Endorectal Ultrasound-Based Habitat Analysis for Pathologic Response Prediction After Neoadjuvant Chemotherapy in Locally Advanced Rectal Cancer
Related Experiment Videos
Hang Yi1, Limei Chen1, Xiaoyin Liu1
1Department of Medical Ultrasonics, The Sixth Affiliated Hospital, Sun Yat-sen University, Guangzhou, China; Guangdong Provincial Key Laboratory of Colorectal and Pelvic Floor Diseases, The Sixth Affiliated Hospital, Sun Yat-Sen University, Guangzhou, China; Biomedical Innovation Center, The Sixth Affiliated Hospital, Sun Yat-Sen University, Guangzhou, China.
Objective:
To investigate the value of endorectal ultrasound (ERUS)-based habitat analysis in predicting pathologic response to neoadjuvant chemotherapy (NCT) in locally advanced rectal cancer (LARC).
Methods:
Consecutive patients treated between January 2018 and May 2025 were randomly divided into training and test sets at a 7:3 ratio. Tumor habitat subregions were defined using K-means clustering, and radiomic features were extracted and selected. Clinical, habitat, and fusion models were developed, and model performance was evaluated using ROC analysis, calibration curves, and decision curve analysis (DCA).
Results:
Among 407 patients, tumor location, cT stage, extramural vascular invasion, and tumor-stroma ratio differed significantly between good- and poor-response groups (p < 0.05). For predicting response to NCT, the clinical, habitat, and fusion models achieved AUCs of 0.885 (95% CI: 0.840-0.927), 0.911 (95% CI: 0.872-0.943), and 0.958 (95% CI: 0.935-0.979) in the training set and 0.757 (95% CI: 0.668-0.841), 0.720 (95% CI: 0.599-0.828), and 0.829 (95% CI: 0.737-0.906) in the test set, respectively. In the test set, the fusion model outperformed the habitat model (p = 0.010), whereas the difference from the clinical model was not statistically significant (p = 0.097). The fusion model exhibited the highest sensitivity in both sets (0.956 and 0.897, respectively). In the test set, it yielded the lowest Brier score (0.174) and generally provided the greatest net benefit across most threshold probabilities of 0.12 to 0.75.
Conclusion:
The fusion model demonstrated promising performance for predicting response to NCT in LARC. Prospective multicenter external validation is required before clinical implementation.