Artificial Intelligence Algorithms to Diagnose Glaucoma and Detect Glaucoma Progression: Translation to Clinical Practice.

Purpose: This concise review aims to explore the potential for the clinical implementation of artificial intelligence (AI) strategies for detecting glaucoma and monitoring glaucoma progression. Methods: Nonsystematic literature review using the search combinations "Artificial Intelligence," "Deep Learning," "Machine Learning," "Neural Networks," "Bayesian Networks," "Glaucoma Diagnosis," and "Glaucoma Progression." Information on sensitivity and specificity regarding glaucoma diagnosis and progression analysis as well as methodological details were extracted. Results: Numerous AI strategies provide promising levels of specificity and sensitivity for structural (e.g. optical coherence tomography [OCT] imaging, fundus photography) and functional (visual field [VF] testing) test modalities used for the detection of glaucoma. Area under receiver operating curve (AROC) values of > 0.90 were achieved with every modality. Combining structural and functional inputs has been shown to even more improve the diagnostic ability. Regarding glaucoma progression, AI strategies can detect progression earlier than conventional methods or potentially from one single VF test. Conclusions: AI algorithms applied to fundus photographs for screening purposes may provide good results using a simple and widely accessible test. However, for patients who are likely to have glaucoma more sophisticated methods should be used including data from OCT and perimetry. Outputs may serve as an adjunct to assist clinical decision making, whereas also enhancing the efficiency, productivity, and quality of the delivery of glaucoma care. Patients with diagnosed glaucoma may benefit from future algorithms to evaluate their risk of progression. Challenges are yet to be overcome, including the external validity of AI strategies, a move from a "black box" toward "explainable AI," and likely regulatory hurdles. However, it is clear that AI can enhance the role of specialist clinicians and will inevitably shape the future of the delivery of glaucoma care to the next generation. Translational Relevance: The promising levels of diagnostic accuracy reported by AI strategies across the modalities used in clinical practice for glaucoma detection can pave the way for the development of reliable models appropriate for their translation into clinical practice. Future incorporation of AI into healthcare models may help address the current limitations of access and timely management of patients with glaucoma across the world.

Differences in Optic Nerve Head Blood Flow Regulation in Normal Tension Glaucoma Patients and Healthy Controls as Assessed With Laser Speckle Flowgraphy During the Water Drinking Test.

PRéCIS:: Optic nerve head (ONH) perfusion was not significantly altered in subjects with normal tension glaucoma (NTG) nor in healthy individuals after performing the water drinking test (WDT), because of its limited effect on ocular perfusion pressure (OPP). PURPOSE: ONH blood flow can be maintained stable in healthy individuals because of a physiological phenomenon called autoregulation. Impairment of autoregulation has been shown especially under condition of NTG. The purpose of this study was to investigate the ONH blood flow autoregulation in patients with NTG by influencing the OPP with the WDT. METHODS: The study included 9 eyes from 9 white patients with a diagnosis of NTG and 9 eyes from age-matched and sex-matched healthy individuals. In the glaucoma group, the antiglaucomatous therapy was paused 3 weeks before the investigations. Measurements of ONH blood flow were performed with laser speckle flowgraphy. After baseline measurements, individuals ingested 800 mL of water in <5 minutes. Measurements were repeated after 15, 30, and 45 minutes. RESULTS: The water ingestion led to a significant rise in intraocular pressure (P<0.001) but also mean arterial pressure (P<0.001) in both groups. This resulted in stable OPP (P=0.051) with no significant difference between the groups (P=0.43). ONH blood flow remained stable over time in both groups (P=0.719). No significant interaction of time and group was shown for all parameters. CONCLUSIONS: Our findings show that the WDT does not significantly influence the OPP and therefore has to be considered inferior to other methods used to assess blood flow autoregulation.

Stromal remodeling and lenticule thickness accuracy in small-incision lenticule extraction: One-year results.

PURPOSE: To prospectively characterize the stromal thickness changes during the first year after myopic small-incision lenticule extraction using spectral-domain optical coherence tomography (SD-OCT). SETTING: Department of Ophthalmology, Kepler University Hospital, Linz, Austria. DESIGN: Prospective case series. METHODS: This study evaluated eyes that had small-incision lenticule extraction to treat myopia or myopic astigmatism. A high-resolution SD-OCT system (RS 3000 Advance) in conjunction with a custom image-segmentation algorithm was applied to directly measure stromal thickness within the central 5.0 mm corneal zone. Measurements were obtained preoperatively and postoperatively at 1 day, 1 week, 6 weeks, 3 months, 6 months, and 1 year. RESULTS: The study enrolled 42 eyes of 21 patients. The mean surgical refractive correction was -4.94 diopters ± 1.75 (SD). Postoperatively, the stromal thickness showed a significant decrease during the first 6 weeks, which amounted to a mean of 10.4 ± 6.3 µm at the apex (P < .001). Subsequently, the central stroma thickened by a mean of 8.8 ± 5.9 µm up until the 1-year follow-up (P < .001). One year postoperatively, the mean observed central stromal thickness reduction was 18.7 ± 5.7 µm smaller than the planned lenticule thickness. This difference was smallest 6 weeks postoperatively (mean 9.8 ± 7.8 µm). CONCLUSIONS: Significant anatomic changes in the corneal stroma were detected during the first year after small-incision lenticule extraction. The achieved lenticule thickness was systematically lower than planned, and the mismatch was more pronounced with higher lenticule thickness. Refractive outcomes did not appear to be influenced by lenticule thickness accuracy.

