Photo Credit: ALIOUI Mohammed Elamine
The following is a summary of “Histological Image-based Ensemble Model to Identify Myenteric Plexitis and Predict Endoscopic Postoperative Recurrence in Crohn’s Disease: A Multicentre, Retrospective Study,” published in the November 2023 issue of Gastroenterology by Wang et al.
Researchers conducted a retrospective study to develop and validate a deep learning system for automated detection of myenteric plexitis features, aiming to improve the prediction of postoperative recurrence in Crohn’s disease (CD).
They analyzed 205 patients who had undergone bowel resection surgery. The patients were categorized into a training group [n = 108], an internal validation group [n = 47], and an external validation group [n = 50]. Using the ResNet50 encoder, 190,960 patches were examined from 278 whole-slide images of surgical specimens, extracting 6,144 features through transfer learning. Five robust algorithms were employed to develop classification models. The models were evaluated using AUC in three cohorts.
The results showed that the stacking model exhibited satisfactory accuracy in predicting the recurrence of CD in the training cohort (AUC: 0.980; 95% CI 0.960–0.999), external validation cohort (AUC: 0.868; 95% CI 0.761–0.975), and internal validation cohort (AUC: 0.908; 95% CI 0.823–0.992). The accuracy for discerning the severity of myenteric plexitis was 0.833, 0.745, and 0.694 in the training, internal validation, and external validation cohorts, respectively.
Investigators concluded that a stacking model based on extracted muscular and myenteric plexus features successfully predicted postoperative recurrence of CD.
Source: academic.oup.com/ecco-jcc/advance-article-abstract/doi/10.1093/ecco-jcc/jjad196/7450131