Published online Apr 28, 2024. doi: 10.3748/wjg.v30.i16.2233
This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial.
Peer-review started: January 2, 2024
First decision: January 31, 2024
Revised: February 8, 2024
Accepted: March 20, 2024
Article in press: March 20, 2024
Published online: April 28, 2024
Processing time: 114 Days and 17.6 Hours
Perineural invasion (PNI), is a potential pathway for the metastatic spread of rectal cancer (RC), and has been used as an important pathological indicator and independent prognostic factor. Preoperative stratification of RC patients according to PNI status facilitates individualized treatment and improves the prognosis of RC patients.
Nowadays, the preoperative predicton of PNI status is still challenging and needs further study.
To evaluate the usefulness of a model based on preoperative magnetic resonance imaging (MRI) radiomics for predicting PNI status in patients with RC and establishing and validating an optimal nomogram model for predicting PNI status preoperatively in RC patients.
We enrolled 244 RC patients from two independent centers from May 2019 to August 2022. The patients from Center 1 were randomly divided into a training group (n = 118) and an internal validation group (n = 52), whereas 74 patients from Center 2 served as an external validation group. Extracted and selected quantitative radiomics features and clinical risk factors to establish and validate the radiomics predictive model and clinical-radiomics (CR) model.
We extracted 944 radiomics features from T2-weighted imaging and contrast-enhanced T1-weighted imaging sequences, combined with PNI-related clinical features (clinical TNM and histological grade) to construct the final CR model, and used internal and external validation groups to evaluate the models. The final CR model showed good performance to predict PNI status, the area under the curve of the CR model in the training and internal and external validation groups were 0.889, 0.889 and 0.894, respectively.
The CR model based on MRI radiomics features and clinical risk factors was able to predict the PNI status of RC noninvasively, showed stable performance, which can provide support for individualized treatment of RC patients.
Further external verification is needed to optimize the model, and explore the feasibility of applying deep learning to automatically describe volume of interest, reduce the difference between observers, and improve the applicability of the model.