Intraindividual comparison between femtosecond laser-assisted and conventional cataract surgery.

PURPOSE: To compare the safety and efficacy between femtosecond laser-assisted cataract surgery using the Victus laser system and conventional cataract surgery. SETTING: Department of Ophthalmology, Kepler University Hospital, Linz, Austria. DESIGN: Prospective randomized case series. METHODS: Both eyes of patients with age-related cataract were randomized to conventional cataract surgery or femtosecond laser-assisted cataract surgery, both with intraocular lens (IOL) implantation. Postoperative follow-up was at 1 day, 1 week, 1 month, 3 months, and 6 months and comprised corrected distance visual acuity, endothelial cell density (ECD), central corneal thickness (CCT), and central retinal thickness. The main outcomes were intraoperative and postoperative complications and the effective phacoemulsification time (EPT). Intraocular lens and capsulotomy centration were evaluated using retroillumination slitlamp photography. RESULTS: The study enrolled 50 patients. No intraoperative complications occurred in either group. The ECD, CCT, and central retinal thickness were similar between the groups at all follow-up examinations (P > .05). The EPT was not statistically significantly different between the groups (P = .22). The IOL centration was similar between the groups (P = .93). CONCLUSION: Femtosecond laser-assisted and conventional cataract surgery using the mentioned system were equally safe and effective.

Corneal Epithelial Remodeling Induced by Small Incision Lenticule Extraction (SMILE).

PURPOSE: To assess the reactive changes of the corneal epithelial thickness (ET) profile induced by small incision lenticule extraction (SMILE) for treatment of myopia. METHODS: This prospective observational study included 46 eyes of 23 patients that were scheduled for myopic SMILE. High-resolution spectral-domain optical coherence tomography (SD-OCT) and a custom image segmentation algorithm were applied to measure corneal ET at multiple time points within the central 5-mm zone. Postoperative ET alterations were correlated with treatment parameters and refractive outcomes. RESULTS: Mean age was 33 ± 6 years and mean spherical equivalent of surgical refractive correction was -4.78 ± 1.75 diopters. The average ET (averaged over the central 5-mm zone) increased from 52.3 ± 3.6 µm preoperatively to 57.7 ± 5.1 µm 6 months postoperatively (P < 0.01). Average ET was 101.9% of the preoperative thickness at 24 hours postoperatively, 103.2% at 1 week, 106.7% at 6 weeks, 109.3% at 3 months, and 110.4% at 6 months postoperatively. The epithelial thickening response could be modeled by an exponential recovery function and stabilized after 3 months. The extent of epithelial hyperplasia was highly dependent on the magnitude of the induced refractive correction (P = 0.002). CONCLUSIONS: In this initial study of corneal ET remodeling after myopic SMILE, significant epithelial thickening was detected as a function of the extent of surgical refractive correction. Moreover, the epithelial remodeling response to the corneal refractive change appeared to decrease with higher age. In our hands, the observed epithelial changes did not appear to affect the refractive outcome of SMILE. (ClinicalTrials.gov number, NCT02614625).

Semiautomated SD-OCT Measurements of Corneal Sublayer Thickness in Normal and Post-SMILE Eyes.

PURPOSE: To test the reliability of a novel algorithm for measuring corneal epithelial thickness (ET) and stromal thickness of normal eyes and post-small incision lenticule extraction (SMILE) corneas with spectral-domain optical coherence tomography. METHODS: In this prospective observational study, a customized semiautomated software algorithm was developed and applied to measure corneal ET and stromal thickness along the horizontal corneal meridian. Measurements were performed by 2 examiners in a randomized fashion on a sample of 40 eyes with previous SMILE for treatment of myopia and a control group composed of 40 normal eyes. The intrauser repeatability and interuser reproducibility were analyzed by calculating typical indices including the coefficient of variation and intraclass correlation coefficient. Corneal sublayer thickness profiles were compared between normal and post-SMILE eyes. RESULTS: In both groups, coefficients of variation were 3.2% or lower and intraclass correlation coefficients were 0.929 or higher indicating excellent reliability of the measurement method. Central ET was on an average 6 µm greater in post-SMILE corneas (58.8 ± 5.4 µm) compared with normal eyes (52.8 ± 4.0 µm), with P < 0.01. Also, there was greater interindividual variability in ET in post-SMILE corneas and their horizontal epithelial profile seemed to show a lenticular appearance. CONCLUSIONS: Highly favorable indices of measurement reliability were achieved for this novel method of measuring corneal sublayer pachymetry not only in normal eyes but also in eyes with previous SMILE. The corneal ET profile was significantly altered in post-SMILE eyes compared with normal corneas.