Deep learning models for MRI-based clinical decision support in cervical spine degenerative diseases

2024

Deep Learning Models for MRI-Based Clinical Decision Support in Cervical Spine Diseases

Sample size: 189 publication 10 minutes Evidence: high

Author Information

Author(s): Li Kai-Yu, Lu Zhe-Yang, Tian Yu-Han, Liu Xiao-Peng, Zhang Ye-Kai, Qiu Jia-Wei, Li Hua-Lin, Zhang Yu-Long, Huang Jia-Wei, Ye Hao-Bo, Tian Nai Feng

Primary Institution: Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University

Hypothesis

Can deep learning models based on MRI provide consistent treatment recommendations for cervical spine degenerative disorders?

Conclusion

The developed deep learning algorithms show substantial consistency with experienced spine surgeons in providing clinical recommendations based on cervical spine MRI.

Supporting Evidence

The study collected 756 MR images from 189 patients to develop the deep learning models.
DL version 1 achieved a Cohen’s Kappa value of 0.874, indicating almost perfect consistency with spine surgeons.
The external validation set included 30 patients and 120 MR images, demonstrating the model's applicability.
Statistical analysis showed significant differences with p-values less than 0.05.
The model's performance was validated using multiple CNN architectures.

Takeaway

Researchers created a computer program that looks at MRI images of the neck to help doctors decide if surgery is needed for neck problems, and it works well with what experienced doctors recommend.

Methodology

The study used MRI images from patients with cervical spine degenerative disorders, divided into training, validation, and external validation sets, to develop and assess deep learning models.