Impact of Image Enhancement Techniques on CNN Performance for Bone Fracture Detection Using X-ray Images

Abstract

The study offers a thorough analysis of the influence of image-enhancing methods on the classification of bone fractures in X-ray images. The FracAtlas labeled dataset is utilized to train and assess a convolutional neural network (CNN) classifier. Precision, accuracy, F1-Score, and recall are used as performance evaluation metrics for the CNN classifier. The outcomes revealed that methods such as contrast-limited adaptive histogram equalizer, wavelet transformation, and Gaussian filters significantly enhance classification accuracy, increasing it from 80.20% to 92.52%. The study proves that the use of optimization methods along with the CNN classifier improves fracture classification and considers an additional diagnostic tool for radiologists. The research's findings advance the image interpretation of medicine realm, showing that the use of a deep learning model combined with image enhancement approaches is clinically diagnostic important.