Channel-Boosted and Transfer Learning Convolutional Neural Network-Based Osteoporosis Detection from CT Scan, Dual...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Bone Health Diagnostics

Background:

• Osteoporosis is characterized by diminished bone density due to an imbalance between bone formation and resorption.
• Accurate diagnosis of osteoporosis is crucial for timely intervention and management.
• Current diagnostic methods rely on medical imaging like CT scans, dual X-ray, and X-ray, with varying effectiveness.

Purpose of the Study:

• To develop a novel framework, CBTCNNOD, for unified osteoporosis diagnosis across multiple imaging modalities.
• To integrate advanced modules for improved diagnostic accuracy and efficiency.
• To provide a robust system for generating clear osteoporosis diagnostic reports.

Main Methods:

• The proposed CBTCNNOD framework integrates three functional modules: a bilinear filter, grey-level zone length matrix, and a convolutional neural network (CB-CNN).
• This integrated approach processes CT scans, X-ray, and dual X-ray images.
• The system is designed to deliver precise diagnostic outcomes based on image analysis.

Main Results:

• CBTCNNOD demonstrated significant improvements in diagnostic performance compared to existing techniques (RCETA, BMCOFA, BACBCT, XSFCV).
• Accuracy increased by up to 14.32%, precision by 16.51%, sensitivity by 12.78%, and specificity by 15.84%.
• Processing time was reduced by up to 33.52%, indicating enhanced efficiency.

Conclusions:

• The CBTCNNOD framework offers a promising, integrated solution for osteoporosis diagnosis across diverse imaging modalities.
• The system's enhanced accuracy, precision, sensitivity, and specificity, coupled with reduced processing time, highlight its clinical potential.
• Further implementation and validation are expected to solidify its role in improving osteoporosis detection.