Abstract
PURPOSE: This finite element analysis assessed lateral compression (LC-1) fracture stability using machine learning for morphological mapping and classification of pelvic ring stability. METHODS: Computed tomography (CT) files of LC-1 pelvic fractures were collected. After morphological mapping and producing matrix data, we used K-means clustering in unsupervised machine learning to classify the fractures. Based on these subtypes, we manually added fracture lines in ANSYS software. Finally, we performed a finite element analysis of a normal pelvis and eight fracture subtypes based on von Mises stress and total deformation changes. RESULTS: A total of 218 consecutive cases were analyzed. According to the three main factors-zone of sacral injury and completion, pubic ramus injury side, and the sagittal rotation of the injured hemipelvis-the LC-1 injuries were classified into eight subtypes (I-VIII). No significant differences in stress or deformation were observed between unilateral and bilateral public ramus fractures. Subtypes VI and VIII showed the maximum stress while subtypes V-VIII showed the maximum deformation in the total pelvis and sacrum. The subtypes did not differ in superior public ramus deformation. CONCLUSIONS: Complete fracture of sacrum zones 2/3 may be a feature of unstable LC-1 fractures. Surgeons should give surgical strategies for subtypes V-VIII.