Deep Learning Models Used in the Diagnostic Workup of Keratoconus: A Systematic Review and Exploratory Meta-Analysis

PURPOSE

The prevalence of keratoconus in the general population is reported to be up to 1 of 84. Over the past 2 decades, diagnosis and management evolved rapidly, but keratoconus screening in clinical practice is still challenging and asks for improving the accuracy of keratoconus detection. Deep learning (DL) offers considerable promise for improving the accuracy and speed of medical imaging interpretation. We establish an inventory of studies conducted with DL algorithms that have attempted to diagnose keratoconus.

METHODS

This systematic review was conducted according to the recommendations of the PRISMA statement. We searched (Pre-)MEDLINE, Embase, Science Citation Index, Conference Proceedings Citation Index, arXiv document server, and Google Scholar from inception to February 18, 2022. We included studies that evaluated the performance of DL algorithms in the diagnosis of keratoconus. The main outcome was diagnostic performance measured as sensitivity and specificity, and the methodological quality of the included studies was assessed using QUADAS-2.

RESULTS

Searches retrieved 4100 nonduplicate records, and we included 19 studies in the qualitative synthesis and 10 studies in the exploratory meta-analysis. The overall study quality was limited because of poor reporting of patient selection and the use of inadequate reference standards. We found a pooled sensitivity of 97.5% (95% confidence interval, 93.6%-99.0%) and a pooled specificity of 97.2% (95% confidence interval, 85.7%-99.5%) for topography images as input.

CONCLUSIONS

Our systematic review found that the overall diagnostic performance of DL models to detect keratoconus was good, but the methodological quality of included studies was modest.

A fluorophotometric comparison of anterior chamber riboflavin delivery via corneal application and fornix instillation

INTRODUCTION AND OBJECTIVES

We investigated whether riboflavin self-administration by patients could be a feasible option for corneal cross-linking, given the considerable resources required to impregnate the cornea with riboflavin. We analysed whether administering riboflavin in the inferior fornix (the site of self-administration) results in non-inferior riboflavin concentrations as when applied directly on the cornea (the site of administration by medical personnel).

PATIENTS AND METHODS

We conducted a prospective study to evaluate riboflavin concentrations at six time-points (baseline, 5, 15, 30, 45 and 60min) in 18 healthy volunteers for each of two application sites: cornea and fornix. Anterior chamber riboflavin (Peschke® TE 0.25%) concentrations were measured by fluorophotometry (Fluorotron™ Master FM-2).

RESULTS

For the two application sites cornea and fornix, participants did not differ in terms of age and sex. At baseline, the autofluorescence in the anterior chamber was 16.7ng/mL (SD 5.5) and 14.6ng/mL (SD 4.6) (P=.221). After 30min, anterior chamber fluorescein concentrations had risen to 55.1ng/mL (SD 25.5) and 46.1ng/mL (SD 25.1) (P=.293) without a further relevant increase by 60min.

CONCLUSIONS

This study found that applying riboflavin drops in the inferior fornix was non-inferior to applying it directly to the cornea, based on fluorophotometric measurements of anterior chamber riboflavin concentrations. This suggests that self-application of riboflavin is feasible in terms of corneal riboflavin impregnation.

Differences in minimal disease knowledge of keratoconus patients: results from an international survey

BACKGROUND/AIMS

The objective of this multicentre, multinational, prospective study was to assess the level of basic understanding that individuals with keratoconus possessed about their condition.

METHODS

We recruited 200 active keratoconus patients who were under regular review, and cornea specialists established a standard of 'minimal keratoconus knowledge' (MKK) that included an understanding of the definition, risk factors, symptoms and treatment options for the condition. We collected data from each participant regarding their clinical characteristics, highest level of education, (para)medical background and experiences with keratoconus within their social circle, and calculated the percentage of MKK attained by each patient.

RESULTS

Our findings revealed that none of the participants met the MKK standard, with the average MKK score being 34.6% and ranging from 0.0% to 94.4%. Furthermore, our study showed that patients with a university degree, previous surgical intervention for keratoconus or affected parents had a higher MKK. However, age, gender, disease severity, paramedical knowledge, disease duration and best-corrected visual acuity did not significantly affect the MKK score.

CONCLUSIONS

Our study demonstrates a concerning lack of basic disease knowledge among keratoconus patients in three different countries. The level of knowledge exhibited by our sample was only one-third of what cornea specialists would typically anticipate from patients. This highlights the need for greater education and awareness campaigns surrounding keratoconus. Further research is needed to determine the most efficient approaches for enhancing MKK and subsequently improving the management and treatment of keratoconus.

