Sharif University of Technology Sharif Journal of Mechanical Engineering 2676-4725 41 2 2025 12 21 A Ring Selection Platform for Treatment of Keratoconus A Ring Selection Platform for Treatment of Keratoconus 17 31 24023 10.24200/j40.2025.64700.1711 FA Amirhossein Khademi Faculty of Mechanical Engineering, Sharif University of Technology, Tehran, Iran. Mohsen Asghari Faculty of Mechanical Engineering, Sharif University of Technology, Tehran, Iran. Mohammad Reza Naderi Eshkaftaki Faculty of Mechanical Engineering, Sharif University of Technology, Tehran, Iran. Journal Article 2024 10 02 This study presents a fully integrated platform for selecting appropriate intrastromal corneal ring segments (ICRS) for keratoconus treatment using a combination of finite element modeling and machine-learning techniques. A patient-specific corneal geometry was reconstructed using Pentacam-derived elevation maps, followed by meshing and biomechanical simulation of ring implantation at various depths and angular positions. Optical parameters of the cornea were calculated using curvature-based relationships (Eqs. (3)–(4)). A comprehensive database of 288 simulated ring-implantation scenarios was generated by varying Ring radius, Implantation depth, ring implantation zone, and arc length (Fig. 7). To predict keratometric outcomes, a random forest regression model and a deep learning architecture were developed and trained on the simulation-derived dataset. Model validation demonstrated acceptable accuracy using an independent rectangular-groove benchmark (Fig. 8). The trained algorithms were finally tested on a separate patient to evaluate generalization capacity. The results indicate that machine-learning prediction of ring-induced corneal response is feasible and can support treatment planning. This platform provides a foundation for developing preoperative decision-support tools to enhance clinical outcomes in keratoconus ring implantation. This study presents a fully integrated platform for selecting appropriate intrastromal corneal ring segments (ICRS) for keratoconus treatment using a combination of finite element modeling and machine-learning techniques. A patient-specific corneal geometry was reconstructed using Pentacam-derived elevation maps, followed by meshing and biomechanical simulation of ring implantation at various depths and angular positions. Optical parameters of the cornea were calculated using curvature-based relationships (Eqs. (3)–(4)). A comprehensive database of 288 simulated ring-implantation scenarios was generated by varying Ring radius, Implantation depth, ring implantation zone, and arc length (Fig. 7). To predict keratometric outcomes, a random forest regression model and a deep learning architecture were developed and trained on the simulation-derived dataset. Model validation demonstrated acceptable accuracy using an independent rectangular-groove benchmark (Fig. 8). The trained algorithms were finally tested on a separate patient to evaluate generalization capacity. The results indicate that machine-learning prediction of ring-induced corneal response is feasible and can support treatment planning. This platform provides a foundation for developing preoperative decision-support tools to enhance clinical outcomes in keratoconus ring implantation. https://sjme.journals.sharif.edu/article_24023_1ea5e6f2837d15cbe7a9989bb9ff07af.pdf