A Study on Gastric Tract Infections Detection and Classification from Wireless Capsule Endoscopy (WCE) Imaging

Due to the unavoidable inherent difficulties in accessing a volute position in the human body, gastrointestinal tract (GI) disorders such as abdominal bleeding, ulcers, tumors, celiac disease, Crohn's disease, and others are difficult to detect. With Wireless Capsule Endoscopy (WCE), the GI tract can be examined in a comfortable, non-invasive, and painless manner. While utilizing DL methods to automatically recognize anomalies in WCE images enhances detection accuracy, model training requires a sizable number of labeled data. However, the explainability of these deep methods is poor, and they do not take into account expert knowledge when making decisions. This study provides an analysis of the performance, complexity, and dataset quality of various DL approaches for anomaly localization and detection methods used in WCE images. In this study, the effectiveness of wireless capsule endoscopy techniques documented in the literature within the previous five years is examined (2019–2022). There is also a comparison table of the works discussed. However, among the suggested possibilities, the use of WCE images has received more attention recently as a useful and distinctive choice.