Journal of dentistry · 2026
External validation of a mHealth tool for detecting gingival inflammation in community-dwelling older adults.
СтатьяВалидационное 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.
Chau Reinhard Chun Wang, Gong Zhuohong, Cheng Andrew Chi Chung, Mao Kaijing, Thu Khaing Myat, Zhang Ruonan, Ling Zhaoting, Chang Tien Hsin, Tan Hong Jin, Tsang Harris Sik-Ho, Tonetti Maurizio S, Schwendicke Falk, Hsung Richard Tai-Chiu, Lam Walter Yu Hang
Аннотация
OBJECTIVE: This study aimed to externally validate the screening performance of a pro-social (equitable, accessible), explainable AI (XAI)-guided mobile health (mHealth) tool, GumAI, that automated the analysis of smartphone photographs to detect signs of gingival inflammation in community-dwelling older adults, a condition that is common yet often undiagnosed due to limited access to dental care.
MATERIAL AND METHODS: Older adults (age 60+) were recruited from nine community centres. Frontal intraoral photographs were captured with smartphones and evaluated using an mHealth tool that classified gingival regions as "no inflammation" (health) in green, "mild changes/questionable" in yellow, or "inflamed" (diseased) in red at the pixel level. In this study, the screening performance was validated only for visible anterior teeth. Two calibrated periodontists independently assessed these photographs, annotated the regions of healthy/mild changes/questionable/diseased, and served as the benchmark for comparison. Performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, F1-score, and mean intersection over union (mIOU), were calculated by comparing the AI outputs to the benchmark.
RESULTS: Of 262 invited older adults, 163 (62.21%) participated, and each centre recruited 17-44 participants. Inter-rater agreement between periodontists was substantial (Prevalence- and Bias-Adjusted Kappa (PABAK) = 0.62). GumAI exhibited sensitivity of 0.91 (95% CI 0.91-0.92), specificity of 0.87 (95% CI 0.87-0.87), PPV of 0.90 (95% CI 0.89-0.90), NPV of 0.89 (95% CI 0.89-0.89), accuracy of 0.89 (95% CI 0.89-0.89), F1-score of 0.90, and mIOU of 0.83.
CONCLUSION: The mHealth tool demonstrated potential to detect visual signs of gingival inflammation in intraoral photographs of older adults in a community setting, with results comparable to those of periodontists' visual annotations. This supports its utility as a pro-social XAI screening solution for scalable, community-based oral health promotion, warranting further refinement and validation against clinical probing.
CLINICAL SIGNIFICANCE: The mHealth tool enables non-invasive, readily accessible, and near-real-time detection of visual signs of gingival inflammation in intraoral photographs. This approach may facilitate early identification of individuals who require referral, support, home care reinforcement, and enable timely intervention for underserved older adults in community settings. Further clinical studies, including validation against periodontal probing, are needed to fully evaluate its potential to promote oral health through scalable community-based screening.
MATERIAL AND METHODS: Older adults (age 60+) were recruited from nine community centres. Frontal intraoral photographs were captured with smartphones and evaluated using an mHealth tool that classified gingival regions as "no inflammation" (health) in green, "mild changes/questionable" in yellow, or "inflamed" (diseased) in red at the pixel level. In this study, the screening performance was validated only for visible anterior teeth. Two calibrated periodontists independently assessed these photographs, annotated the regions of healthy/mild changes/questionable/diseased, and served as the benchmark for comparison. Performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, F1-score, and mean intersection over union (mIOU), were calculated by comparing the AI outputs to the benchmark.
RESULTS: Of 262 invited older adults, 163 (62.21%) participated, and each centre recruited 17-44 participants. Inter-rater agreement between periodontists was substantial (Prevalence- and Bias-Adjusted Kappa (PABAK) = 0.62). GumAI exhibited sensitivity of 0.91 (95% CI 0.91-0.92), specificity of 0.87 (95% CI 0.87-0.87), PPV of 0.90 (95% CI 0.89-0.90), NPV of 0.89 (95% CI 0.89-0.89), accuracy of 0.89 (95% CI 0.89-0.89), F1-score of 0.90, and mIOU of 0.83.
CONCLUSION: The mHealth tool demonstrated potential to detect visual signs of gingival inflammation in intraoral photographs of older adults in a community setting, with results comparable to those of periodontists' visual annotations. This supports its utility as a pro-social XAI screening solution for scalable, community-based oral health promotion, warranting further refinement and validation against clinical probing.
CLINICAL SIGNIFICANCE: The mHealth tool enables non-invasive, readily accessible, and near-real-time detection of visual signs of gingival inflammation in intraoral photographs. This approach may facilitate early identification of individuals who require referral, support, home care reinforcement, and enable timely intervention for underserved older adults in community settings. Further clinical studies, including validation against periodontal probing, are needed to fully evaluate its potential to promote oral health through scalable community-based screening.