U-Net-Based Deep Learning Hybrid Model: Research and Evaluation for Precise Prediction of Spinal Bone Density on...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Bone Metabolism

Background:

• Osteoporosis, a metabolic bone disorder, causes progressive bone loss and increases fracture risk.
• Current bone mineral density assessment methods like DXA and QCT have limitations, especially for high-risk female populations.
• There is a need for improved, accessible, and accurate methods for osteoporosis diagnosis.

Purpose of the Study:

• To develop and validate a hybrid AI model for precise L2 vertebra bone mineral density measurement.
• To address limitations of existing bone density assessment techniques in female patients.
• To enhance skeletal analysis using abdominal X-ray images.

Main Methods:

• A hybrid model combining U-Net for image preprocessing and artificial neural networks for analysis was developed.
• The model focuses on anteroposterior abdominal X-ray images of female patients.
• U-Net enhanced bone features by reducing noise, and ANNs performed nonlinear regression for bone mineral density prediction.

Main Results:

• The hybrid model achieved a high correlation coefficient of 0.77.
• A low mean absolute error of 0.08 g/cm² was recorded, indicating high accuracy.
• The model demonstrated significant effectiveness, particularly compared to dual-energy X-ray absorptiometry.

Conclusions:

• The proposed hybrid U-Net and ANN model offers a promising, accurate, and potentially more accessible method for assessing bone mineral density.
• This AI-driven approach shows significant potential for improving osteoporosis diagnosis in at-risk female populations.
• The study highlights the effectiveness of integrating advanced imaging techniques with AI for skeletal health assessment.