Deep Learning and Minimally Invasive Endoscopy: Automatic Classification of Pleomorphic Gastric Lesions in Capsule Endoscopy

Abstract

Introduction – Capsule endoscopy (CE) is a minimally invasive exam for evaluating the gastrointestinal tract. However, it’s diagnostic yield for detecting gastric lesions is suboptimal. Convolutional Neural Networks (CNN) are artificial intelligence models with great performance for image analysis. Nonetheless, their role in gastric evaluation by wireless CE (WCE) has not been explored. Methods – Our group developed a CNN-based algorithm for the automatic classification of pleomorphic gastric lesions, including vascular lesions (angiectasia, varices and red spots), protruding lesions, ulcers and erosions. 12918 gastric images from three different CE devices (PillCam Crohn’s; PillCam SB3; OMOM HD capsule endoscopy system) were used from the construction of the CNN: 1407 from protruding lesions; 994 from ulcers and erosions; 822 from vascular lesions and 2851 from hematic residues, the remaining images from normal mucosa. The images were divided in a training (split for 3-fold cross validation) and validation dataset. The model’s output was compared to a consensus classification by two WCE experienced gastroenterologists. The networks’ performance was evaluated by its sensitivity, specificity, accuracy, positive predictive (PPV) and negative predictive values (NPV), and area under the precision-recall curve (AUPRC). Results – The trained CNN had a 97.4% sensitivity, 95.9% specificity, PPV and NPV of 95.0% and 97.8% for gastric lesions, with 96.6% overall accuracy. The CNN had an image processing time of 115 images per second. Conclusions – Our group developed, for the first time, a CNN capable of automatically detecting pleomorphic gastric lesions in both small bowel and colon capsule endoscopy devices.

More than 20 procedures are necessary to learn small bowel capsule endoscopy: Learning curve pilot study of 535 trainee cases

“…This can be seen as a sign of missing dedication and prioritization during a busy workday, too high reading speed, or a sign of confidence without achieving enough competencies, which is alarming 24 . The rapidly developing use of artificial intelligence (AI) with promising diagnostic accuracy can potentially assist learning and change the reviewing process toward deep learning algorithms instead of in-person evaluation 25 . The use of AI to assist in learning SBCE calls for further studies to ensure sufficient learning and diagnostic accuracy by the reviewers.…”

Artificial intelligence in capsule endoscopy: development status and future expectations

“…The severity assessment given (normal, mild, and severe) by the model was shown to correlate well with those manually assigned by experts. While multiple machine learning models have achieved reasonable sensitivities and specificities in this field [79][80][81] , since 2018, deep learning systems have predominated research [81][82][83][84][85][86][87][88][89][90][91][92] . In 2022, Ferreira et al developed a CNN using a total of 8,085 images to detect ulcers and erosions in images from the PillCam™ Crohn's Capsule, with an overall sensitivity of 90% and specificity of 96% [89] .…”

The Use of Artificial Intelligence in Gastroenterology: A Glimpse Into the Present

“…AI can also improve patient access to GI diagnostics, as some technologies may become more easily utilized by primary care providers or allied health professionals. To that end, we present data on AI's ability to diagnose gastric lesions using only capsule endoscopy ( 4 ) and as a tool to render a diagnosis during anorectal manometry ( 5 ).…”