Deep learning for automated bowel preparation assessment during colonoscopy: time to embrace a new approach?
Menée en Chine à partir d'images coloscopiques, d'images vidéo et de données portant sur 616 patients ayant bénéficié d'une coloscopie de dépistage entre mai et août 2020 (âge : entre 18 et 75 ans), cette étude évalue la performance, du point de vue du taux de détection d'adénomes colorectaux, d'un système d'évaluation de la préparation colique reposant sur un algorithme d'apprentissage profond
Some of the most translationally mature artificial intelligence (AI) applications in health care belong to colonoscopy, where algorithms can now be used in clinical practice to assist colorectal polyp detection and characterisation. The rapid pace of translation reflects the desperate need to identify solutions to drive up quality in colonoscopy and overcome the operator dependence associated with variable colorectal cancer protection. It is not surprising that AI developers are expanding research to further use cases applied to colonoscopy, particularly those that relate to performance measures and quality metrics.
Suboptimal bowel preparation is a major barrier to effective colonoscopy as it is associated with missed polyps, incomplete procedures, unsatisfactory patient experience, shorter surveillance intervals, and increased health-care costs.
The Lancet Digital Health , commentaire en libre accès, 2020