Journal of pediatric surgery · 2026
Artificial intelligence in the diagnosis of Hirschsprung disease: A scoping review and rationale for a multicentric approach.
СтатьяScoping ReviewНазвание 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.
Alzate-Ricaurte Sergio, Ocampo Osorio Felipe, Costa-Roig Adrià, Gómez Ayala Maria Camila, Gomez Inés Elvira, Arnold Michael A, Campos Garcia Clara Ivette, Sanchez-Sosa Sergio, Lovell Mark, Cruz-Suarez Gustavo Adolfo, Portilla-Buenaventura Ana Maria, Alzate Gallego Edgar Dario, De La Torre Luis
Аннотация
BACKGROUND: Hirschsprung disease (HD) is a congenital disorder characterized by absence of enteric ganglion cells, leading to functional bowel obstruction. Despite advances in diagnosis, global disparities persist, particularly in low- and middle-income countries (LMICs), where access to pathology remains limited. Artificial intelligence (AI) offers potential to improve diagnostic accessibility through histopathologic analysis.
OBJECTIVE: To systematically synthesize evidence on AI-based diagnostic strategies for HD and identify current gaps to guide future multicenter research.
METHODS: Following PRISMA-ScR guidelines, PubMed, EMBASE, Epistemonikos, and CENTRAL were searched through June 2025. Studies applying AI to HD diagnosis in patients aged 0-18 years were included. Data on model design, validation, performance, and interpretability were extracted and narratively synthesized.
RESULTS: Six retrospective studies (2018-2024) met inclusion criteria. Four applied machine learning and two deep learning. Input data included histologic slides (n = 4) and clinical or radiologic features (n = 2). Reported accuracies ranged from 42 % to 100 %, recall 81.8 %-100 %, and AUC 0.82-0.99. All studies were internally validated and characterized by heterogeneous designs, small datasets, and no use of explainable AI (XAI) frameworks. Reporting of computational costs, model weight, or latency was absent, limiting evaluation of real-world feasibility.
CONCLUSIONS: First review to systematically synthesize data on the topic. AI can identify histologic and imaging features relevant to HD diagnosis, but current models remain exploratory. Limited dataset diversity, lack of validation, and absence of explainability limit clinical translation. Future work should prioritize multicenter collaboration, diverse data inclusion from LMICs, and simple, interpretable architectures to enable equitable, real-world diagnostic application.
OBJECTIVE: To systematically synthesize evidence on AI-based diagnostic strategies for HD and identify current gaps to guide future multicenter research.
METHODS: Following PRISMA-ScR guidelines, PubMed, EMBASE, Epistemonikos, and CENTRAL were searched through June 2025. Studies applying AI to HD diagnosis in patients aged 0-18 years were included. Data on model design, validation, performance, and interpretability were extracted and narratively synthesized.
RESULTS: Six retrospective studies (2018-2024) met inclusion criteria. Four applied machine learning and two deep learning. Input data included histologic slides (n = 4) and clinical or radiologic features (n = 2). Reported accuracies ranged from 42 % to 100 %, recall 81.8 %-100 %, and AUC 0.82-0.99. All studies were internally validated and characterized by heterogeneous designs, small datasets, and no use of explainable AI (XAI) frameworks. Reporting of computational costs, model weight, or latency was absent, limiting evaluation of real-world feasibility.
CONCLUSIONS: First review to systematically synthesize data on the topic. AI can identify histologic and imaging features relevant to HD diagnosis, but current models remain exploratory. Limited dataset diversity, lack of validation, and absence of explainability limit clinical translation. Future work should prioritize multicenter collaboration, diverse data inclusion from LMICs, and simple, interpretable architectures to enable equitable, real-world diagnostic application.