THOSE OF US seeing patients regularly know that obtaining the best-fitted contact lens can be difficult on the first attempt due to the corneal irregularities of a keratoconus patient. Practitioners have the options of using fitting guides provided by lens manufacturers, fitting software, and consultations to assist in getting the best fit. It is common knowledge among contact lens practitioners that the posterior radius of curvature is a critical parameter in the corneal GP fitting process (Ortiz-Toquero et al, 2019).
Recently, there has been a shift toward artificial intelligence (AI) for the detection and assessment of disease, especially in the areas of diabetic retinopathy, glaucoma, and age-related macular degeneration (Ting et al, 2019). In terms of keratoconus, deep learning (DL) techniques are being used effectively to detect and classify clinical and subclinical keratoconus (Kamiya et al, 2019). There have only been 2 studies that explore the use of DL to predict rigid lens parameters for keratoconus patients with hopeful results (Hashemi et al, 2020).
This year, Abadou and colleagues (2025) compared the efficiency of a few AI frameworks against a reference method of the mean radius of curvature (K) to predict the posterior radius of curvature for a rigid corneal GP contact lens in keratoconic eyes. This retrospective study looked at 197 keratoconic eyes with 135 patients fitting conventional GP lenses and 1 or more topographies that were taken between January 2020 and September 2022. For the AI analysis, 2 types of topographic data were used: indices and reconstructed maps from raw data. Three approaches were utilized for AI: standard machine learning (ML) methods and DL (multilayer perception [MLP] based on topographic indices and convolution neural networks [CNNs] based on topographic maps).
The results of the study showed that a CNN showed the best accuracy. This framework had a higher r2 coefficient compared to both the mean K method and the commercial GP guidelines. The CNN combined 3 topographic maps, and this outperformed both the methods (mean K and GP fitting guide). This AI method allowed a predicted best-fitted rigid contact lens (RBFCL) using the combination of 3 topographic map analyses.
The paper by Abadou and colleagues (2025) is setting the stage for the future of corneal GP and scleral lens fittings. As the research starts to use GP lenses in comparison to AI, the next step will be to research and implement the results into the process of scleral lens fittings. Many practitioners already use advanced technology (scleral mapping, wavefront, impression-based) to assist in fittings. The incorporation of AI to reduce fitting time and followups will follow.
CNNs have also been used for keratoconus detection and classification (Subramanian et al, 2022). Its focus on image analysis and filters to detect patterns and features helps it capture information about the cornea’s structure and curvature. The use of this information could result in a reduction of the trial-and-error period often associated with difficult-to-fit keratoconus patients. It is important to note that there are many limitations to the Subramanian and colleagues study, including size, not all lens parameters included, and comparison to only one GP lens design.
Practitioners are seeing that AI will be a part of the future in all aspects of their lives, and now in their careers. Optometry is already evolving in using AI for detection and analysis, and it will likely be part of their future practice. The use of AI in scleral lens fittings is in our imminent future, and more studies like this one will need to be performed on scleral lenses.
References
1. Ortiz-Toquero S, Rodriguez G, de Juan V, Martin R. Gas permeable contact lens fitting in keratoconus: Comparison of different guidelines to back optic zone radius calculations. Indian J Ophthalmol. 2019;67(9):1410-1416. doi: 10.4103/ijo.IJO_1538_18
2. Ting DSW, Pasquale LR, Peng L, et al. Artificial intelligence and deep learning in ophthalmology. Br J Ophthalmol. 2019;103(2):167-175. doi: 10.1136/bjophthalmol-2018-313173
3. Kamiya K, Ayatsuka Y, Kato Y, et al. Keratoconus detection using deep learning of colour-coded maps with anterior segment optical coherence tomography: a diagnostic accuracy study. BMJ Open. 2019;9(9):e031313. doi: 10.1136/bmjopen-2019-031313
4. Hashemi S, Veisi H, Jafarzadehpur E, Rahmani R, Heshmati Z. Multi-view deep learning for rigid gas permeable lens base curve fitting based on Pentacam images. Med Biol Eng Comput. 2020;58(7):1467-1482. doi: 10.1007/s11517-020-02154-4
5. Abadou J, Dahan S, Knoeri J, Leveziel L, Bouheraoua N, Borderie VM. Artificial intelligence versus conventional methods for RGP lens fitting in keratoconus. Cont Lens Anterior Eye. 2025 Feb;48(1):102321. doi: 10.1016/j.clae.2024.102321.
6. Subramanian P, Ramesh GP. Keratoconus Classification with Convolutional Neural Networks Using Segmentation and Index Quantification of Eye Topography Images by Particle Swarm Optimisation. Biomed Res Int. 2022 Mar 22:2022:8119685. doi: 10.1155/2022/8119685.