Artificial Intelligence-Assisted Three-Dimensional Morphometric Analysis of Lumbar Vertebral Canal and Pedicle Parameters in Indian Adults Using Computed Tomography

Background: Anatomy research is increasingly moving from traditional descriptive cadaveric work to clinically relevant imaging-based morphometric analysis supported by digital tools. Lumbar vertebral and pedicle dimensions are important for spinal instrumentation, minimally invasive fixation, radiological interpretation, and understanding population-specific anatomical variation. Most available morphometric datasets are based on Western populations, small cadaveric samples, or manual measurements. Artificial intelligence-assisted image segmentation and three-dimensional reconstruction may improve anatomical precision and reproducibility, but region-specific evidence remains limited. Aim:To evaluate lumbar vertebral canal and pedicle morphometry in Indian adults using artificial intelligence-assisted computed tomography-based three-dimensional analysis, and to assess sex-wise and vertebral level-wise variation. Materials and Methods:This cross-sectional observational study was designed on archived abdominal or lumbosacral computed tomography scans of 240 Indian adults aged 20–70 years. Cases with congenital vertebral anomalies, fractures, severe deformity, prior spinal surgery, tumour, or destructive pathology were excluded. DICOM data were processed using semi-automated artificial intelligence-assisted segmentation software to reconstruct L1 to L5 vertebrae in three dimensions. The primary parameters included transverse canal diameter, midsagittal canal diameter, pedicle width, pedicle height, pedicle transverse angle, chord length, and vertebral body transverse diameter. Each parameter was measured bilaterally where appropriate. Inter-observer and intra-observer reliability were assessed in a random 15% subset. Statistical analysis included descriptive statistics, independent t test, repeated-measures analysis of variance, and intraclass correlation coefficient. Results:Pedicle width increased progressively from L1 to L5, while pedicle height showed a decreasing trend caudally. Transverse canal diameter was greatest at lower lumbar levels, whereas midsagittal canal diameter demonstrated smaller interlevel variation. Male participants showed significantly larger pedicle width, chord length, and vertebral body transverse diameter than female participants at most lumbar levels. Pedicle transverse angle increased from upper to lower lumbar vertebrae, with the highest values observed at L5. Artificial intelligence-assisted segmentation produced excellent measurement reliability, with intraclass correlation coefficients above 0.90 for major parameters. The findings indicate clear population-specific dimensional patterns relevant to transpedicular screw placement and spinal canal assessment. Conclusion:Artificial intelligence-assisted three-dimensional CT morphometry provides robust and reproducible anatomical data on the lumbar vertebrae. The present study demonstrates significant sex-based and vertebral level-based variation in lumbar canal and pedicle parameters among Indian adults. These findings may support safer implant selection, preoperative planning, and development of population-specific spinal instrumentation guidelines.