Abstract
The definition of target alignment for spinal fusion surgery follows anatomical criteria and strongly relies on surgical experience. However, the optimal patient-specific alignment often remains unknown. Statistical models could provide information about physiological alignments, and musculoskeletal models are powerful tools to investigate biomechanics. We aimed to statistically predict alignments and hypothesized they would be biomechanically favorable. A statistical model was trained with 60 annotated radiographs to predict physiological sagittal alignment based on position of femoral heads and sacrum. Predicted alignments for 11 back pain patients were clinically evaluated in terms of balance and compared to Original alignments. The normative ranges for spinal balance parameters were obtained from Surgimap™. Musculoskeletal loads were furthermore simulated in upright standing and 30° forward flexion, using alignment-specific musculoskeletal models. For the majority of Predicted alignments (n = 9) at least two of three investigated balance parameters were within the normative range, as opposed to the minority of the Original alignments (n = 4). Predicted alignments resulted in significantly lowered overall muscle activity and compressive loads (all levels, both postures). Shear force magnitudes in upright standing decreased significantly at levels L1L2 (-68 N) and L2L3 (-69 N) and clearly yet not significantly at L3L4 (-39 N) and L4L5 (-152 N). Shear loads at level L5S1 remained the same. In flexed postures identical trends were observed. The statistical model was able to predict spinal alignments that led to both improved balance and reduced musculoskeletal loads. Further studies are needed to investigate clinical validity of such models.
Caprara, Sebastiano