Machine Learning Base Bone Density Detection Using T-Score

Abstract

Early and accurate identification of bone-health conditions such as Osteopenia and Osteoporosis is critical for preventing fractures and improving clinical outcomes. In this study, machine-learning techniques were applied to classify bone-health status into three categories such as Normal, Osteopenia, and Osteoporosis using features extracted from biomedical signals. Six classifiers, including Naive Bayes, Support Vector Machine, Decision Tree, Gradient Boosting, Random Forest, and K-Nearest Neighbors (KNN), were systematically evaluated using confusion matrices and overall accuracy metrics. The results indicate substantial differences in model performance, with KNN achieving the highest accuracy of 70.62 %, followed by the ensemble-based Gradient Boosting and Random Forest models at 63.47 %. In contrast, traditional models such as SVM, Decision Tree, and Naive Bayes exhibited comparatively lower accuracies below 41 %. The confusion matrix analysis further revealed that osteopenia, being a transitional state between normal and osteoporotic density, was the most frequently misclassified category across several models. Overall, the findings demonstrate that non-parametric and ensemble learning methods are more effective for capturing the nonlinear and overlapping feature patterns present in biomedical data. This study underscores the potential of machine-learning-based decision-support systems for aiding early detection and classification of bone-health conditions, providing a foundation for future advancements incorporating deep learning and multimodal data integration.

Ähnliche Arbeiten