AI Framework Achieves 90% Accuracy in Lumbar Spinal Stenosis Diagnosis With Explainable Results

Researchers developed an explainable vision-language model that diagnoses lumbar spinal stenosis from MRI scans with 90.69% accuracy while generating radiologist-style clinical reports. The framework uses a novel adaptive loss function to handle class imbalance in medical imaging datasets, addressing a critical clinical challenge that currently relies on labor-intensive manual interpretation. The explainable AI approach balances diagnostic precision with transparency, potentially reducing inter-observer variability and diagnostic delays in spine imaging.