Machine Learning Achieves Pathologist-Level Celiac Disease Diagnosis
Publication New England Journal of Medicine
Florian Jaeckle, James Denholm, Benjamin Schreiber, Shelley C. Evans, Mike N. Wicks, James Y. H. Chan, Adrian C. Bateman, Sonali Natu, Mark J. Arends, Elizabeth Soilleux
27 March 2025
Diagnosing coeliac disease (CD), an autoimmune disorder that has an estimated global prevalence of around 1%, mainly relies on examination of biopsies of the duodenum under a microscope – called histologic examination. However when pathologists look at the same sample, only 80% of the time do they agree on a CD diagnosis. Jaeckle et al aimed to improve the CD diagnosis by developing an accurate, machine-learning-based diagnostic classifier. Initial results revealed that the model was able to diagnose samples with an accuracy, sensitivity and specificity exceeding 95%. It was concluded that the tool may be able to assist pathologists in making a CD diagnosis, reducing the time required to make a diagnosis.
