Reviewed by Lexie CornerApr 15 2025
In a recent paper, Dr. Li Li, Dr. Jifeng Yu, and Dr. Nan Liu from the Department of Ophthalmology at Capital Medical University in China discussed the applications and challenges of artificial intelligence in myopia, including detection, risk factor evaluation, and prediction models.
Artificial intelligence models can diagnose myopia by scanning the patient’s retina. Image Credit: Chris Urbanowicz at Openverse
Myopia (nearsightedness) affects approximately two billion people globally. If left uncorrected, myopia can lead to vision impairment, impacting education, employment opportunities, and overall quality of life. By 2050, it is estimated that about half of the world's population will be myopic.
High myopia is often associated with complications that can result in visual impairment, further reducing patients' quality of life and contributing to the global medical and economic burden. Early detection of myopia is, therefore, essential to prevent visual impairment in affected individuals.
Artificial intelligence (AI) has advanced medical practices and may offer solutions to address this global health concern. Subsets of AI, such as machine learning (ML) and deep learning (DL), can assist in data analysis for diagnosing diseases, predicting risk factors, and identifying biomarkers.
AI models, trained using ML/DL, can detect myopia in fundus photographs and optical coherence tomography images. By inputting a large number of fundus images from myopic patients, AI can learn to identify subtle changes in retinal color and patterns associated with myopia, allowing the model to diagnose future patients based on these images.
Additionally, self-monitoring devices, such as the SVOne—a portable device that measures eye defects using a wavefront sensor—can detect refractive errors in the eyes through AI algorithms. These devices could connect to online image libraries, which AI can use to help diagnose myopia. AI could also be trained to identify behavioral changes linked with the onset of myopia.
This technology is particularly useful for early diagnosis of myopia in children, a demographic in which the condition is often overlooked. For example, the Vivior monitor employs ML algorithms to detect changes in visual behavior, such as time spent on near-vision tasks, in children aged 6 to 16.
Furthermore, ML techniques such as support vector machines, logistic regression, and XGBoost can be applied to identify risk factors for myopia.
An XGBoost-based model can be fed large quantities of longitudinal data, allowing it to learn the outcomes and associated risk factors of myopia in numerous patients. This, in turn, allows the model to assess the risk factors of new patients based on their genetics, family history, environment, and physiological parameters.
Dr. Li Li, Professor, Department of Ophthalmology, Capital Medical University
Predicting the progression and prognosis of myopia can help doctors customize treatment strategies. On a broader scale, it can inform clinical practices and policy decisions to support myopia management. By inputting large datasets that include biometric data, refractive information, treatment responses, and ocular images from various myopia patients, AI can be trained to predict outcomes for new patients.
While AI offers significant potential in myopia management, several challenges must be addressed. First, the dataset used to train AI models must be accurate and of high quality. Issues such as bias, false positives/negatives, and poor data quality can undermine the model's diagnostic and predictive accuracy. Second, most AI models are trained using data from large hospitals, which may not represent the patient populations seen in smaller clinics, leading to disparities between real-world and training datasets.
Additionally, since AI models are not licensed medical professionals, they may lack the clinical reasoning necessary to justify their diagnoses, potentially leading to rejection by healthcare providers. Lastly, due to the large amounts of patient data required for training AI models, safeguarding the privacy of patient medical records becomes a critical concern.
While our study highlights the remarkable progress made in the clinical application of AI in myopia, further studies are needed to overcome the technological challenges. By building high-quality datasets, improving the model’s capacity to process multimodal image data, and improving human-computer interaction capability, the AI models can be further improved for widespread clinical application.
Dr. Jifeng Yu, Department of Ophthalmology, Capital Medical University
Journal Reference:
Liu, N., et al. (2025) Application of artificial intelligence in myopia prevention and control. Pediatric Investigation. doi:10.1002/ped4.70001.