Predicting Post-Refractive Surgery Corneal Stiffness And Ectasia Risk Using Novel Simulation Software Before The Surgery
Published 2023 - 41st Congress of the ESCRS
Reference: FP11.04 | Type: Free paper | DOI: 10.82333/z1my-sn37
Authors: Mathew Francis* 1 , Rohit Shetty 2 , Prema Padmanabhan 3 , Riccardo Vinciguerra 4 , Paolo Vinciguerra 5 , Myrta Lippera 5 , Himanshu Matalia 6 , Pooja Khamar 6 , Nandini Chinnappaiah 6 , Deepa Mukundan 3 , Rudy M.M.A. Nuijts 7 , Abhijit Sinha Roy 8
1IBMS Lab,Narayana Nethralaya Foundation,Bangalore,India;MHeNs - Neuroscience,Maastricht University,Maastricht,Netherlands, 2Department of Corneal and Refractive surgery,Narayana Nethralaya,Bangalore,India;MHeNs - Neuroscience,Maastricht University,Maastricht,Netherlands, 3Department of Corneal and Refractive surgery,Sankara Nethralaya,Chennai,India, 4Humanitas San Pio X Hospital,Milan,Italy;The School of Engineering,University of Liverpool,Liverpool,United Kingdom, 5Humanitas Clinical and Research Center – IRCCS,Rozzano,Italy;Department of Biomedical Sciences,Humanitas University,Milan,Italy, 6Department of Corneal and Refractive surgery,Narayana Nethralaya,Bangalore,India, 7Department of Ophthalmology,Maastricht University Medical Centre,Maastricht,Netherlands, 8IBMS Lab,Narayana Nethralaya Foundation,Bangalore,India
Purpose
To predict postoperative (post-op) corneal stiffness (Kc) using a novel artificial intelligence adjusted finite element simulation software called AcuSimX™ (Narayana Ophthalmic Research and Development LLP, India) using preoperative (pre-op) data.
Setting
Refractive surgery patients from Narayana Nethralaya eye hospital (India), Sankara Nethralaya (India) and Humanitas Clinical and Research Center (Italy) who underwent photorefractive keratectomy (PRK), femtosecond laser-assisted in situ keratomileusis (LASIK) and small incision lenticule extraction (SMILE).
Methods
The study had 529 normal eyes of 529 patients, along with 11 SMILE and 2 LASIK ectasia eyes. Pre-op Corvis-ST and Pentacam HR (OCULUS Optikgerate Gmbh) data were analysed using AcuSimX™ to predict post-op Kc for all eyes. Prediction accuracy was evaluated using regression analysis (RA), intraclass correlation coefficient (ICC) and mean absolute error (MAE). The predicted post-op Kc along with pre-op parameters were used in a sigmoid-calibrated decision tree classifier (Orange: Data Mining Toolbox) to differentiate between ectasia from normal eyes. Classification efficiency was analysed using the area under the receiver operating characteristics curve (AUROC), precision and recall with “leave one out” validation.
Results
The post-op follow-up time was 7.19 [6.73, 7.65] months (mean and 95% confidence interval). AcuSimX™ predicting post-op Kc had an ICC of 0.84 [0.8, 0.87] in normal eyes. RA in the above eyes was significant (p<0.001) with a slope of 1.0, intercept of 0.51 and r value of 0.75. MAE in the above eyes was 6.24 N/m. Post-op Kc in ectasia eyes was similar (p=0.96) between in-vivo measured (74.01 [70.01, 78.01] N/m) and AcuSimX™ prediction (74.1 [69.03, 79.17] N/m). MAE in the ectasia eyes was 5 N/m. The decision tree ectasia classifier had an AUROC, precision, and recall of 1,0.99 and 0.99, respectively. The first node of the decision tree was AcuSimX™ predicted post-op Kc <= 82.85 N/m. AcuSimX™ run time was ~1.4 hours on an 8 GB Core-i7 system.
Conclusions
AcuSimX™ showed excellent predictability of post-refractive surgery corneal stiffness in normal and ectasia eyes using just the pre-op information. The near-perfect ectasia classification from normal eyes indicates that these software-predicted parameters could be used to identify post-refractive surgery ectasia risk. In clinics, the excellent predictability and ectasia classification feature of the AcuSimX™ could help in selecting the best procedure for the patients.