Frontiers in medicine · 2025
Development and evaluation of machine learning models for individualized prediction of myopia control efficacy treated with overnight orthokeratology.
СтатьяНазвание 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.
Zhang Lan, Gao Mingjun, Wang Yiru, Zhang Siqi, Zhu Huailin, Zhao Qi
Аннотация
PURPOSE: The primary objective of this study is to develop a predictive model utilizing fundamental clinical and ocular measurements to predict the effect of overnight orthokeratology on myopia control. Accordingly, this study aims to assist ophthalmologists in selecting adolescent myopia control methods.
METHODS: This retrospective study used one-year follow-up data of 225 myopia children treated with orthokeratology. Using the random sampling method, 225 samples were randomly divided into a training set (n = 180) and a test set (n = 45). LASSO regression identified predictive factors correlated with controlling myopia. The final features are input into the machine learning model for prediction model construction to predict 1-year axial length elongation. The prediction performance was evaluated according to the accuracy and AUC of the training set and the test set. DCA was used to assess the clinical benefits of the model.
RESULTS: Five features (age, diopter, flat keratometry, corneal higher-order aberrations (6 mm), and intraocular trefoil (6 mm) were used to build the machine learning model (p < 0.01)). Based on the accuracy, ROC, and DCA curves, the prediction performance and clinical practicability of five prediction models: KNN, SVM, RF, Extra Trees, and XGBoost were compared. In the DCA, all machine learning models consistently achieved greater net benefits within the clinical threshold range. SVM demonstrated the highest predictive quality with an AUC of 0.877 in the training and 0.828 in the external validation set.
CONCLUSION: We developed and validated several prediction models for individualized prediction of myopia control efficacy treated with overnight orthokeratology through machine learning, using easily obtained clinical and corneal topography features. This cost effective strategy helps ophthalmologists predict the effect of using orthokeratology in children, and make timely adjustments to myopia control methods. The differential features selected by this model can also provide insights for optimizing lens design.
METHODS: This retrospective study used one-year follow-up data of 225 myopia children treated with orthokeratology. Using the random sampling method, 225 samples were randomly divided into a training set (n = 180) and a test set (n = 45). LASSO regression identified predictive factors correlated with controlling myopia. The final features are input into the machine learning model for prediction model construction to predict 1-year axial length elongation. The prediction performance was evaluated according to the accuracy and AUC of the training set and the test set. DCA was used to assess the clinical benefits of the model.
RESULTS: Five features (age, diopter, flat keratometry, corneal higher-order aberrations (6 mm), and intraocular trefoil (6 mm) were used to build the machine learning model (p < 0.01)). Based on the accuracy, ROC, and DCA curves, the prediction performance and clinical practicability of five prediction models: KNN, SVM, RF, Extra Trees, and XGBoost were compared. In the DCA, all machine learning models consistently achieved greater net benefits within the clinical threshold range. SVM demonstrated the highest predictive quality with an AUC of 0.877 in the training and 0.828 in the external validation set.
CONCLUSION: We developed and validated several prediction models for individualized prediction of myopia control efficacy treated with overnight orthokeratology through machine learning, using easily obtained clinical and corneal topography features. This cost effective strategy helps ophthalmologists predict the effect of using orthokeratology in children, and make timely adjustments to myopia control methods. The differential features selected by this model can also provide insights for optimizing lens design.