Investigating the Performance of Machine Learning Algorithms for Bone Fracture Identification Using Filtering Techniques

Bone fracture detection plays a crucial role in orthopaedic diagnosis and timely medical intervention. Conventional fracture identification relies on manual inspection of X-ray images by radiologists, which may lead to delayed diagnosis and human error, especially under noisy or lowcontrast imaging conditions. This paper investigates the effectiveness of various machine learning algorithms for automated bone fracture identification from X-ray images, with a comparative study conducted using filtering and non-filtering preprocessing techniques. Logistic Regression, K-Nearest Neighbour, Decision Tree, Support Vector Machine, and Random Forest classifiers are evaluated under both conditions. Additionally, a hybrid SVM–KNN model is proposed to improve classification performance. Image preprocessing techniques such as Gaussian and Median filtering are applied to enhance image quality and feature extraction. The models are implemented using Python with Scikitlearn and OpenCV libraries. Performance is evaluated using accuracy, precision, recall, and F1-score. Experimental results indicate that filtering techniques significantly enhance classification performance, with the hybrid SVM–KNN model achieving the highest accuracy of 96.7%. The findings demonstrate the potential of hybrid machine learning models for reliable and efficient bone fracture detection.