**3.4 AI in EoE diagnostics**

The applications of AI in EoE have been on rise. One of the most recent applications by Guimarães et al. is the utilization of CNN networks in endoscopic images of EoE. Their study examined 484 real-world endoscopic images taken from 134 subjects within three distinct categories (normal, EoE, and candidiasis). In their results, they found that global accuracy (0.915 [95% confidence interval (CI) 0.880–0.940]), specificity (0.936 [95%CI 0.910–0.955]), and sensitivity (0.871 [95%CI 0.819–0.910) were all higher than for the endoscopists on the test set. The global area under the receiver operating characteristic curve was 0.966 [95%CI 0.954–0.975] [82]. One study by Dnaiel et al. applied machine learning to EoE biopsies and created a dataset for training a multilabel segmentation deep network. Their model was able to segment intact and notintact eosinophils with a mean intersection over union (mIoU) value of 0.93. This segmentation was able to quantify intact eosinophils with a mean absolute error of 0.611 eosinophils and to classify EoE disease activity with an accuracy of 98.5%. Their model achieved 94.8% accuracy, 94.3% sensitivity, and 95.14% specificity in detecting EoE disease activity when using whole slide images from the validation cohort [83]. EoE diagnosis could be flourished with the introduction of AI as more already ongoing research on it in the literature [84–87].
