Deep Learning Model for Classifying Metastatic Epidural Spinal Cord Compression on MRI.

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Metastatic epidural spinal cord compression (MESCC) is a severe complication of advanced cancer.
• Early diagnosis and referral are crucial for managing MESCC.
• Automated classification of MESCC on MRI using deep learning (DL) could enhance diagnostic workflows.

Purpose of the Study:

• To develop and evaluate a DL model for automated MESCC classification on MRI.
• Assess the DL model's performance against expert classifications.

Main Methods:

• A DL model was trained on 215 MRI spine studies using axial T2-weighted images.
• Internal and external datasets were used for testing the model's performance.
• The DL model's classifications were compared to those of radiologists, a spine surgeon, and a radiation oncologist using Gwet's kappa for agreement.

Main Results:

• The DL model achieved almost perfect agreement (kappa = 0.92-0.98) with the reference standard for dichotomous MESCC classification on internal test sets.
• Similar high performance (kappa = 0.94-0.95) was observed on an external test set.
• The DL model demonstrated comparable agreement to subspecialist radiologists and clinical specialists.

Conclusions:

• A DL model can reliably classify MESCC on MRI with performance comparable to human experts.
• This AI tool has the potential to optimize earlier diagnosis and surgical referral for patients with MESCC.
• The study highlights the utility of DL in improving the management of advanced cancer complications.