SpineReport AI automates 3D lumbar MRI analysis with 95% accuracy

A team led by Nathan Molinier and colleagues has introduced SpineReport, an open-source framework that automates 3D quantification and reporting of lumbar spine degeneration from MRI. Traditional clinical analysis relies on 2D measurements, which suffer from poor reproducibility when anatomical structures are not aligned with the imaging plane. SpineReport overcomes this by performing robust automated segmentation of key structures—including the spinal canal, spinal cord, vertebrae, intervertebral discs, and foramina—and extracting both morphological and signal-based features. This enables cross-subject and longitudinal assessment, with subject-specific reports that allow comparison against cohort distributions.

Clinical relevance was evaluated against radiologist severity grades for central canal, lateral recess, and foraminal stenosis. The strongest associations were found for central canal stenosis, with T2-weighted cerebrospinal fluid (CSF) signal achieving an AUC of 0.95. Canal AP diameter and area ratios also showed strong discriminative ability (AUC > 0.80). For lateral recess stenosis, lateral CSF signal was most informative (AUC = 0.73), while foraminal stenosis showed no significant associations despite robust region extraction. SpineReport is released as an open-access tool, making it freely available to the research and clinical community.