# Study on predicting microsatellite instability in rectal cancer using T2 mapping combined with ADC value

## TL;DR

This study shows that combining T2 mapping and ADC values from MRI can accurately predict microsatellite instability in rectal cancer before surgery.

## Contribution

A novel combined model using T2 and ADC values improves prediction of MSI status in rectal cancer compared to individual metrics.

## Key findings

- MSI tumors had significantly lower T2 values and higher ADC values compared to MSS tumors.
- The combined model achieved an AUC of 0.915, outperforming individual T2 and ADC metrics.
- The model demonstrated excellent stability with Bootstrap mean AUC of 0.913.

## Abstract

To investigate the feasibility of T₂ mapping combined with apparent diffusion coefficient (ADC) value for preoperatively predicting microsatellite instability (MSI) status in rectal cancer.

This retrospective study included 152 patients with pathologically confirmed rectal cancer (40 MSI, 112 microsatellite stable [MSS]). All patients underwent MRI including T₂ mapping and diffusion-weighted imaging (DWI). Tumor T₂ and ADC values were measured and compared. A combined prediction model was constructed using multivariate logistic regression with SMOTE oversampling and L2 regularization. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curve analysis, with model stability assessed via 10-fold cross-validation and Bootstrap resampling (1000 times).

The MSI group had a significantly lower T₂ value (92.18 ± 7.21 ms vs. 99.47 ± 7.85 ms, p < 0.001) and a higher ADC value (1.06 ± 0.18 vs. 0.91 ± 0.19 × 10⁻³ mm²/s, p < 0.001) compared to the MSS group. The AUC for predicting MSI status was 0.865 for T₂ value and 0.741 for ADC value. The combined model significantly improved the AUC to 0.915 (95% CI: 0.865–0.965), with a sensitivity of 82.5% and specificity of 89.3%. The model demonstrated excellent stability (Bootstrap mean AUC = 0.913).

T₂ mapping combined with ADC value provides a reliable, non-invasive method for preoperative prediction of MSI status in rectal cancer. The combined model demonstrates higher diagnostic efficacy than either parameter alone and shows promising potential for clinical translation to support precision treatment decision-making in conjunction with histopathological assessment.