Study on the Diagnostic Efficacy of ICL Selection and Prediction Depth Model Based on Eye Images

Diagnostic Efficacy of Deep Neural Network Algorithm Based on Preoperative Scheimpflug-based Anterior Segment Image for Implantable Collamer Lens Selection and Prediction

To evaluate the diagnostic efficacy of deep learning network model in implantable collamer lens selection and prediction in a multicenter cross-sectional study

Posterior chamber intraocular lens implantation is an main choice for myopia correction. Implantable collamer lens (ICL) is currently the most widely used, and the official reference index is mainly based on biological parameters obtained from eye images. The parameter acquisition and selection of ICL design are often controversial, forcing the doctors to synthesize multiple modal data, making the optimization of ICL formula being a focus of attention in refractive surgery. This research aimed to build an image-based ICL prediction algorithm to assist human physicians in decision-making and improve the accuracy, safety and predictability of ICL implantation.

Inclusion Criteria:

1. Aged 18-45 years ;
2. Myopia, with or without astigmatism, annual diopter change ≤ 0.50 D for 2 consecutive years ;
3. Anterior chamber depth ≥ 2.80 mm ;
4. Corneal endothelial cell count ≥ 2000 / mm2, stable cell morphology ;
5. There were no other ocular diseases that significantly affected vision and / or systemic organic lesions that affected surgical recovery.

Exclusion Criteria:

1. There were no other ocular diseases that significantly affected vision and / or systemic organic lesions that affected surgical recovery;
2. Have a history of corneal refractive surgery or intraocular surgery ;
3. Corneal endothelial cell count is low ;
4. Those with systemic diseases ;
5. Lactating or pregnant women.