, and HD. The results indicate that the proposed boundary refinement approach effectively improves the segmentation quality of lung X-rays, excelling in refining well-defined structures and achieving superior boundary alignment, showcasing its potential for clinical applications. However, limitations exist when dealing with irregular or unpredictable shapes, suggesting areas for future enhancement.

This study proposes a deep learning vision-language model for the automated diagnosis of pediatric dental diseases, with a focus on differentiating between caries and periapical infections. The model integrates visual features extracted from panoramic radiographs using methods of non-linear dynamics and textural encoding with textual descriptions generated by a large language model. These multimodal features are concatenated and used to train a 1D-CNN classifier. Experimental results demonstrate that the proposed model outperforms conventional convolutional neural networks and standalone language-based approaches, achieving high accuracy (90%), sensitivity (92%), precision (92%), and an AUC of 0.96. This work highlights the value of combining structured visual and textual representations in improving diagnostic accuracy and interpretability in dental radiology. The approach offers a promising direction for the development of context-aware, AI-assisted diagnostic tools in pediatric dental care.

Advanced diagnostic instruments are crucial for the accurate detection and treatment of lung diseases, which affect millions of individuals globally. This study examines the effectiveness of deep learning and transfer learning models using a hybrid dataset, created by merging four individual datasets from Bangladesh and global sources. The hybrid dataset significantly enhances model accuracy and generalizability, particularly in detecting COVID-19, pneumonia, lung opacity, and normal lung conditions from chest X-ray images. A range of models, including CNN, VGG16, VGG19, InceptionV3, Xception, ResNet50V2, InceptionResNetV2, MobileNetV2, and DenseNet121, were applied to both individual and hybrid datasets. The results showed superior performance on the hybrid dataset, with VGG16, Xception, ResNet50V2, and DenseNet121 each achieving an accuracy of 99%. This consistent performance across the hybrid dataset highlights the robustness of these models in handling diverse data while maintaining high accuracy. To understand the models implicit behavior, explainable AI techniques were employed to illuminate their black-box nature. Specifically, LIME was used to enhance the interpretability of model predictions, especially in cases of misclassification, contributing to the development of reliable and interpretable AI-driven solutions for medical imaging.

With the growing capabilities of language models like ChatGPT to process text and images, this study evaluated their accuracy in detecting supernumerary teeth on periapical radiographs. A customized GPT-4V model (CGPT-4V) was also developed to assess whether domain-specific training could improve diagnostic performance compared to standard GPT-4V and GPT-4o models. One hundred eighty periapical radiographs (90 with and 90 without supernumerary teeth) were evaluated using GPT-4 V, GPT-4o, and a fine-tuned CGPT-4V model. Each image was assessed separately with the standardized prompt "Are there any supernumerary teeth in the radiograph above?" to avoid contextual bias. Three dental experts scored the responses using a three-point Likert scale for positive cases and a binary scale for negatives. Chi-square tests and ROC analysis were used to compare model performances (p < 0.05). Among the three models, CGPT-4 V exhibited the highest accuracy, detecting supernumerary teeth correctly in 91% of cases, compared to 77% for GPT-4o and 63% for GPT-4V. The CGPT-4V model also demonstrated a significantly lower false positive rate (16%) than GPT-4V (42%). A statistically significant difference was found between CGPT-4V and GPT-4o (p < 0.001), while no significant difference was observed between GPT-4V and CGPT-4V or between GPT-4V and GPT-4o. Additionally, CGPT-4V successfully identified multiple supernumerary teeth in radiographs where present. These findings highlight the diagnostic potential of customized GPT models in dental radiology. Future research should focus on multicenter validation, seamless clinical integration, and cost-effectiveness to support real-world implementation.

