Osteoporosis Prognosis Through Machine Learning Analysis of X-Ray Films

This study aims to assess the efficacy of machine learning models, specifically k-nearest neighbor (KNN) and convolutional neural network (CNN), for predicting osteoporosis prognosis using X-ray films. A dataset comprising 2,445 X-ray images, including 1,175 healthy bone images and 1,270 osteoporosis bone images, was utilized. The dataset was divided into a learning subset of 2,417 images and a testing subset of 24 images, with equal representation of osteoporosis and normal bone images. Performance evaluation revealed an accuracy of 78. 57%for the KNN model $(K=11)$ and 97.57%for the CNN model (after 40 epochs). These results suggest that the CNN model outperforms the KNN model in accurately detecting osteoporosis prognosis. The findings highlight the potential of machine learning in improving osteoporosis assessment and management. Further research is warranted to validate these models in clinical settings and explore their broader clinical implications.