Related Experiment Video
Updated: Aug 6, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
A feature-efficient dual-task machine learning framework for predicting bone mineral density and osteoporosis
Alina Maryum1, Arslan Shaukat1, Ehsan Yousaf2
1Department of Computer and Software Engineering, National University of Sciences and Technology (NUST), Islamabad, Pakistan.
Plos One
|July 20, 2026
Summary
This study introduces a machine learning model for early osteoporosis detection using accessible clinical markers like blood type and serum levels. The model achieves high accuracy, offering a cost-effective, non-invasive screening tool for widespread use.
Area of Science:
- Biomedical Engineering
- Computational Biology
- Medical Diagnostics
Background:
- Osteoporosis is a prevalent skeletal disorder marked by bone density loss and structural decline, leading to increased fracture risk.
- Current diagnostic methods like Dual-Energy X-ray Absorptiometry (DXA) have limitations including high cost, limited accessibility, and inability to detect vertebral fractures directly.
- Early detection of osteoporosis is crucial but challenging due to its asymptomatic nature and the limitations of conventional diagnostics.
Purpose of the Study:
- To develop a cost-effective and accessible machine learning model for early osteoporosis screening.
- To utilize basic clinical markers, avoiding expensive imaging techniques for wider applicability, especially in resource-limited settings.
- To create a non-invasive, radiation-free complementary tool for osteoporosis screening and fracture risk assessment.
Main Methods:
- A dataset of 159 patient records was utilized, incorporating demographics, genetic/blood type, clinical history, and lab test parameters.
- Data preprocessing included feature selection, standardization, hyperparameter tuning, and derivation of clinically relevant features and biomarker combinations.
- An ensemble voting classifier was trained on Weight, Potassium, Calcium, and Total Vitamin D for osteoporosis severity prediction, and an extreme gradient boosting Regressor was trained on Age, Weight, ABO Group, and Total Vitamin D for lumbar spine Bone Mineral Density (BMD) estimation.
Main Results:
- The ensemble voting classifier achieved 90% accuracy and an AU-ROC score of 0.93 in predicting osteoporosis severity.
- The extreme gradient boosting Regressor demonstrated an R2 of 0.536 for lumbar spine BMD estimation.
- The model effectively uses accessible markers like ABO blood groups, serum calcium, and potassium levels, removing financial and technical barriers to screening.
Conclusions:
- Machine learning can be leveraged for non-invasive osteoporosis screening and fracture risk assessment.
- The proposed framework offers a radiation-free, clinically accessible, and cost-effective complementary pre-screening tool.
- This approach enhances early detection possibilities, particularly in remote or underfunded healthcare settings.