Alimentary pharmacology & therapeutics · 2025
Systematic Review: Use of Artificial Intelligence and Unmet Needs in Eosinophilic Oesophagitis.
СтатьяСистематический обзорПоддержка исследования: вне правительства СШАНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Castagnaro Ester, Felici Enrico, Spaccapelo Roberta, Amoroso Mariangela, Moretti Devis, Saab Jean Pierre, Stassaldi Annalisa, Tassone Andrea, Borrelli Osvaldo, De Bortoli Nicola, Corso Giulio Del, Gaynor Edward, Goh Leanne, Oliva Salvatore, Rea Francesca, Renzo Sara, Romano Claudio, Savarino Edoardo V, Visaggi Pierfrancesco
Аннотация
BACKGROUND: Artificial intelligence (AI) has been applied extensively to eosinophilic oesophagitis (EoE), but its clinical impact and comprehensiveness remain unclear.
AIMS: To summarise the state of the art of applications of AI in EoE, identify gaps in the literature and disclose unmet needs.
METHODS: We performed a systematic review of applications of AI in diagnosis, prognosis, precision medicine and follow-up in EoE. We searched MEDLINE, Embase and Embase Classic (via Ovid) from inception until 28th February 2025. Clinical trials, cohort, case-control, cross-sectional and case series studies were eligible for inclusion.
RESULTS: We identified 28 studies on AI in EoE, with 18 focusing on diagnosis. AI was used for histology image analysis in digital pathology in seven and endoscopic image analysis in three studies. AI-based non-invasive diagnostic predictive models showed high accuracy in detecting EoE. As a prognostic tool, AI showed potential in predicting food impaction, histologic disease severity and fibrosis. AI was used in precision medicine for biomarker discovery, endotype prediction based on transcriptomics and treatment personalisation, as well as in disease monitoring and follow-up. Unmet needs included insufficient studies on children/adolescents, prediction of treatment response and improvement of study designs.
CONCLUSIONS: Although AI showed promise at improving EoE management across various domains, most tools require validation on larger, independent and prospective cohorts. Several challenges, including algorithmic bias, data privacy, explainability and regulatory hurdles, need to be addressed for clinical implementation. Future studies should focus on developing non-invasive diagnostic tools for younger populations, treatment response prediction and disease monitoring.
AIMS: To summarise the state of the art of applications of AI in EoE, identify gaps in the literature and disclose unmet needs.
METHODS: We performed a systematic review of applications of AI in diagnosis, prognosis, precision medicine and follow-up in EoE. We searched MEDLINE, Embase and Embase Classic (via Ovid) from inception until 28th February 2025. Clinical trials, cohort, case-control, cross-sectional and case series studies were eligible for inclusion.
RESULTS: We identified 28 studies on AI in EoE, with 18 focusing on diagnosis. AI was used for histology image analysis in digital pathology in seven and endoscopic image analysis in three studies. AI-based non-invasive diagnostic predictive models showed high accuracy in detecting EoE. As a prognostic tool, AI showed potential in predicting food impaction, histologic disease severity and fibrosis. AI was used in precision medicine for biomarker discovery, endotype prediction based on transcriptomics and treatment personalisation, as well as in disease monitoring and follow-up. Unmet needs included insufficient studies on children/adolescents, prediction of treatment response and improvement of study designs.
CONCLUSIONS: Although AI showed promise at improving EoE management across various domains, most tools require validation on larger, independent and prospective cohorts. Several challenges, including algorithmic bias, data privacy, explainability and regulatory hurdles, need to be addressed for clinical implementation. Future studies should focus on developing non-invasive diagnostic tools for younger populations, treatment response prediction and disease monitoring.