Predictive Modeling of Osteoporosis: A Machine Learning Approach Based on Electromagnetic Signals

Osteoporosis is a medical condition characterized by a reduction in bone mineral density (BMD) and consequent increase in bone permittivity and is generally diagnosed using BMD measurements. However, the diagnosis relies on estimated BMD compared to healthy individuals, which can inherently exhibit a range of variability. In addition, the existing BMD measurement methods often use ionizing radiation and may be harmful to body tissues. In contrast, electromagnetic (EM) wave-based methods offer a nonionizing, noninvasive alternative. In this study, an EM wave-based system operating between 1.5 and 4.5 GHz is proposed to analyze the changes in the dielectric permittivity of bone tissues when subjected to mineral loss. The objective is to classify the progressive stages of osteoporosis using machine learning (ML) techniques. For this purpose, the transmitted signal through bone samples of varying dimensions and permittivities is simulated using a phantom model and measured experimentally using human subjects. The simulated and experimental datasets are preprocessed, the supervised classification of simulated samples is performed, and the classification model is tested on experimental data to observe its efficacy. The strategic use of EM waves and ML is effective in establishing a predictive model for understanding bone health and progression of osteoporosis.