EoE App: Deep-learning based detection of eosinophilic esophagitis

Despite a growing awareness for eosinophilic esophagitis (EoE), diagnostic delay is still a clinically relevant phenomenon. Deep learning (DL) based algorithms have demonstrated potential in medical image analysis. Our convolutional neural network was trained to analyze endoscopic images and distinguish EoE from normal findings and candida esophagitis using real-world endoscopic still images.

We trained and tested our approach using 484 real-world images (e.g. various scope positions, distances, angles, illumination, food and mucus contaminations). Results were highly reproducible with a global accuracy of 0.911, performing significantly better than the combined results of 3 independent endoscopists. Of note, the algorithm independently identified typical endoscopic signs also used for classification by endoscopists.

With this web tool we want to provide wide access to this software. It allows you to upload your own endoscopic images for analysis, and provides an indicative diagnosis.