Clinical usefulness of AI-assisted small bowel localization and lesion detection in capsule endoscopy
Endoscopy · 01.04.2024 · 56 (S 02 S 02) · S121
DOI: 10.1055/s-0044-1782956 ↗Kurz gefasst
Fachartikel in der Zeitschrift Endoscopy, erschienen am 01.04.2024. Verfasst von Y. J. Yang und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Aims
Although various AI models have been developed to detect small bowel (SB) abnormalities in capsule endoscopy (CE), AI models that simultaneously enable the localization of SB and detection of SB abnormalities were rarely developed. This study aimed to develop an AI model that automatically discriminated SB from stomach and colon, diagnosed multiple abnormalities in CE, and compared its performance to that of experienced endoscopists.
Methods
We developed a CNN model using 87,005 CE images (11,925 stomach, 33,781 SB, and 41,299 colon) for localization and 28,405 CE images (1,337 erosion/ulcer, 126 angiodysplasia, 494 bleeding, and 26,448 normal and other images) for detection of SB abnormalities. The whole dataset was divided into training, validation, and test sets of 7:1:2 for localization and 8:1:1 for detection of SB abnormalities. To evaluate the clinical usefulness of the CNN model, we compared the performance (reading time, the accuracy of localization, and lesion detection in per-patient analysis) of conventional reading (endoscopist-alone reading) to that of AI-assisted reading, which was performed using the summary version of CE video made by newly developed CNN model.
Results
For the internal test dataset, the sensitivity and specificity to discriminate stomach, SB, and colon were 98.7% and 99.6%, 96.1% and 97.9%, and 97.1% and 97.8%, respectively. And the accuracy for detection of erosion/ulcer, angiodysplasia, bleeding, and normal and others was 88%, 100%,100%, and 100%, respectively. Comparing the performance of AI-assisted reading and conventional reading of 22 external validation datasets, the reading time was significantly reduced in AI-assisted reading than conventional reading (53.9 min vs. 8.7 min, P
Zitieren
APA
Yang, Y. J., Cho, B. J., & Jang, H. J. (2024). Clinical usefulness of AI-assisted small bowel localization and lesion detection in capsule endoscopy. Endoscopy, 56(S 02 S 02), S121. https://doi.org/10.1055/s-0044-1782956
Vancouver
Yang YJ, Cho BJ, Jang HJ. Clinical usefulness of AI-assisted small bowel localization and lesion detection in capsule endoscopy. Endoscopy. 2024;56(S 02 S 02):S121. doi:10.1055/s-0044-1782956
BibTeX
@article{Yang2024,
author = {Yang, Y. J. and Cho, B. J. and Jang, H. J.},
title = {Clinical usefulness of AI-assisted small bowel localization and lesion detection in capsule endoscopy},
journal = {Endoscopy},
year = {2024},
volume = {56},
number = {S 02 S 02},
pages = {S121},
doi = {10.1055/s-0044-1782956}
}
EUVIMED