View abstract

This abstract is assigned to sessions

Title

A new nomogram for laser corneal refractive surgery based on Artificial intelligence

Purpose	To predict delta between programmed laser treatment and really delivered treatment. To compare the predicted delta with the wavelight nomogram To study relative importance of each preoperative parameter.
Methods	We reviewed 2728 myopic eyes operated with Wavelight Refractive Suite. Preoperative and one-month postoperative refractions and programmed treatment defined the observed delta. Delta is then predicted from clinical, laser and topographical preoperative parameters using machine learning regression algorithm.
Results	Machine learning nomogram reduced the error and improve outcomes. Outcomes are also improved after using machine learning and Wavelight nomogram. Programmed spherical equivalent is the more influential parameter to predict sphere and programmed astigmatism is the more influential parameter to predict cylinder.
Conclusion	Machine learning could allow to increase laser treatment accuracy and our understanding of causes of residual errors.
Conflict of interest	No

Authors 1

Last name	CRAHAY
Initials	F-X
Department	CHR Citadelle
City	Liège

Authors 2

Last name	Debellemaniere
Initials	G
Department	Fondation Ophtalmologique Adolphe de Rothschild
City	Paris

Authors 3

Last name	Rampat
Initials	R
Department	Fondation Ophtalmologique Adolphe de Rothschild
City	Paris

Authors 4

Last name	Moran
Initials	S
Department	Fondation Ophtalmologique Adolphe de Rothschild
City	Paris

Authors 5

Last name	Gatinel
Initials	D
Department	Fondation Ophtalmologique Adolphe de Rothschild
City	Paris