Osteoporosis screening using machine learning and electromagnetic waves.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Data Science

Background:

• Osteoporosis significantly impacts global health and economies due to fractures.
• Dual-energy X-ray absorptiometry (DXA) is the diagnostic standard but is costly and inaccessible in many regions.
• There is a need for accessible, low-cost osteoporosis screening tools.

Purpose of the Study:

• To evaluate the performance of Osseus, a novel, low-cost portable device for osteoporosis screening.
• To assess the efficacy of machine learning models in predicting bone mineral density using Osseus measurements and risk factors.
• To compare Osseus-based screening with DXA, the current gold standard.

Main Methods:

• Osseus device measures signal attenuation through the medial phalanx of the middle finger.
• Supervised classification models, including Random Forest, were trained using Osseus data and patient risk factors.
• Dual-energy X-ray absorptiometry (DXA) results served as the ground truth for model training and validation.
• A dataset of 505 patients was used, with 5-fold cross-validation and 20% for testing.

Main Results:

• The best performing model, Random Forest, achieved a sensitivity of 0.853, specificity of 0.879, and F1 score of 0.859.
• Key predictors identified by the Random Forest model were age, body mass index, and Osseus signal attenuation.
• The study demonstrates the effectiveness of Osseus in conjunction with machine learning for osteoporosis screening.

Conclusions:

• The Osseus device, coupled with Random Forest analysis, offers an effective and accessible method for early osteoporosis screening.
• This approach can significantly reduce healthcare costs associated with osteoporosis diagnosis, treatment, and hospitalization.
• Early detection through accessible screening can improve patient quality of life by enabling timely intervention.