Finger Bone Fracture Classification in X-ray Images Using Convolutional Neural Network with Transfer Learning

Abstract

Finger Bone Fractures are among the most common injuries and are generally diagnosed through X-ray imaging. However, the diagnostic process is still largely performed manually by radiologists or physicians, requiring a high level of expertise and potentially leading to interpretation errors, particularly in cases involving subtle fractures. Therefore, an intelligent system is needed to assist medical personnel in detecting fractures more quickly and accurately. This study aims to develop a web-based system for classifying finger Bone Fractures from X-ray images using a Convolutional Neural Network (CNN) with a Transfer Learning approach based on the EfficientNet-B0 architecture. The datasets used in this study are obtained from FracAtlas and HBFMID, which contain finger X-ray images. The research methodology consists of dataset verification, image preprocessing, model training and testing using EfficientNet-B0, and performance evaluation based on Accuracy, precision, recall, F1-score, and the Confusion Matrix. The developed system also provides features for user management, X-ray image upload, examination history, prediction visualization using Gradient-weighted Class Activation Mapping (Grad-CAM), and diagnosis verification by physicians. The expected outcome of this study is a classification model capable of accurately detecting finger Bone Fractures and a web-based system that can assist radiographers and physicians in the diagnostic process. Furthermore, this research is expected to improve the efficiency of radiological examinations and support faster, more objective clinical decision-making.Keywords: EfficientNet-B0, Finger Fracture Detection, X-ray Image, Deep Learning, Web-Based System.