Deep Learning Assisted Classification of T1ρ-MR Based Intervertebral Disc Degeneration Phases

Area of Science:

• Biomedical Imaging
• Artificial Intelligence in Medicine
• Spine Degenerative Diseases

Background:

• Intervertebral disc degeneration (IDD) can be classified into three phases using T1ρ-Magnetic Resonance (MR) imaging.
• T1ρ-MR is not routinely used due to long scan times and high costs, limiting its clinical application for IDD staging.
• Routine T1- and T2-MR sequences are commonly used for IDD assessment.

Purpose of the Study:

• To develop a deep learning model for classifying IDD phases based on T1ρ-MR parameters from conventional T1-MR images.
• To enable the use of readily available T1-MR scans for advanced IDD staging.

Main Methods:

• A retrospective study involving 60 patients with low back pain or radiculopathy.
• A deep learning model was trained on T1-MR images to predict T1ρ-MR based IDD phases.
• The model's performance was evaluated using metrics including accuracy, Area Under the Receiver Operating Characteristic Curve (AUC), and average precision.

Main Results:

• The deep learning model achieved high accuracy in classifying IDD phases from T1-MR images.
• The model demonstrated a diagnostic accuracy of 0.840 and an AUC of 0.871 for T1ρ-MR based IDD phase classification.
• The model achieved a mean average precision of 0.996 for detecting intervertebral disc (IVD) levels.

Conclusions:

• A deep learning model can effectively classify T1ρ-MR based IDD phases using routine T1-MR images.
• This approach may facilitate the broader clinical application of T1ρ-MR imaging in the management of IDD.
• The developed model shows promise for improving the assessment and timing of treatments for IDD.