[Advancements in machine learning applications in refractive surgery].

Abstract

Refractive error is a significant factor contributing to visual impairment, imposing a relatively large burden on the social economy. Although refractive surgery is an important corrective method, it faces challenges in clinical practice, such as precise preoperative screening, personalized surgical plan design, and prevention of postoperative complications. This study focuses on the application of machine learning in the field of refractive surgery. Through a comprehensive analysis of relevant literature, it is found that machine learning plays a positive role in multiple key aspects. In preoperative screening, it can effectively improve the accuracy of keratoconus screening and assist in precisely selecting surgical candidates and determining surgical methods. During surgical design, it can optimize the plans for corneal refractive surgery and implantable Collamer lens implantation, enhancing the predictability of surgeries. In postoperative evaluation and prediction, it helps to assess surgical outcomes, identify high-risk patients for refractive regression, and assist in calculating the power of intraocular lenses. However, machine learning has limitations in practical applications, such as the "black box" nature of algorithms, uneven data quality, and lack of multimodal data integration. By systematically reviewing its application status and limitations, this review hopes to provide references for subsequent research, help overcome difficulties, and promote the more in-depth and rational application of machine learning in the field of refractive surgery, thereby improving the overall level of refractive surgery.