Intraocular Lens Power Formula Selection Using Support Vector Machines

Abstract

Purpose: In cataract surgery, the defected lens is replaced with an artificial intraocular lens (IOL). The refraction power of this lens is specified by ophthalmologists before the surgery. There are different formulas that propose the IOL power based on corneal power and axial length. Six common formulas is used in this study and the one which minimizes the postoperative error for a specific patient have to be selected.
Methods: Refraction is measured three times at most, during six month after surgery and the best result is considered as postoperative refraction for each patient. A Support Vector Machine (SVM) is used to classify the data to two groups based on axial length and corneal power. Each class needs a formula with a specific tendency toward stronger or weaker IOL lenses according to the feature vector.
Results: Experimental tests lead to a nearly diagonal confusion matrix which supports the performance of the proposed method strongly. Mean Absolute Error (MAE) is 0.47 which shows 6% decrease in postoperative refraction error compared to the best reported result.
Conclusions: In calculating IOL power, we expect stronger IOL powers for eyes having shorter axial length or weaker corneal power. In the contrary, a weaker IOL power is expected for longer axial length and stronger corneal power. But experimental results show that for the first group, formulas proposing weaker powers win the race for decreased postoperative refraction error while for the second group, formulas leading to stronger powers perform better. This shows that these formulas overestimate and underestimate for marginal cases.