Preoperative MRI Segmentation of Paraspinal Muscles Is Associated with Intraoperative Findings of Traumatic Brachial
Yun-Cong Zheng1,2,3, Johnny Chuieng-Yi Lu2,4, Tommy Nai-Jen Chang2,4
1Department of Biomedical Engineering, National Taiwan University, No. 1, Section 4, Roosevelt Rd, Zhongzheng District, Taipei City 10617, Taiwan.
Abstract:
Background Surgical decisions for traumatic brachial plexus injury (TBPI) depend on the severity of nerve root injury, especially total root avulsion and partial root avulsion, determined with MRI and intraoperative findings. However, the use of quantitative MRI indexes to distinguish normal roots, total root avulsion, and partial root avulsion remains to be explored. Purpose To develop a diagnostic system using paraspinal muscle volumetric segmentation in MRI assessments of TBPI. Materials and Methods In this single-center retrospective study, data from patients diagnosed with TBPI from December 2014 to June 2023 were reviewed. Cervical paraspinal muscles were volumetrically segmented using preoperative gadolinium-enhanced T1-weighted MRI. Muscles were labeled according to injury side (right vs left), root level (C4 through C7), and layer depth (superficial to deep layers). Diagnostic accuracy of the ratio of the segmented muscle volume on the lesion side to that on the normal side (hereafter, L/N volume ratio), neurologic and/or physical examination, and MRI reports differentiating the severity of root injury (normal roots, partial root avulsion, and total root avulsion) were evaluated using the area under the receiver operating characteristic curve (AUC) analysis, with surgical findings as the reference standard. A P value of less than .05 was considered to indicate a statistically significant difference. Results In total, 145 patients with TBPI (mean age, 30 years ± 13 [SD]; 121 men) were evaluated. The L/N volume ratio in layer 4 paraspinal muscles (semispinalis cervicis and multifidus) was associated with TBPI severity at C4 through C7 (higher in patients with normal root vs partial root avulsion: P < .001 for C4 through C6, P = .16 for C7; higher in patients with partial vs total root avulsion: P = .03 for C4, P < .001 for C5 through C7). Optimal cutoff values were 0.95 (normal vs avulsion) and 0.80 (partial vs total root avulsion) based on pooled data. Diagnostic accuracy (AUC) for normal roots versus injured roots was 0.80 with neurologic and/or physical examination, 0.85 with MRI, and 0.88 with L/N volume ratio and that for partial versus total root avulsion was 0.59, 0.76, and 0.91, respectively. The L/N volume ratio cutoffs enabled accurate diagnosis of all four roots in 55.2% (80 of 145) and three of four roots in 29.0% (42 of 145) of patients. Conclusion MRI-based volumetric analysis of paraspinal muscles in patients with TBPI accurately distinguished the severity of root injury, which was associated with intraoperative findings. © RSNA, 2025 Supplemental material is available for this article. See also the editorial by Sneag and Ensle in this issue.