Foundation models leveraging vision-language pretraining have shown promise in chest X-ray (CXR) interpretation, yet their real-world performance across diverse populations and diagnostic tasks remains insufficiently evaluated. This study benchmarks the diagnostic performance and generalizability of foundation models versus traditional convolutional neural networks (CNNs) on multinational CXR datasets. We evaluated eight CXR diagnostic models - five vision-language foundation models and three CNN-based architectures - across 37 standardized classification tasks using six public datasets from the USA, Spain, India, and Vietnam, and three private datasets from hospitals in China. Performance was assessed using AUROC, AUPRC, and other metrics across both shared and dataset-specific tasks. Foundation models outperformed CNNs in both accuracy and task coverage. MAVL, a model incorporating knowledge-enhanced prompts and structured supervision, achieved the highest performance on public (mean AUROC: 0.82; AUPRC: 0.32) and private (mean AUROC: 0.95; AUPRC: 0.89) datasets, ranking first in 14 of 37 public and 3 of 4 private tasks. All models showed reduced performance on pediatric cases, with average AUROC dropping from 0.88 +/- 0.18 in adults to 0.57 +/- 0.29 in children (p = 0.0202). These findings highlight the value of structured supervision and prompt design in radiologic AI and suggest future directions including geographic expansion and ensemble modeling for clinical deployment. Code for all evaluated models is available at https://drive.google.com/drive/folders/1B99yMQm7bB4h1sVMIBja0RfUu8gLktCE

Visual grounding (VG) is the capability to identify the specific regions in an image associated with a particular text description. In medical imaging, VG enhances interpretability by highlighting relevant pathological features corresponding to textual descriptions, improving model transparency and trustworthiness for wider adoption of deep learning models in clinical practice. Current models struggle to associate textual descriptions with disease regions due to inefficient attention mechanisms and a lack of fine-grained token representations. In this paper, we empirically demonstrate two key observations. First, current VLMs assign high norms to background tokens, diverting the model's attention from regions of disease. Second, the global tokens used for cross-modal learning are not representative of local disease tokens. This hampers identifying correlations between the text and disease tokens. To address this, we introduce simple, yet effective Disease-Aware Prompting (DAP) process, which uses the explainability map of a VLM to identify the appropriate image features. This simple strategy amplifies disease-relevant regions while suppressing background interference. Without any additional pixel-level annotations, DAP improves visual grounding accuracy by 20.74% compared to state-of-the-art methods across three major chest X-ray datasets.

Self-supervised learning has emerged as a powerful paradigm for training deep neural networks, particularly in medical imaging where labeled data is scarce. While current approaches typically rely on synthetic augmentations of single images, we propose VET-DINO, a framework that leverages a unique characteristic of medical imaging: the availability of multiple standardized views from the same study. Using a series of clinical veterinary radiographs from the same patient study, we enable models to learn view-invariant anatomical structures and develop an implied 3D understanding from 2D projections. We demonstrate our approach on a dataset of 5 million veterinary radiographs from 668,000 canine studies. Through extensive experimentation, including view synthesis and downstream task performance, we show that learning from real multi-view pairs leads to superior anatomical understanding compared to purely synthetic augmentations. VET-DINO achieves state-of-the-art performance on various veterinary imaging tasks. Our work establishes a new paradigm for self-supervised learning in medical imaging that leverages domain-specific properties rather than merely adapting natural image techniques.

Tuberous breast deformity (TBD) is a congenital condition characterized by constriction of the breast base, parenchymal hypoplasia, and areolar herniation. The absence of a universally accepted classification system complicates diagnosis and surgical planning, leading to variability in clinical outcomes. Artificial intelligence (AI) has emerged as a powerful adjunct in medical imaging, enabling objective, reproducible, and data-driven diagnostic assessments. This study introduces an AI-driven diagnostic tool for tuberous breast deformity (TBD) classification using a Siamese Network trained on paired frontal and lateral images. Additionally, the model generates a continuous Tuberosity Score (ranging from 0 to 1) based on embedding vector distances, offering an objective measure to enhance surgical planning and improved clinical outcomes. A dataset of 200 expertly classified frontal and lateral breast images (100 tuberous, 100 non-tuberous) was used to train a Siamese Network with contrastive loss. The model extracted high-dimensional feature embeddings to differentiate tuberous from non-tuberous breasts. Five-fold cross-validation ensured robust performance evaluation. Performance metrics included accuracy, precision, recall, and F1-score. Visualization techniques, such as t-SNE clustering and occlusion sensitivity mapping, were employed to interpret model decisions. The model achieved an average accuracy of 96.2% ± 5.5%, with balanced precision and recall. The Tuberosity Score, derived from the Euclidean distance between embeddings, provided a continuous measure of deformity severity, correlating well with clinical assessments. This AI-based framework offers an objective, high-accuracy classification system for TBD. The Tuberosity Score enhances diagnostic precision, potentially aiding in surgical planning and improving patient outcomes.