Application of Deep Learning in Bone Digital Densitometer Based on Electrical Impedance Spectroscopy

Authors

  • Sari Ayu Wulandari udinus

Abstract

- In Indonesia, over 23% of women between the ages of 50 and 80 suffer from osteoporosis, a serious bone health condition. This condition may lead to fractures that affect quality of life. The disadvantages of traditional bone mineral density (BMD) measurement methods, like DEXA and QCT, include high costs and radiation exposure. By combining Deep Learning technology with Electrical Impedance Spectroscopy (EIS), this study seeks to provide a different, safer, and more portable approach. A total of 12,480 measurement data were obtained by collecting EIS data from four levels of bone density (30%, 50%, 70%, and 90%).This process involves signal pre-processing using end point detection techniques and artificial feature extraction. Five machine learning algorithms Decision Tree, Naive Bayes, Support Vector Machine, Logistic Regression, and XGBoost are used for bone density classification. As a result, the integration of EIS with machine learning algorithms has been shown to improve the accuracy of osteoporosis detection, where XGBoost shows the best performance and is able to overcome overfitting. This innovation is expected to expand access to osteoporosis diagnosis in health facilities, especially in developing countries like Indonesia.

Published

2026-10-22