Purpose: This study aimed to explore a combined transrectal ultrasound (TRUS) and radiomics model for predicting tumor regression grade (TRG) after neoadjuvant chemoradiotherapy (NCRT) in patients with locally advanced rectal cancer (LARC).
Methods: Among 190 patients with LARC, 53 belonged to GRG and 137 to PRG. Eight TRUS parameters were identified as statistically significant (P < 0.05) for distinguishing between the groups, including PSVpre, LDpost, TDpost, CEUS-IGpost, LD change rate, TD change rate, RI change rate, and CEUS-IG downgrade. The accuracies of these individual parameters in predicting TRG were 0.42, 0.62, 0.56, 0.68, 0.67, 0.70, 0.63, and 0.71, respectively. The AUC values were 0.596, 0.597, 0.630, 0.752, 0.686, 0.660, 0.650, and 0.666, respectively. The multi-parameter ultrasonic logistic regression (MPU-LR) model achieved an accuracy of 0.816 and an AUC of 0.851 (95% CI: [0.792-0.909]). The optimal pre- and post-treatment radiomics models were RF (Mean-PCA-RFE-6) and AE (Zscore-PCA-RFE-12), with accuracies of 0.563 and 0.596 and AUCs of 0.601 (95% CI: [0.561-0.641]) and 0.662 (95% CI: [0.630-0.694]), respectively. The combined model (US-RADpre-RADpost) showed the highest predictive power with accuracy and AUC of 0.863 and 0.913.
Conclusions: The combined model based on TRUS and radiomics demonstrated remarkable predictive capability for TRG after NCRT. It serves as a precision tool for assessing NCRT response in patients with LARC, impacting treatment strategies.
Keywords: Locally advanced rectal cancer (LARC); Neoadjuvant chemoradiotherapy (NCRT); Radiomics; Transrectal ultrasound (TRUS); Tumor regression grade (TRG).
© 2025. The Author(s).