Agreement of IOL power calculation between IOLMaster 700 and Anterion swept source optical coherence tomography-based biometers

PURPOSE

To assess agreement of measurements by two swept source optical coherence tomography (SS-OCT) biometers and to evaluate the prediction error (PE) in intraocular lens (IOL) power calculation with seven formulas.

SETTING

Tertiary public eye hospital.

DESIGN

Consecutive observational.

METHODS

Axial length (AL), keratometry (K), anterior chamber depth (ACD), lens thickness (LT), and corneal diameter (CD) were measured with IOL Master 700 (Biometer A) and Anterion (Biometer B). Agreement was quantified by the limits of agreement (LoAs) and concordance correlation coefficient (CCC). The PE, the median absolute error (MedAE), and the mean absolute error (MAE) of the Barrett Universal II (BUII), EVO 2.0, Haigis, Hoffer Q, Holladay 1, Kane, and SRK/T formulas was inverstigated after constant optimization.

RESULTS

In 78 eyes from 78 patients, excellent agreement was obtained for AL (CCC >0.99), very good agreement for K, ACD, and LT (CCC >0.95), and strong agreement for CD (CCC>0.72). An additive offset of 0.07 mm was measured for ACD and LT whose mean values were higher with Biometer B (p<0.001). No statistically significant difference was found between the PEs and their absolute values when comparing the results of each formula between the two biometers.

CONCLUSION

Agreement of biometric measurements by the two biometers was high, although Biometer B provided higher mean values of ACD and LT by 0.07 mm.In cataract patients with normal eye length, measurements by the two biometers do not lead to different refractive outcomes with the seven formulas investigated.

Multifocal Femto-PresbyLASIK in Pseudophakic Eyes

BACKGROUND

Presbyopia treatment in pseudophakic patients with a monofocal IOL is challenging. This study investigates the refractive results of femto-PresbyLASIK and analyzes presbyopia treatment in pseudophakic eyes.

METHODS

14 patients with 28 pseudophakic eyes were treated with femto-PresbyLASIK. The dominant eye was targeted at a distance and the non-dominant eye at -0.5 D. The presbyopic algorithm creates a steepness in the cornea center by using an excimer laser that leads to corneal multifocality.

RESULTS

6 months after surgery a refraction of -0.11 ± 0.13 D (p = 0.001), an uncorrected distance visual acuity of 0.05 ± 1.0 logMAR (p < 0.001) and an uncorrected near visual acuity of 0.15 ± 0.89 logMAR (p = 0.001) were achieved in the dominant eye. For the non-dominant eye, the refraction was -0.28 ± 0.22 D (p = 0.002), the uncorrected distance of visual acuity was 0.1 ± 1.49 logMAR, and the uncorrected near visual acuity was 0.11 ± 0.80 logMAR (p < 0.001). Spherical aberrations (Z400) were reduced by 0.21-0.3 µm in 32% of eyes, and by 0.31-0.4 µm in 26% of eyes.

CONCLUSION

By steepening the central cornea while maintaining spherical aberrations within acceptable limits, PresbyLASIK created a corneal multifocality that safely improved near vision in both eyes. Thus, femto-PresbyLASIK can be used to treat presbyopia in pseudophakic eyes without performing intraocular surgery.

Mismatch of corneal specialists' expectations and keratoconus knowledge in general ophthalmologists - a prospective observational study in Switzerland

BACKGROUND

To assess whether Swiss general ophthalmologists have the minimal keratoconus knowledge that corneal specialists would expect them to have.

METHODS

Corneal specialists defined "minimal keratoconus knowledge" (MKK) with respect to definition, risk factors, symptoms and possible treatment options of keratoconus. A telephone interview survey was conducted among one hundred ophthalmologists (mean age 51.9 years (SD 9.5), 60 % male) from the German-speaking part of Switzerland. For each participant, years of work experience, number of keratoconus patients seen per year and access to a topography device were obtained. We calculated the proportion of MKK and examined in multivariate analyses whether ophthalmologists with access to topography and with greater work experience performed better than other groups.

RESULTS

No single ophthalmologist had MKK. The mean MKK was 52.0 %, and the range was 28.6-81.0 %. Per 10 years of working in private practice, the MKK decreased by 8.1 % points (95 % CI: -14.2, -2.00; p = 0.01). Only 24 % of participants correctly recalled the definition of keratoconus, 9 % all risk factors, 5 % all symptoms and 20 % all treatment modalities. The MKK values were not associated with the number of keratoconus patients seen per year and the availability of topography to diagnose keratoconus.

