Main Session

Sep 29

PQA 07 - Head and Neck Cancer, Lung Cancer/Thoracic Malignancies, and Nursing and Supportive Care

3524 - Automated Volumetric Muscle Assessment Refines Identification of High-Risk Sarcopenia in Head and Neck Radiotherapy

03:45pm - 05:00pm ET

Poster Hall - Exhibit Hall A

Screen: 4

POSTER

Presenter(s)

Sang Min Lee, MD - Seoul National University College of Medicine, Seoul, Seoul

S. M. Lee¹, W. Y. Suh¹, H. G. Wu¹, J. H. Kim¹, K. Y. Eom², and J. H. Lee¹; ¹Department of Radiation Oncology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea, Republic of (South), ²Department of Radiation Oncology, Seoul National University Bundang Hospital, Seongnam, Korea, Republic of (South)

Purpose/Objective(s):

Accurate identification of high-risk sarcopenia during radiotherapy (RT) for head and neck cancer (HNC) remains challenging, as conventional L3-based skeletal muscle index (SMI) relies on single-slice measurements that may inadequately represent whole-body muscle burden. We evaluated whether automated deep learning–based volumetric skeletal muscle assessment refines identification of patients with high-risk sarcopenia and improves prognostic stratification compared with conventional L3-based methods.

Materials/Methods:

A total of 227 patients with HNC treated with RT or chemoradiotherapy between 2006 and 2018 at two institutions were retrospectively analyzed. Whole body computed tomography (CT) images obtained from routine staging or treatment work-up PET-CT scans performed before and after RT were analyzed using a deep learning–based automated segmentation. Whole-body skeletal muscle index (WB-SMI) was calculated as skeletal muscle volume (cm³) between the vertex and the ischial tuberosity, normalized by scan length (cm) and height squared (m²). Sarcopenia was defined using WB-SMI cut-offs derived from maximally selected rank statistics, while conventional sarcopenia was defined using established L3-based cut-offs. Overall survival (OS) was calculated from the completion date of RT.

Results:

With a median follow-up of 63.3 months, the 5-year OS rate was 80.5%. WB-SMI significantly decreased after treatment (mean difference, 5.12 cm²/m²; 95% confidence interval [CI], 4.38–5.85; p < 0.001). In multivariable analysis, post-treatment WB-SMI remained independently associated with OS (HR, 0.93; 95% CI, 0.88–0.99; p = 0.030), along with age, sex, and baseline albumin. Using WB-SMI cut-offs optimized for OS, sarcopenia was identified in 28.2% of patients before treatment and 56.8% after treatment. WB-SMI demonstrated superior prognostic stratification for OS compared with conventional L3-based SMI (Harrell’s C-index, 0.64 vs. 0.56; p = 0.029). Notably, WB-SMI identified 48.9% of high-risk patients missed by L3-based assessment, and this reclassified group demonstrated a higher risk of admission due to treatment-related toxicity (odds ratio, 1.92; 95% CI, 1.25–2.94; p = 0.003). OS differed significantly among muscle trajectory groups, with 5-year OS rates of 90.8%, 80.1%, and 63.2% for the muscle maintenance, muscle loss, and persistent sarcopenia groups, respectively (p < 0.001).

Conclusion:

Automated volumetric skeletal muscle assessment refines identification of high-risk sarcopenia and improves prognostic discrimination beyond conventional L3-based methods in HNC patients undergoing RT. Integration of volumetric muscle quantification into routine imaging workflows may enhance risk-adapted supportive care strategies and improve treatment tolerance and survival.