Neural Network-Based Method for Continuous Curvilinear Capsulorhexis Edges Segmentation and Its Effect Evaluation

With the increasing popularity of electronic products and more and more people with excessive eye use, the incidence of ophthalmic diseases is increasing year by year, of which cataract is the world's first blinding ophthalmic diseases, ultrasonic emulsification cataract aspiration is the main method of treatment of cataracts, and the continuous curvilinear capsulorhexis is one of the most important step, the effectiveness of continuous curvilinear capsulorhexis has a bearing on the effect of the subsequent surgery. The current method of assessing the capsulorhexis results still has a lot of shortcomings, mainly relying on the doctor's observation to assess, which is easy to be affected by disturbing factors, for this reason, we proposed a deep learning-based segmentation model of capsulorhexis results in order to achieve the quantitative assessment of capsulorhexis results, the segmentation model of capsulorhexis results is based on the construction of the U-Net and SegFormer, and at the same time we use the segmentation results to perform quantitative calculations to evaluate capsulorhexis results. The experimental results show that our proposed model has a good effect compared with other methods, MIoU and MPA reach 71.16% and 85.41%, so this method can achieve better evaluation of capsulorhexis results.