A multivariate IVIM-DWI-based model for preoperative prediction of perineural invasion status in rectal cancer: a feasibility study

AI Summary

Multivariate IVIM-DWI identified ADC and D as independent preoperative predictors of perineural invasion in rectal cancer.
Combined ADC+D model achieved AUC 0.85, sensitivity 86.10%, specificity 73.70%, accuracy 77.00%, outperforming ADC alone on reclassification and calibration.
Fivefold internal validation mean AUC 0.84 ± 0.04; decision curve analysis showed higher net benefit for the combined model across 0–0.50 thresholds.

AI Summary

OBJECTIVE: To evaluate intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) for preoperative diagnosis of perineural invasion (PNI) in rectal cancer (RC). Net reclassification improvement (NRI) and integrated discrimination improvement (IDI), along with calibration metrics and decision curve analysis, were used to further evaluate model performance. Further analyses revealed that the combined model provided significant improvements over the ADC model in individual risk reclassification (continuous NRI = 0.65, 95% CI 0.33-0.95), overall predictive accuracy (IDI = 0.07, 95% CI excluding 0), and calibration (Brier score: 0.16 vs. 0.17; MAE: 0.01 vs. 0.04; MSE: 2.3×10⁻⁴ vs. 1.91×10⁻³).

Basic summary

MAGMA. 2026 Jun 6. doi: 10.1007/s10334-026-01374-3. Online ahead of print.

ABSTRACT

OBJECTIVE: To evaluate intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) for preoperative diagnosis of perineural invasion (PNI) in rectal cancer (RC).

MATERIALS AND METHODS: A total of 148 patients with pathology-confirmed RC (PNI+, n = 72; PNI-, n = 76) were enrolled. Parameters from mono-exponential (ADC), bi-exponential (D, D*, f), and stretched-exponential (DDC, α) IVIM models were analyzed. Univariate and multivariate logistic regression analyses were used to construct diagnostic models. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curve analysis. The DeLong test was used to compare the AUC of the models. Internal validation was employed to assess model performance. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI), along with calibration metrics and decision curve analysis, were used to further evaluate model performance. P-value < 0.05 was considered statistically significant.

RESULTS: ADC, D, f, and DDC differed significantly between groups. Multivariate analysis identified ADC and D as independent PNI predictors. The D value yielded the highest AUC (0.84), while ADC showed the highest sensitivity (81.94%). A combined model (ADC + D) achieved an AUC of 0.85, sensitivity of 86.10%, specificity of 73.70%, and accuracy of 77.00%. The fivefold internal validation mean AUC was 0.84 ± 0.04. No significant AUC differences were found among parameters or models (DeLong test, P > 0.05). Further analyses revealed that the combined model provided significant improvements over the ADC model in individual risk reclassification (continuous NRI = 0.65, 95% CI 0.33-0.95), overall predictive accuracy (IDI = 0.07, 95% CI excluding 0), and calibration (Brier score: 0.16 vs. 0.17; MAE: 0.01 vs. 0.04; MSE: 2.3×10⁻⁴ vs. 1.91×10⁻³). Decision curve analysis demonstrated consistently higher net benefit for the combined model across threshold probabilities of 0-0.50.

CONCLUSION: IVIM-DWI demonstrates potential value for the preoperative assessment of PNI status in rectal cancer and may facilitate individualized treatment planning.

PMID:42250045 | DOI:10.1007/s10334-026-01374-3

Document this CPD