Spectral CT-derived extracellular volume fraction for pathological characterization and noninvasive T staging in colon cancer.
Authors
Affiliations (5)
Affiliations (5)
- Department of Radiology, The Affiliated Hospital, Southwest Medical University, Luzhou, Sichuan, China; Department of Radiology, Luzhou People's Hospital, Luzhou, Sichuan, China.
- Department of Radiology, The Affiliated Hospital, Southwest Medical University, Luzhou, Sichuan, China; Precision Imaging and Intelligent Analysis Key Laboratory of Luzhou, Southwest Medical University, Luzhou, Sichuan, China.
- Department of Radiology, Suining People's Hospital, Suining, Sichuan, China.
- Department of Oncology, The Affiliated Hospital, Southwest Medical University, Luzhou, Sichuan, China.
- Department of Radiology, The Affiliated Hospital, Southwest Medical University, Luzhou, Sichuan, China; Precision Imaging and Intelligent Analysis Key Laboratory of Luzhou, Southwest Medical University, Luzhou, Sichuan, China. Electronic address: [email protected].
Abstract
To investigate the association between the extracellular volume fraction (ECV) quantified by spectral CT and multiparameter pathological features (Ki-67, mismatch repair status, WHO Grade, Perineural/Lymphovascular Invasion, pT/pN Stage) in colon cancer, and to develop a noninvasive preoperative prediction model for any indicator showing a significant association. This single-center retrospective study enrolled 158 patients with pathologically confirmed colon cancer. All patients underwent contrast-enhanced spectral CT and hematocrit testing within 1 week before surgery for ECV calculation. Univariate logistic regression was used to evaluate associations between ECV and each pathological feature. For pathological features showing significant associations, three predictive models were developed: a clinical model, an ECV-only model and a fusion model. Models were trained using elastic net-regularized logistic regression with stratified 3-fold cross-validation. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), Brier score, and calibration. Univariate analysis showed that spectral CT-derived ECV was significantly associated with pathological T (pT) stage (OR = 1.07, 95% CI: 1.01-1.13, P = 0.02), with higher ECV values in advanced pT3-T4 tumors. No statistically significant associations were observed with other pathological features (P > 0.05). For pT staging prediction, the fusion model achieved the highest discrimination performance (mean AUC = 0.815, 95% CI: 0.69-0.96) and best calibration (Brier score = 0.225), numerically outperforming the clinical model (AUC = 0.772) and the ECV-only model (AUC = 0.699). Decision curve analysis suggested that the fusion model provided higher net clinical benefit than single-modality approaches. Spectral CT-derived ECV is associated with invasion depth in colon cancer. A multimodal model integrating ECV with clinical indicators showed promising internal performance for preoperative pT risk estimation, providing proof-of-concept for future validation studies.