JMIR Publications

ABSTRACT

Background:

The undifferentiated-type of early gastric cancer (U-EGC) is included among the expanded indications of endoscopic submucosal dissection (ESD), yet the rate of curative resection remains unsatisfactory. Endoscopists predict the probability of curative resection using the size and shape of the lesion and whether ulcers are present or not. The location of the lesion, indicating the likely technical difficulty, is also considered.

Objective:

To establish machine learning (ML) models to better predict the possibility of curative resection in U-EGC prior to ESD.

Methods:

A nationwide cohort of 2,703 U-EGCs treated by ESD or surgery were adopted for the training and internal-validation cohorts. Separately, an independent dataset of the Korean ESD registry (n=275) and an Asan medical center dataset (n=127) treated by ESD were chosen for external-validation. Eighteen ML classifiers were selected to establish prediction models of curative resection with the following variables; age; sex; location, size, and shape of the lesion; and whether ulcers were present or not.

Results:

Among the 18 models, the XGBoost classifier showed the best performance [internal-validation accuracy: 93.4% (95% confidence interval: 90.4%–96.4%), precision: 92.6% (89.5%–95.7%), recall: 99.0% (97.8%–99.9%), and F1 score: 95.7% (93.3%–98.1%)]. Attempts at external-validation showed substantial accuracy [81.5% (76.9%–86.1%) (first external-validation) and 89.8% (84.5%–95.1%) (second external-validation), respectively]. Lesion size was the most important feature per explainable artificial intelligence analysis.

Conclusions:

We established an ML model capable of accurately predicting the curative resection of U-EGC before ESD by incorporating lesion morphologic and ecologic factors.