Health Information Technology and Celiac Disease Diagnosis: A Scoping Review of Methods and Applications

In recent years, health information technology methods have been increasingly used to diagnose celiac disease (CD). In order to improve diagnosis, image analysis, deep learning, and machine learning methods have been used to improve the accuracy of diagnosis in the shortest time while requiring the least human resources. This article summarizes these techniques and discusses the latest advances in the diagnosis of CD.

All articles that use patient data for diagnosis are presented in this article, and systems designed for follow-up were excluded. EMBASE, PubMed, Web of Science, and Scopus databases have been searched in this inquiry. After searching different databases to eliminate duplicate and review articles, 2266 articles were evaluated in the first stage, and then the titles, abstracts, and concepts were reviewed, and 33 articles were finally selected.

The results showed that 14 studies were conducted in children (42.42%), eight studies were conducted in adults (27.58%), and 10 studies were conducted in two populations (roughly 30.30%). 14 studies were completed in the United States, 16 studies were completed in Europe, and two studies were completed in Asia. 12 articles (37.93%) used the integration of image processing and artificial intelligence methods. 23 articles (69.69%) used endoscopy images, and nine articles (27.27%) used patient information in their research.

The results express a general acceptance of using information technology to improve the diagnosis of CD, but further research is needed to improve and optimize the method of diagnosis. Recent artificial intelligence techniques using deep learning (DL), machine learning methods such as convolutional neural network, support vector machines, and K-nearest neighbors, and also image processing have emerged as the breakthrough computer technology that can be used for computer-aided diagnosis of CD. The current review summarizes methods used in clinical studies to diagnose CD, from feature extraction methods to artificial intelligence and machine learning techniques.