Otolaryngology--head and neck surgery : official journal of American Academy of Otolaryngology-Head and Neck Surgery · 2025
Toward an Unbiased Deep Learning Classifier of Pediatric Middle Ear Disease.
СтатьяМногоцентровое исследованиеНазвание и аннотация: Перевод пока не подготовлен
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.
Surapaneni Sruthi, Rangarajan Nikhil, Davis Kyle, Pletcher Katherine, Flowers Jeffrey, Strub Graham, Nolder Abby, King Deanne, Marston Alexander P, Vecchiotti Mark, Alfonso Kristan, Evans Sean, Deshpande Anita, Prickett Kara, Landry April, Goudy Steven, Govil Nandini, Messner Anna, Richter Gresham, Scott Andrew R, Mehta Deepak, Hill Courtney Anne
Аннотация
OBJECTIVE: Otitis media is the leading cause of healthcare visits and antibiotic prescriptions for children in the United States. Differentiating acute otitis media (AOM) from otitis media with effusion (OME) is crucial for antibiotic stewardship but is often difficult. The objective was to train an artificial intelligence algorithm that accurately predicts the presence and nature of middle ear effusion in pediatric patients using pediatric tympanic membrane (TM) images captured with inexpensive, consumer-grade otoscopes.
STUDY DESIGN: Prospective cohort study.
SETTING: Tertiary Children's Hospitals.
METHODS: A multicenter study gathered ear images from children aged 6 months to 10 years undergoing myringotomy and tube placement at four pediatric hospitals in the United States. Images were taken with over-the-counter digital otoscopes. Intraoperative middle ear findings were used to label the images. A deep learning algorithm was trained to classify middle ear disease. Performance was assessed by weighted accuracy.
RESULTS: From a diverse population of 219 children (42.14% black, Hispanic, Asian, and other), 737 images were obtained, categorized as AOM (73), OME (190), no effusion or infection (274), and no TM in image (200). The classification model achieved a weighted accuracy of 92.5%, ranging 88.4% to 98.8% per individual category.
CONCLUSION: The model demonstrated high accuracy in classifying the middle ear state in young, anesthetized children. Developing an effective deep learning model using diverse, age-representative images from affordable digital otoscopes may move us closer to real-world applications of such technology in clinical practice to validate the role of telemedicine and improve antibiotic stewardship.
STUDY DESIGN: Prospective cohort study.
SETTING: Tertiary Children's Hospitals.
METHODS: A multicenter study gathered ear images from children aged 6 months to 10 years undergoing myringotomy and tube placement at four pediatric hospitals in the United States. Images were taken with over-the-counter digital otoscopes. Intraoperative middle ear findings were used to label the images. A deep learning algorithm was trained to classify middle ear disease. Performance was assessed by weighted accuracy.
RESULTS: From a diverse population of 219 children (42.14% black, Hispanic, Asian, and other), 737 images were obtained, categorized as AOM (73), OME (190), no effusion or infection (274), and no TM in image (200). The classification model achieved a weighted accuracy of 92.5%, ranging 88.4% to 98.8% per individual category.
CONCLUSION: The model demonstrated high accuracy in classifying the middle ear state in young, anesthetized children. Developing an effective deep learning model using diverse, age-representative images from affordable digital otoscopes may move us closer to real-world applications of such technology in clinical practice to validate the role of telemedicine and improve antibiotic stewardship.