Semi-automatic CAD for upright lumbar spine MRI: reproducibility, time efficiency, and workload reduction with Q-Spine

Abstract Purpose To evaluate the performance of Q-Spine, a semi-automatic CAD system, in improving reproducibility, efficiency, and workload in morphometric assessment of upright lumbar spine MRI. Materials and methods This retrospective study included 16 patients (6 men, 10 women; mean age 45 years; range 21–80) who underwent upright lumbar MRI. Imaging was performed on a 0.25-T open MRI system using standardized sagittal and axial sequences. Two radiologists (12 year vs 2 year experience) independently analyzed spinal canal area, sagittal canal thickness, spinal curvature, vertebral collapse, intervertebral angles, foraminal area, vertebral wedging, and spondylolisthesis index, using both manual measurements and the Q-Spine CAD system. Inter- and intra-observer agreement was assessed with Spearman’s correlation, intraclass correlation coefficients (ICC), and Concordance Correlation Coefficients (CCC). Processing times and observer workload (NASA-TLX) were compared between manual and CAD-assisted analyses. Results Compared with manual assessment, Q-Spine significantly improved inter- and intra-observer agreement, particularly for spinal canal area, intervertebral angles, vertebral wedging, and spondylolisthesis index (CCC range: 0.869–0.956 vs 0.800 for manual). Median processing time was reduced by > 90% with Q-Spine (5.3 vs 56.0 min, p < 0.0001). Subjective workload scores (NASA-TLX) were significantly lower with Q-Spine, especially for mental demand, temporal demand, and frustration. Conclusion The Q-Spine semi-automatic CAD system enhances reproducibility, dramatically reduces analysis time, and lowers observer workload in morphometric assessment of upright lumbar spine MRI. These findings highlight the potential of AI-assisted tools to improve both diagnostic consistency and workflow efficiency in dynamic spinal imaging.