Predictive Correction Model for Corneal Back Surface Astigmatism With IOLMaster700 Keratometry Data in a Cataractous...

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Optical Metrology

Background:

• Accurate prediction of total keratometry (TK) is crucial for intraocular lens (IOL) power calculation in cataract surgery.
• Classical keratometry (K) measurements may not fully represent the total corneal power, necessitating predictive models.

Purpose of the Study:

• To develop and validate models for predicting TK power vector components (TKC0, TKC45) from K measurements (KC0, KC45).
• To evaluate the performance of various predictive models using a large dataset of IOLMaster 700 measurements.

Main Methods:

• A retrospective, multicentric study utilized 13,6378 IOLMaster 700 measurements.
• Developed and compared global and segmented constant, linear, and harmonic models against classical and segmented linear reference models.
• Performance was cross-validated using root-mean-squared error (RMSE) on 80% training and 20% test data.

Main Results:

• Correction models reduced RMSE by 40%-42%, from 0.173 D to 0.100-0.104 D.
• Segmented and linear models showed slight performance improvements over global and constant models, respectively.
• Individually adjusted segmented models (CMS, LMS) slightly outperformed reference models (CMR, LMR) with no evidence of overfitting.

Conclusions:

• All tested correction models demonstrated similar predictive performance for TK vector components.
• A simple global constant model is recommended for its ease of implementation in clinical software.
• This approach can simplify and potentially improve IOL power calculations in routine cataract surgery.