Prediction of total corneal power from measuredanterior corneal power on the IOLMaster 700using a feedforward shallow neural network
Sprache des Titels:
Englisch
Original Kurzfassung:
ABSTRACT.Background: The corneal back surface is known to add some astigmatism against-the-rule, which has to be considered incataract surgery with toric lens implantation. Thepurposeof this study was to set up a deep learning algorithm which predicts thetotal corneal power from keratometry and biometric measures.Methods: Based on a large data set of meas urements with the IOLMaster 700 from two clinical centres, data fromN = 21 108 eyes were included, each record containing valid data for keratometry K, total keratometry TK, axial lengthAL, central corneal thickness CCT, anterior chamber depth ACD, lens thickness LT and horizontal corneal diameterW2W from an individual eye. After a vector decom position of K and TK into equivalent power (.EQ) and projections ofastigmatism to the 0°/90° (.AST0°) and 45°/135° (.AST45°) axis, a multi-output feedf orward shallow neural network wasderived to predict TK from K, AL, CCT, ACD, LT, W2W and patient age.Results: After some trial and error, the neural network having a Levenberg?Marquardt training function and three hiddenlayers (10/8/5 neurons) perfo rmed best and showed a fast convergence. The data set was split into training data (70%),validation data (15%) and test data (15%). The prediction error (predicted corneal power CPpredminus TK) of the networktrained with the training and cross-validated with test data showed systematically narrower distributions for CPEQ-TKEQ, CPAST0°-TKAST0°and CPAST45°-TKAST45°compared with KEQ-TKEQ, KAST0°-TKAST0°and KAST45°-TKAST45°. There was no systematic o?set in the components between CPpredand TK.Conclusion: Unlike any ?xed correction term, which can compensate only for a static intercept of the astigmatic componentsTKEQ, TKAST0°andTKAST45°compared with KEQ, KAST0°andKAST45°,ourtrainedneuralnetwork was ableto reduce thevariance in the prediction error signi?cantly. This neural network could be used to account for the corneal back surfaceastigmatism for biometers where the corneal back surface measurement or total keratometry is not available.