Development of a nomogram for predicting extramural venous invasion in colon cancer using spectral CT multiparameter imaging.

PURPOSE: To develop a predictive model based on spectral CT-derived parameters for preoperative assessment of extramural venous invasion (EMVI) in patients with colon cancer.

METHODS

This retrospective study included 235 patients with pathologically confirmed colon cancer from two medical centers between September 2022 and July 2024 (192 from center 1 and 43 from center 2). Virtual monoenergetic images (VMIs) ranging from 40 to 120 keV (20 keV increments) were evaluated to identify the optimal energy level based on tumor CT value, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR). On the optimal VMI, spectral CT quantitative parameters were measured in the arterial, venous, and delay phases, and dynamic multiphase parameters were calculated. Patients from center 1 were randomly divided into a training cohort and an internal testing cohort at a 7:3 ratio. The nomogram model was developed using the training cohort and subsequently evaluated in the internal testing cohort and an independent external validation cohort from center 2.

RESULTS

(1) The 60 keV VMI demonstrated superior image quality, showing significantly higher tumor CT value, SNR, and CNR compared with other monoenergetic images (all P < 0.05). (2) Compared with EMVI negative patients, positive patients exhibited significantly higher values of ΔHU40 (delay-arterial), ΔHU60 (delay-arterial), ΔIC (delay-arterial), delay enhancement fraction (DEF), and ΔZeff (venous-arterial) (all P < 0.05). Multivariable analysis further identified these five parameters as independent predictors of EMVI. (3) A nomogram was constructed based on these parameters and demonstrated AUCs of 0.93, 0.91, and 0.84 in the training, testing, and external validation cohorts, respectively.

CONCLUSIONS

The 60 keV VMI provides improved lesion delineation in colon cancer. A nomogram incorporating dynamic multiphase spectral CT parameters enables accurate preoperative prediction of EMVI and may support individualized clinical decision-making.