Development and Verification of a Deep Learning Algorithm to Evaluate Small-Bowel Preparation Quality

Abstract

Capsule endoscopy (CE) quality control requires an objective scoring system to evaluate the preparation of the small bowel (SB). We propose a deep learning algorithm to calculate SB cleansing scores and verify the algorithm’s performance. A 5-point scoring system based on clarity of mucosal visualization was used to develop the deep learning algorithm (400,000 frames; 280,000 for training and 120,000 for testing). External validation was performed using additional CE cases (n = 50), and average cleansing scores (1.0 to 5.0) calculated using the algorithm were compared to clinical grades (A to C) assigned by clinicians. Test results obtained using 120,000 frames exhibited 93% accuracy. The separate CE case exhibited substantial agreement between the deep learning algorithm scores and clinicians’ assessments (Cohen’s kappa: 0.672). In the external validation, the cleansing score decreased with worsening clinical grade (scores of 3.9, 3.2, and 2.5 for grades A, B, and C, respectively, p < 0.001). Receiver operating characteristic curve analysis revealed that a cleansing score cut-off of 2.95 indicated clinically adequate preparation. This algorithm provides an objective and automated cleansing score for evaluating SB preparation for CE. The results of this study will serve as clinical evidence supporting the practical use of deep learning algorithms for evaluating SB preparation quality.

Small bowel cleanliness in capsule endoscopy: a case–control study using validated artificial intelligence algorithm

“…However, since small bowel cleanliness is subjective and takes a long time to be measured by clinicians, intra-observer variation is inevitable 14 . For this reason, several studies have been conducted to identify the cleanliness of SBCE using an AI algorithm objectively 12 , 15 – 17 . We confirmed the objective SB bowel cleanliness scores by using a validated AI algorithm that was trained using SBCE images 12 .…”

Diagnostic Yields and Clinical Impacts of Capsule Endoscopy

“…Furthermore, CE is considered a therapeutic endoscope for biopsies, hemostasis, and drug delivery [117]. In addition, using computer-assisted diagnosis and artificial intelligence, CE reading could be achieved, reducing reading time and improving diagnosis [118,119]. With the development of these technologies, it is expected to be used in clinical applications as actual products in the near future.…”

Where do we stand in AI for endoscopic image analysis? Deciphering gaps and future directions

“…The test results on 927 images from the external dataset showed an overall accuracy of 96.7%. Similarly, a study focused on understanding the percentage of mucosal visualisation in small bowel during VCE used a simple, fully connected convolution neural network 81 . Similarly, most landmark classification works only applied off-the-shelf CNN networks showing good accuracy in the classification of the landmark sites (e.g., above 90% recall values for 9 out of 11 site classes 82 ), widely based on the OGD procedures that include the oesophagus, stomach and duodenum 82 , 83 .…”