A Novel Approach for Diagnosis of Osteoporosis Integrating Clinical Decision Support with Generative Adversarial Networks

Abstract Osteoporosis is a bone illness that minimizes bone strength and increases fracture risk. Machine learning methods have been applied to diagnose osteoporosis. However, the accuracy of osteoporosis disease prediction has not improved, nor has the time taken been reduced. To improve the accuracy of osteoporosis disease prediction efficiently, the Rosenthal canonical correlative explainable deep convolutional generative adversarial network (RCCEDCGAN) is proposed. It consists of four main processes: data acquisition, preprocessing, feature selection, and classification. Data samples are collected during data acquisition. Preprocessing includes using multivariate linear regression to fill in missing values and a Q statistical proximal test to identify outliers. Feature selection is carried out using Rosenthal canonical variants analysis for identifying and selecting pertinent features with less time. Finally, explainable deep convolutional generative adversarial network (EDCGAN) is employed for classifying and predicting osteoporosis. The clinical decision support system utilizes EDCGANs for osteoporosis risk prediction analysis based on the Rand indexive decision stump model to assist healthcare professionals in diagnosis and treatment planning. The quantitatively analyzed results show that the RCCEDCGAN method improved by 5% in disease prediction accuracy, precision, recall, F1 score, and 9% specificity compared to the RR model and modified GP classifier. RCCEDCGAN method showed a p-value of less than 0.05 and a confidence interval of 95% for developing osteoporosis. In addition, the RCCEDCGAN method reduced prediction time by 8% compared to the RR model and modified GP classifier techniques. Hence, the RCCEDCGAN is an effective approach for early diagnosis and risk reduction of osteoporosis, aiding in prevention and management strategies.