Towards A Standardized Scoring System for Lumbar MRI: Development, Validation, and Future AI Integration
Description
Back and leg pain are among the most common medical complaints, with treatment largely dependent on lumbar MRI evaluation. However, the literature documents significant variability in interpretation, with both inter-rater and intra-rater reliability concerns. This inconsistency impacts diagnostic accuracy and treatment decisions. Our study aims to develop a standardized scoring system for lumbar MRI assessment, designed by a multidisciplinary team of neurosurgeons, radiologists, and spine care specialists. By synthesizing insights from established grading systems, including the Pfirrmann and Modic classifications, we are creating an objective, reliable, and reproducible evaluation framework. Calibration will be conducted using a subset of MRI images, followed by validation on 150 diverse scans. Inter-rater reliability will be assessed using Cohen’s Kappa statistic. A larger dataset of 1,000 MRI images is being assembled for future convolutional neural network (CNN) training. The goal is to establish a validated system that enhances diagnostic consistency and accuracy. Future CNN integration is expected to further improve reliability and efficiency, ultimately optimizing clinical decision-making in spine care.
Towards A Standardized Scoring System for Lumbar MRI: Development, Validation, and Future AI Integration
Back and leg pain are among the most common medical complaints, with treatment largely dependent on lumbar MRI evaluation. However, the literature documents significant variability in interpretation, with both inter-rater and intra-rater reliability concerns. This inconsistency impacts diagnostic accuracy and treatment decisions. Our study aims to develop a standardized scoring system for lumbar MRI assessment, designed by a multidisciplinary team of neurosurgeons, radiologists, and spine care specialists. By synthesizing insights from established grading systems, including the Pfirrmann and Modic classifications, we are creating an objective, reliable, and reproducible evaluation framework. Calibration will be conducted using a subset of MRI images, followed by validation on 150 diverse scans. Inter-rater reliability will be assessed using Cohen’s Kappa statistic. A larger dataset of 1,000 MRI images is being assembled for future convolutional neural network (CNN) training. The goal is to establish a validated system that enhances diagnostic consistency and accuracy. Future CNN integration is expected to further improve reliability and efficiency, ultimately optimizing clinical decision-making in spine care.
https://scholar.uwindsor.ca/we-spark-conference/2025/postersessions/47