Evaluation of the prognostic impact of artificial intelligence–based body composition assessment in colorectal cancer

Colorectal cancer remains one of the most common malignancies worldwide and represents a major public health concern due to its high mortality rates. Recent studies have suggested that body composition parameters may have a significant impact on cancer prognosis. However, manual quantification methods are time-consuming and observer-dependent, limiting their integration into clinical practice. Therefore, this study aimed to evaluate the prognostic significance of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) volumes measured using artificial intelligence (AI)–based automatic segmentation in patients with colorectal cancer. A total of 113 patients diagnosed with CRC between 2022 and 2025 at Alanya Alaaddin Keykubat University Faculty of Medicine, Department of Oncology, were included in this retrospective study. VAT and SAT volumes were quantified from abdominal CT scans using the RECOMIA platform, which employs a deep learning–based automatic segmentation algorithm. The relationships between body composition parameters and survival outcomes were analyzed using ROC, Kaplan–Meier, and Cox regression models. The analyses revealed that higher SAT levels were significantly associated with improved overall survival (OS) and progression-free survival (PFS) (OS HR=0.17, p=0.005; PFS HR=0.46, p=0.042). Multivariate Cox analysis confirmed that elevated SAT was an independent favorable prognostic factor. In contrast, VAT and BMI values showed no statistically significant associations with OS or PFS. In conclusion, the amount of subcutaneous adipose tissue (SAT) measured through artificial intelligence (AI)–assisted volumetric analysis emerges as an important prognostic marker in patients with colorectal cancer. AI-based automatic segmentation provides substantially faster, more reproducible, and observer-independent results compared with manual measurements, offering a significant advantage for clinical integration. This approach may contribute to improved patient risk stratification and the development of personalized treatment strategies. Keywords: Colorectal cancer, artificial intelligence, body composition, subcutaneous adipose tissue, survival