Spinal Disease Classification Using Deep Learning on Dual-View Videos.

Area of Science:

• Biomechanics
• Medical Imaging
• Artificial Intelligence

Background:

• Adult spinal deformity (ASD) presents with symptoms like balance disorders, impacting patient mobility.
• Current gait observation methods for ASD lack standardized quantitative evaluation protocols.
• Accurate diagnosis of ASD is crucial for effective treatment and management.

Purpose of the Study:

• To develop and validate a novel approach for distinguishing ASD from non-ASD using AI.
• To enhance the quantitative assessment of gait dynamics in patients with ASD.
• To improve the diagnostic accuracy of gait analysis for adult spinal deformity.

Main Methods:

• Utilized a neural network trained on gait videos captured from two perspectives.
• Integrated spatiotemporal gait dynamics for comprehensive gait assessment.
• Employed a two-stream 3D Convolutional Neural Network (CNN) architecture for gait analysis.

Main Results:

• The two-perspective model showed a modest enhancement in classification accuracy compared to single-perspective models.
• Achieved a significant 10% increase in F1 score, reaching 71.86%, indicating improved diagnostic performance.
• Demonstrated the effectiveness of integrating multi-view gait data for ASD detection.

Conclusions:

• Integrating two-perspective videos significantly improves gait analysis for adult spinal deformity.
• The proposed AI-driven method offers a more comprehensive and quantitative approach to ASD diagnosis.
• Further research into video fusion techniques could further enhance diagnostic capabilities for ASD.