Bone Health Classification using Machine Learning

Osteoporosis is a serious medical condition. Currently, the healthcare system employs a number of techniques to identify osteoporosis at an early stage. However, these techniques are damaging to skin and tissue, ineffective economically, and difficult to carry around. Microwaves have become more widely used in the medical field recently. It is a non-ionizing, non-invasive method that causes minor tissue damage. Previously, the Bone Health Analyser was designed using a microstrip patch antenna for frequencies ranging from 3GHz to 8GHz, and it was discovered that at lower frequencies, the transfer characteristics of the antenna for different bones are more distinct and easier to classify. Thus, in this paper, microwave signal ranging from 1.5GHz to 4.5GHz is transmitted from a Bow-Tie antenna, and a change in attenuation of the signal is received at another Bow - tie antenna. The simulated signal attenuation (S21) is recorded as it passes through various osteoporotic bones. The motive behind this research is to classify the different stages of osteoporosis using machine learning. Three supervised classification algorithms, Support Vector Machine for Classification (SVC), Random Forest (RF), and Decision Tree (DT) are applied to classify the different stages and their performance is analyzed. Features are extracted from attenuated signals and used to train the model. The results demonstrate that implementing machine learning to classify stages of osteoporosis is promising.