This content is only available for ASCRS Members
This content from the 2020 ASCRS Virtual Annual Meeting is only available to ASCRS members. To log in, click the teal "Login" button in the upper right-hand corner of this page.
Filed Under
Astigmatism
keratotomy
femtosecond
astigmatic keratotomy
machine learning
astigmatism correction
2020 paper presentation
Purpose
To analyze femtosecond laser arcuate keratotomy (FLAK) using modern machine learning techniques, thereby finding important corneal variables affecting results, and algorithms to better predict results.
Methods
125 retrospective patient records were analyzed with preoperative topography, FLAK with iris registration and cataract surgery with recording of 2.4 mm incision was performed, and postop topography recorded minimum 6 weeks postop. Results were analyzed using Alpins analysis. SIA was measured, and a new definition of Moment was developed. Moment measures the resultant misalignment of SIA from TIA. SIA magnitude was the first outcome measure, and all variables were entered into a machine learning ensemble. A mixed model was developed to predict SIA magnitude, and second model predicted Moment. A 2 stage genetic algorithm predicts best parameters to predict desired SIA and smallest Moment.
Results
A machine learning ensemble 92% weighted with a linear model, predicted SIA magnitude with Rsquared= 0.89, and the Moment with Rsquared = 0.88; The most important variables are Ks at 3mm, Total Corneal Astigmatism, including posterior astigmatism, and Q shape Factor.
Conclusion
Machine Learning can use all corneal variables to arrive at better predictive models to predict the FLAK SIA magnitude, and the risk of misalignment of the SIA angles.
To analyze femtosecond laser arcuate keratotomy (FLAK) using modern machine learning techniques, thereby finding important corneal variables affecting results, and algorithms to better predict results.
Methods
125 retrospective patient records were analyzed with preoperative topography, FLAK with iris registration and cataract surgery with recording of 2.4 mm incision was performed, and postop topography recorded minimum 6 weeks postop. Results were analyzed using Alpins analysis. SIA was measured, and a new definition of Moment was developed. Moment measures the resultant misalignment of SIA from TIA. SIA magnitude was the first outcome measure, and all variables were entered into a machine learning ensemble. A mixed model was developed to predict SIA magnitude, and second model predicted Moment. A 2 stage genetic algorithm predicts best parameters to predict desired SIA and smallest Moment.
Results
A machine learning ensemble 92% weighted with a linear model, predicted SIA magnitude with Rsquared= 0.89, and the Moment with Rsquared = 0.88; The most important variables are Ks at 3mm, Total Corneal Astigmatism, including posterior astigmatism, and Q shape Factor.
Conclusion
Machine Learning can use all corneal variables to arrive at better predictive models to predict the FLAK SIA magnitude, and the risk of misalignment of the SIA angles.
View More Presentations from this Session
This presentation is from the session "SPS-102 Toric IOLs - Astigmatism" from the 2020 ASCRS Virtual Annual Meeting held on May 16-17, 2020.