An Integrative Approach of K-Means Algorithm and Convolutional Neural Networks (CNN) for Hand Fracture Classification: An Applied Study on Local Clinical Data

Abstract

Accurate diagnosis of bone fractures in radiographic images presents a central challenge in emergency departments due to the anatomical complexity of hand bones and the potential for human error resulting from fatigue. This study aims to develop an advanced hybrid system based on integrating unsupervised learning techniques with deep neural networks to enhance the efficiency of automated classification of X-ray images. The proposed methodology employed the K-Means algorithm to perform image segmentation and extract detailed structural features, followed by Convolutional Neural Networks (CNN) to classify the images into two categories (healthy, fractured).

The study was distinguished by using a real-world database collected from the radiology department archive at Yefren General Hospital in Libya, giving the research an applied character that goes beyond ideal laboratory data. The results showed a notable superiority of the proposed hybrid model, achieving high levels of accuracy and specificity compared to traditional models. This indicates the system's potential to function as a reliable clinical decision-support tool, assisting non-specialist doctors in remote areas and reducing the workload in busy medical centers. The study concludes that the integration of clustering and deep learning provides a robust diagnostic mechanism capable of handling the complexities of real-world medical images.