Related Experiment Video
Updated: Jun 2, 2026
Segmentation and Linear Measurement for Body Composition Analysis using Slice-O-Matic and Horos
Published on: March 21, 2021
Artificial Intelligence-Based Body Composition Analysis Reveals Sex-Specific Prognostic Markers and Their Clinical
Tianxiang Li1,2,3, Tingting Liu1,2,3, Qian Yang4
1Department of Radiology, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China.
Background And Aims:
Conventional body composition assessment fails to capture its multidimensional complexity in gastric cancer (GC). This study aimed to systematically evaluate multidimensional body composition and its clinical relevance in GC.
Methods:
1196 GC patients and 983 healthy controls were retrospectively enrolled. Body composition was segmented using nnU-Net. Propensity score matching was used to compare body composition differences between patients and healthy controls. Prognostic value was assessed across treatment cohorts. TCGA pathological and transcriptomic data were integrated for exploratory analyses. Sex-specific prognostic models were developed and externally validated.
Results:
AI-drived body compositions parameters showed sex-specific associations with survival. Higher muscle/fat area/index and lower fat density correlated with longer survival in surgical patients (L1 muscle area: HR = 0.438, ad p = 0.001). In ICI-treated females, higher subcutaneous fat area (SFA) improved survival (L2SFA: HR = 0.972, ad p = 0.045). Potential associations were observed between body composition and tumor microenvironmental features, including stromal composition, EMT-related pathways, and immune infiltration. Sex-specific prognostic models achieved C-indices of 0.723 (males) and 0.705 (females) in test cohorts, outperforming conventional predictors.
Conclusion:
Multidimensional body composition is associated with prognosis in GC and may serve as a complementary biomarker for risk stratification.
