A multivariate model of IVIM-DWI in the preoperative diagnosis of tumor budding grade in rectal cancer

TL;DR

This study shows that a multivariate model using IVIM-DWI and clinical factors can noninvasively predict tumor budding grade in rectal cancer before surgery.

Contribution

A novel noninvasive multivariate model combining IVIM-DWI parameters and clinical factors for preoperative tumor budding grade diagnosis in rectal cancer.

Findings

The combined model achieved an AUC of 0.920 with high sensitivity and specificity.

Parameters f, DDC, MRF, and age were identified as independent risk factors for tumor budding grade.

IVIM-DWI parameters like f and DDC showed significant differences between tumor budding grades.

Abstract

The aim of this study was to explore the application value of intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) in the preoperative evaluation of the tumor budding (TB) grade in patients with rectal cancer (RC).

Patients with RC who underwent rectum resection from January 2018 to October 2023 were collected retrospectively. All patients underwent magnetic resonance (MR) examination, including collection of IVIM sequences, within 1 week before surgery. Among them, 17 low-grade and 13 intermediate-grade budding cases were classified into the low-intermediate-grade group, while 30 high-grade budding cases were classified as high grade. After processing the IVIM sequences, the apparent diffusion coefficient (ADC) from the mono-exponential (ME) model; D, D*, and f from the bi-exponential (BE) model; and the DDC and α from the stretching (SE) model were obtained. Clinical factors, including age, gender, and CEA levels, and imaging factors, including location, mriT, and N stage, were collected. Differences between the low-intermediate- and high-grade groups were compared. The diagnostic efficiency was evaluated from the ROC curve, AUC, sensitivity, and specificity. Significance was set at p < 0.05.

A total of 60 patients with RC were enrolled. Significant differences between low-intermediate and high grade were found in f value (p = 0.001), DDC value (p < 0.001), age (p = 0.002), location (p = 0.047), and mesorectal fascia (MRF) (p = 0.01). Multivariate binary logistic regression analysis identified f, DDC, MRF, and age as independent risk factors (p < 0.05). The AUC of the combined model (f, DDC, MRF, and age) was 0.920 (95% CI, 0.820–0.974), with a sensitivity and specificity of 1.000 and 0.767, respectively.

The combined model constructed using multiple IVIM mathematical model parameters and clinical and imaging factors could be used as a noninvasive tool to facilitate the preoperative auxiliary diagnosis of TB in RC.

Figures