Objectives Lung cancer’s propensity for spinal metastasis leads to fractures, dysfunction, pain, and reduced quality of life. Spinal interventions are selectively offered to patients deemed fit for surgery. Sarcopenia, assessed by psoas muscle (PM) and whole-abdominal muscle (WAM) measurements, is proposed as a fitness marker, but consensus on thresholds and segmentation tools is lacking. This study aims to validate sarcopenia metrics as imaging biomarkers using both open-source and locally tailored neural networks in the context of bone-metastatic lung cancer and spinal surgery. Materials and methods A retrospective cohort of 63 lung cancer patients (age 64 ± 9, 46% female) with spinal metastases who underwent surgery between 2010 and 2020 was analyzed. PM and lumbar vertebrae segmentation were validated by a musculoskeletal radiologist on CT scans. A local PM segmentation model was trained using nnUNet, and TotalSegmentator (TS) was used for PM and WAM segmentation. Sarcopenia metrics (i.e., PLVI, PM L4 vertebral index (PLVI), psoas muscle index (PMI), skeletal muscle index (SMI), and total muscle area (TMA)) and radiomic features were evaluated. Survival analysis was conducted based on sarcopenia classification using the Wilcoxon log-rank test. Results The locally tailored psoas segmentation model outperformed TS in seven metrics. PMI and PLVI thresholds showed significant survival differences only when measured with the local model (p < 0.05), but not SMI or TMA. Percentile-based classification revealed significant survival differences, especially in local PM metrics (p < 0.001). Of 108 radiomic feature clusters, 38 showed significance with the local models, whereas none did with TS WAM segmentation. Conclusion The locally tailored model demonstrated superior performance compared to TS. Percentile-based thresholds and PM features were more predictive of survival, underscoring the need for disease-specific cutoffs. Radiomic features warrant further investigation. Key Points <Emphasis Type="BoldItalic">Question</Emphasis> Can automated computed tomography assessment of sarcopenia via PM segmentation predict surgical fitness in patients with metastatic lung cancer undergoing spinal surgery? <Emphasis Type="BoldItalic">Findings</Emphasis> Locally tailored PM segmentations validated known sarcopenia metrics and demonstrated greater significance in predicting patient survival compared to WAM segmentation, using both percentile-based and radiomics features. <Emphasis Type="BoldItalic">Clinical relevance</Emphasis> The decision to proceed with surgery requires a confident assessment of surgical fitness. As in other surgical contexts, PM sarcopenia assessment through imaging has been validated as a predictor of post-operative survival. Graphical Abstract

Automated psoas muscle segmentation: imaging features and surgical fitness in spinal metastatic lung cancer

期刊论文
发表日期：

2025年8月21日

查看全文

Marco Pérez Caceres,
Omer Ahmed,
Véronique Freire,
Jesse Shen,
Fidaa Al-Shakfa,
Danielle Boulé,
Zhi Wang

摘要

Objectives

Lung cancer’s propensity for spinal metastasis leads to fractures, dysfunction, pain, and reduced quality of life. Spinal interventions are selectively offered to patients deemed fit for surgery. Sarcopenia, assessed by psoas muscle (PM) and whole-abdominal muscle (WAM) measurements, is proposed as a fitness marker, but consensus on thresholds and segmentation tools is lacking. This study aims to validate sarcopenia metrics as imaging biomarkers using both open-source and locally tailored neural networks in the context of bone-metastatic lung cancer and spinal surgery.

Materials and methods

A retrospective cohort of 63 lung cancer patients (age 64 ± 9, 46% female) with spinal metastases who underwent surgery between 2010 and 2020 was analyzed. PM and lumbar vertebrae segmentation were validated by a musculoskeletal radiologist on CT scans. A local PM segmentation model was trained using nnUNet, and TotalSegmentator (TS) was used for PM and WAM segmentation. Sarcopenia metrics (i.e., PLVI, PM L4 vertebral index (PLVI), psoas muscle index (PMI), skeletal muscle index (SMI), and total muscle area (TMA)) and radiomic features were evaluated. Survival analysis was conducted based on sarcopenia classification using the Wilcoxon log-rank test.

Results

The locally tailored psoas segmentation model outperformed TS in seven metrics. PMI and PLVI thresholds showed significant survival differences only when measured with the local model (p < 0.05), but not SMI or TMA. Percentile-based classification revealed significant survival differences, especially in local PM metrics (p < 0.001). Of 108 radiomic feature clusters, 38 showed significance with the local models, whereas none did with TS WAM segmentation.

Conclusion

The locally tailored model demonstrated superior performance compared to TS. Percentile-based thresholds and PM features were more predictive of survival, underscoring the need for disease-specific cutoffs. Radiomic features warrant further investigation.

Key Points

Question Can automated computed tomography assessment of sarcopenia via PM segmentation predict surgical fitness in patients with metastatic lung cancer undergoing spinal surgery?

Findings Locally tailored PM segmentations validated known sarcopenia metrics and demonstrated greater significance in predicting patient survival compared to WAM segmentation, using both percentile-based and radiomics features.

Clinical relevance The decision to proceed with surgery requires a confident assessment of surgical fitness. As in other surgical contexts, PM sarcopenia assessment through imaging has been validated as a predictor of post-operative survival.

Graphical Abstract