A Study on Intelligent Optical Bone Densitometry.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Artificial Intelligence

Background:

• Osteoporosis is a widespread condition, especially in aging populations.
• Dual-energy X-ray Absorptiometry (DXA) is the standard for diagnosis but is costly and requires expertise.
• There is a need for more accessible bone density screening methods.

Purpose of the Study:

• To develop and validate a novel, rapid screening method for bone density.
• To utilize near-infrared light and deep learning for bone density assessment.
• To provide an accessible alternative to DXA for osteoporosis screening.

Main Methods:

• Near-infrared light is used to capture local body information.
• Deep learning algorithms analyze the captured data for bone density prediction.
• Multi-linear regression was initially used to establish correlation with DXA-measured Bone Mineral Density (BMD).

Main Results:

• Initial multi-linear regression showed a strong correlation (r = 0.98, p < 0.003) with DXA BMD.
• Deep learning models predicted bone density at the wrist, hip, and spine.
• Prediction error was below 10% for the wrist and below 20% for the hip and spine.

Conclusions:

• The proposed near-infrared and deep learning method is a viable and accurate approach for bone density screening.
• This technique offers a potentially more accessible and cost-effective alternative to DXA.
• Further development could improve accessibility for widespread osteoporosis screening.