Researchers at the Tokyo Medical and Dental University (TMDU) have developed an advanced artificial intelligence (AI) model that shows promise in predicting the risk of long-term visual impairment in individuals with high myopia. High myopia, a condition characterized by extreme nearsightedness, is among the leading causes of irreversible blindness in many parts of the world.
The innovative machine-learning model developed by the TMDU team has demonstrated remarkable accuracy in predicting and visualizing the long-term risk of visual impairment. Machine learning, a form of AI that enables software systems to learn from data, holds immense potential in improving its performance over time.
While individuals with high myopia can see nearby objects clearly, focusing on objects at a distance poses significant challenges. While corrective measures such as glasses, contact lenses, or surgery can aid in vision correction, high myopia can lead to a condition known as pathologic myopia, which is a leading cause of blindness.
Lead author of the study, Yining Wang, explains, "We know that machine-learning algorithms work well in identifying changes and complications in myopia. However, in this study, we aimed to evaluate their effectiveness in long-term predictions."
The study, published in the esteemed journal JAMA Ophthalmology, involved analyzing the visual acuity of 967 Japanese patients at TMDU's Advanced Clinical Center for Myopia over a span of 3 and 5 years. The researchers compiled a dataset consisting of 34 variables commonly collected during ophthalmic examinations, including age, current visual acuity, and corneal diameter.
Various machine-learning models were tested, including random forests and support vector machines. Ultimately, the logistic regression-based model emerged as the most accurate in predicting visual impairment over a 5-year period.
However, the researchers emphasized that predicting outcomes is just one aspect of the study. They also underscored the importance of presenting the model's output in a manner that patients can easily comprehend, facilitating convenient clinical decisions.
To address this, the researchers employed a nomogram, which visually represents the classification model. Each variable is assigned a line, with the length indicating its predictive importance for visual acuity. These lengths can be converted into points, which can then be added to yield a final score, explaining the individual's risk of future visual impairment.
The loss of vision can have devastating consequences, both financially and physically, significantly impacting an individual's independence. In 2019, severe visual impairment resulted in an estimated global productivity decrease of USD 94.5 billion.
While the model still requires further evaluation on a broader population, this study highlights the immense potential of machine-learning models in addressing this pressing public health concern. The implications extend beyond individuals, benefiting society as a whole.
