An Application to Scanalyse a Given X-ray Image to Defects in Bones, Muscles, and Nerves

Abstract

Bone fractures, muscle injuries, and nerve defects require timely and accurate diagnosis to prevent serious medical complications. Traditional manual interpretation of X-ray images is prone to human error and diagnostic delays, emphasizing the necessity for automated, reliable, and efficient diagnostic systems. This study aims to develop a lightweight and scalable deep learning framework that enables real-time detection of bone, muscle, and nerve defects from X-ray images. The proposed method integrates MobileNet-based deep feature extraction with Random Forest ensemble classification to balance high diagnostic accuracy with low computational overhead. The Bone Fracture Detection X-ray Dataset, comprising 1,029 labeled images, was utilized, and preprocessing steps including resizing, normalization, augmentation, and noise suppression were applied to enhance model robustness. Hyperparameters were optimized using grid search, and model evaluation was performed using 5-fold cross-validation on stratified train-validation-test splits. Experimental results demonstrate that the proposed model achieved an overall accuracy of 92.7%, with a precision of 91.8%, recall of 92.6%, and F1-score of 92.2%, outperforming baseline CNN and MobileNet-only architectures. Inference time was significantly reduced to 18 milliseconds per image, confirming its real-time applicability. Statistical significance testing further validated the superiority of the proposed model with a p-value of 0.018. This research presents a practical, deployable solution for fracture detection in clinical and remote healthcare settings, setting a foundation for future work incorporating explainable AI, attention mechanisms, and multimodal data integration to further enhance performance and trust in automated medical diagnostics