An Optimized Approach for Detecting Osteoporosis Using Advanced Machine Learning and Genetic Algorithms
Year: 2025
Type: article
Abstract: Artificial intelligence has transformed healthcare through improved disease prediction and diagnostic accuracy. However, the prediction of osteoporosis remains challenging due to limitations in current diagnostic methods, which are based on DXA scans and basic risk assessments that lack personalization and comprehensive integration of risk factors. This study employs machine learning techniques to predict osteoporosis using data from two sources—a Kaggle dataset and the NHANES dataset—each containing around 2,000 patients. The datasets were considered and formatted to include variables such as age, gender, weight, family history, lifestyle habits, and medical history, allowing for a comparative analysis. We used eight traditional machine learning methods and two deep learning models, CNN and RNN, along with a BERT transformer model. To enhance performance, we integrated Genetic Algorithms with the RNN model for parameter optimization. Our models were validated using a 10-fold cross-validation. The BERT model achieved the best performance, with 92% accuracy, 88% sensitivity, 96% specificity, and an AUC of 0.92 on the Kaggle dataset. It also demonstrated strong results on the NHANES dataset, which was originally imbalanced and was balanced using SMOTE. The RNN model initially performed poorly, but improved significantly when optimized with Genetic Algorithms, achieving an AUC of 0.932 and matching the overall performance of BERT. Our findings demonstrate that combining deep learning with evolutionary optimization techniques can significantly improve osteoporosis prediction accuracy, offering potential for early diagnosis and improved preventive care in bone health management.
Source: IEEE Access
Authors: Sajida Chamass, Layal Abu Daher, May Itani
Institution: Beirut Arab University
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W4417201591