Prediction of total corneal power from measured anterior corneal power on the IOLMaster 700 using a feedforward shallow neural network

Abstract

Background: The corneal back surface is known to add some astigmatism against-the-rule, which has to be considered in cataract surgery with toric lens implantation. The purpose of this study was to set up a deep learning algorithm which predicts the total corneal power from keratometry and biometric measures. Methods: Based on a large data set of measurements with the IOLMaster 700 from two clinical centres, data from N = 21 108 eyes were included, each record containing valid data for keratometry K, total keratometry TK, axial length AL, central corneal thickness CCT, anterior chamber depth ACD, lens thickness LT and horizontal corneal diameter W2W from an individual eye. After a vector decomposition of K and TK into equivalent power (.EQ) and projections of astigmatism to the 0°/90°(.AST 0°) and 45°/135°(.AST 45°) axis, a multi-output feedforward shallow neural network was derived 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 hidden layers (10/8/5 neurons) performed 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 CP pred minus TK) of the network trained with the training and cross-validated with test data showed systematically narrower distributions for CPEQ-TKEQ, CPAST 0°-TKAST 0°a nd CPAST 45°-TKAST 45°c ompared with KEQ-TKEQ, KAST 0°-TKAST 0°a nd KAST 45°-TKAST 45°. There was no systematic offset in the components between CP pred and TK. Conclusion: Unlike any fixed correction term, which can compensate only for a static intercept of the astigmatic components TKEQ, TKAST 0°a nd TKAST 45°c ompared with KEQ, KAST 0°a nd KAST 45°, our trained neural network was able to reduce the variance in the prediction error significantly. This neural network could be used to account for the corneal back surface astigmatism for biometers where the corneal back surface measurement or total keratometry is not available.

Comparison of 2 modern swept-source optical biometers—IOLMaster 700 and Anterion

“…As the most relevant data for lens power calculation, the AL, CCT, ACD and LT plus the corneal front and back surface curvature were compared. As the locations of the flat and steep meridians for the corneal front and back surfaces do not necessarily coincide, the radii of curvature cannot be directly compared [12]. Instead, we performed a vector analysis and decomposed the corneal front and back surface power into 3D power vectors [15] including equivalent power and projections of the astigmatism to the 0°/90° and to the 45°/135° meridians.…”

“…Using this method, we could easily access the respective 3D power vector of the equivalent power (independent of any internal calculation scheme implemented in the biometer software), which is referenced to the secondary principal plane of the cornea. This equivalent power derived from the Gullstrand formula is based on the curvature of both corneal surfaces, the CCT and the refractive index of the cornea and aqueous humour, but is independent from any keratometer index [12].…”

“…However, we have to be aware that the measures do not match perfectly, even with modern high-end optical biometers, and therefore, the instruments cannot be used interchangeably. It is well known that measures such as AL or CCT give mostly consistent results between instruments [5], whereas other measures such as corneal curvature data, ACD or LT show a larger degree of variation [12]. As a consequence, if a biometer is replaced by another model in a clinical environment, either the biometric measures should be transformed to correspond to the respective measures of the previously used biometer or the formula constants should be adapted to obtain the best results with the replacement device [13].…”

Evaluation of corneal power from an AS-OCT thick lens model and ray tracing: reliability of the keratometer index

“…2,3 In addition, since the keratometer does not take account of either corneal thickness or the corneal back surface, the conversion of corneal front surface radius to corneal power using a keratometer index is always a simplification strictly linked to a given schematic model of the cornea. 3,[7][8][9][10][11][12] The validity of such a model in an individual case cannot be verified by the Submitted: July 27, 2023 | Final revision submitted: November 6, 2023 | Accepted: November 7, 2023 From the Department of Experimental Ophthalmology, Saarland University, Homburg/Saar, Germany (Langenbucher, Wendelstein); Augen-und Laserklinik Castrop-Rauxel, Castrop-Rauxel, Germany (Hoffmann); School of Physical Sciences, The Open University, Milton Keynes, United Kingdom (Cayless); Department of Ophthalmology, Johannes Kepler University Linz, Linz, Austria (Wendelstein); Dr. Rolf M. Schwiete Center for Limbal Stem Cell and Aniridia Research, Saarland University, Homburg/Saar, Germany (Szentm áry); Department of Ophthalmology, Semmelweis-University, Budapest, Hungary (Szentm áry).…”

Evaluation of statistical correction strategies for corneal back surface astigmatism with toric lenses: a vector analysis

“…In addition to the keratometric astigmatism, these strategies allow the corneal back surface astigmatism and the total corneal astigmatism resulting from the corneal front and back surface to be analyzed. [21][22][23] Along with modern biometers, such as the IOLMaster 700 (IOLM), ANTE-RION, or Pentacam AXL, there are several Scheimpflug or optical coherence-based tomographers, which evaluate both corneal surfaces.…”