Divergent Prognostic Value of Primary Tumor Segmentation Metrics on Baseline FDG PET/CT in Colorectal Cancer

Colorectal cancer (CRC) remains a major global health concern, with increasing incidence and mortality projected over the coming decades. Despite the central role of staging systems, substantial heterogeneity in clinical outcomes persists among patients within the same stage, highlighting the need for additional prognostic biomarkers. This study aimed to evaluate whether segmentation-derived morphological and metabolic features of the primary tumor could serve as prognostic biomarkers associated with subsequent tumor evolution in CRC.

In this retrospective, single-center study, 91 patients with histologically confirmed CRC who underwent baseline FDG PET/CT prior to treatment were analyzed. Morphological (tumor shape, cranio-caudal extension, volume) and metabolic (SUVmean, SUVmax, MTV, TLG) parameters of the primary tumor were extracted using 3D segmentation. Clinical benefit (CB) was defined according to RECIST criteria at six months. Logistic regression and Cox proportional hazards models were applied to identify predictors of short- and long-term outcomes, with performance assessed using ROC curves and Kaplan-Meier survival analyses.

Cranio-caudal extension was the strongest prognostic biomarker of short-term clinical benefit (AUC = 0.89), with a threshold of 6.2 cm discriminating favorable from unfavorable outcomes. In multivariate analysis, early UICC stage and lower cranio-caudal extension were independently associated with CB. For long-term outcomes, MTV emerged as a consistent prognostic factor: higher MTV predicted shorter progression-free survival (HR = 1.03, p < 0.01) and overall survival (HR = 1.03, p < 0.01). In addition, UICC stage IV significantly increased the risk of progression (HR = 9.65, p < 0.01).

Segmentation of the primary tumor on baseline FDG PET/CT provides valuable prognostic information in CRC. While cranio-caudal extension was the strongest prognostic biomarker of short-term treatment response, MTV was independently associated with long-term outcomes, particularly progression-free survival. These findings highlight the complementary prognostic roles of morphological and metabolic tumor features and support the integration of PET/CT-based biomarkers into personalized treatment strategies for colorectal cancer.