Leveraging 3D Convolutional Neural Networks for Accurate Recognition and Localization of Ankle Fractures

Abstract

Abstract Background Ankle fractures are common injuries with substantial implications for patient mobility and quality of life. Traditional imaging methods, while standard, have limitations in detecting subtle fractures and distinguishing them from complex bone structures. The advent of 3D Convolutional Neural Networks (3D-CNNs) offers a promising avenue for enhancing the accuracy and reliability of ankle fracture diagnoses. Methods In this study, we acquired 1,453 high-resolution CT scans and processed them through three distinct 3D-CNN models: 3D-Mobilenet, 3D-Resnet101, and 3D-EfficientNetB7. Our approach involved rigorous preprocessing of images, including normalization and resampling, followed by a comparative evaluation of the models using accuracy, Area Under the Curve (AUC), and recall metrics. Additionally, the integration of Gradient-weighted Class Activation Mapping (Grad-CAM) provided visual interpretability of the models' predictive focus points. Results The 3D-EfficientNetB7 model demonstrated superior performance, achieving an accuracy of 0.91 and an AUC of 0.94 after 20 training epochs. Furthermore, Grad-CAM visualizations aligned closely with expert radiologists' assessments, validating the model's diagnostic precision. Spatial localization techniques further enhanced the interpretability of fracture detection, providing clear visual cues for medical professionals. Conclusions The implementation of 3D-CNNs, particularly the 3D-EfficientNetB7 model, significantly improved the detection and localization of ankle fractures. The use of Grad-CAM has also proved essential in providing transparency to AI-driven diagnostics. Our research supports the integration of 3D-CNNs in clinical settings, potentially revolutionizing the standard of care in fracture diagnosis and paving the way for their application in broader medical imaging tasks.