CONCLUSIONS

There is a substantial mismatch between corneal specialist' expectations and general ophthalmologists' knowledge about keratoconus. The low recall of symptoms and risk factors may explain why ophthalmologists diagnose relatively few cases of keratoconus, resulting in inefficient care delivery and delayed intervention.

Don't be afraid of the dark - OCT angiography through a black intraocular lens

Purpose

To report a case in which optical coherence tomography (OCT) and OCT angiography (OCT-A) allowed imaging of the posterior pole in a patient fitted with a black occlusive intraocular lens (IOL).

Observations

Following retinal central artery occlusion, a 52-year-old patient suffered from disturbing residual light perception. Occlusive contact lenses blocked the light insufficiently, so that the patient had to rely on an eye patch for relief of symptoms. After no neovascularization had formed during an observation period of 12 months, a black IOL (Morcher 85F) was implanted, blocking wavelengths in the visible spectrum but allowing transmission in the near-infrared spectrum. Slit lamp photography, OCT and OCT-A were performed pre- and postoperatively. Postoperatively, slit lamp photography could no longer provide images of the posterior pole, proving the effective blockade of wavelengths in the visible light spectrum. In contrast, transmission in the near-infrared spectrum allowed for OCT and OCT-A imaging of the fundus. The complete suppression of the disturbing perception of light succeeded only temporarily.

Conclusions and Importance

The implantation of a black IOL does not prevent the imaging of the retinal microvasculature by OCT-A. Black IOLs can therefore be considered even if continued monitoring of the vascular situation of the posterior pole is required.

Changes in higher order aberrations after central corneal regularization - a comparative two-year analysis of a semi-automated topography-guided photorefractive keratectomy combined with corneal cross-linking

Background

The optical quality in progressive keratoconus deteriorates due to ectasia and distortion of the corneal shape and optics. While corneal cross-linking (CXL) aims at stopping disease progression, “CXL-Plus” combines CXL with excimer laser ablation to improve visual function. Central Corneal Regularization (CCR) represents a therapeutic excimer laser modality specifically designed to smoothen the ectatic corneal shape and to reduce higher order aberrations (HOA). We set out to compare CXL-Plus, consisting of CXL combined with CCR, with CXL by itself for patients with progressive keratoconus.

Methods

Retrospective 2-year matched group analysis of patients who either underwent CXL-Plus (n = 28) or CXL as a sole procedure (n = 28) for progressive keratoconus. Main outcome parameters were HOA, visual function and tomographic results 12 and 24 months postoperatively.

Results

After 12 months, the total HOA root mean square wavefront error was reduced from 0.79 ± 0.30 to 0.40 ± 0.19 μm (CXL-Plus; p <  0.0001) and changed from 0.71 ± 0.28 to 0.73 ± 0.36 μm (CXL; p = 0.814). Uncorrected distance visual acuity improved from 0.70 ± 0.35 to 0.36 ± 0.29 logMAR (CXL-Plus; p = 0.0002) and from 0.65 ± 0.39 to 0.46 ± 0.37 logMAR (CXL; p = 0.067), translating to gains of three or more lines in 50% (CXL-Plus) and 36% (CXL) of patients. The steepest keratometry value (Kmax) regressed by 5.84 D (CXL-Plus; p <  0.0001) and 0.66 D (CXL; p = 0.752). For none of the investigated parameters a statistically significant change could be shown between 12 and 24 months.

Conclusions

CXL-Plus in the form of a CCR reduces HOA and Kmax more effectively than CXL as a sole procedure.

Fluorophotometric Determination of Riboflavin Concentrations in a Human Artificial Anterior Chamber Model

PURPOSE

Corneal cross-linking (CXL) requires an adequate corneal riboflavin impregnation, which is clinically assessed by verification of a riboflavin "flare" in the anterior chamber. We set out to replace this subjective assessment with an objective measurement method and evaluated fluorophotometry as an apparatus-based technique for riboflavin detection in the anterior chamber.

METHODS

In an artificial anterior chamber model using human corneas and a modified Fluorotron fluorophotometer, we determined the detection limits of riboflavin concentrations across native corneas by comparison measurements of the same concentrations in glass cuvettes. Subsequently, standard CXL procedures with corneal application of riboflavin were simulated and the proportions of riboflavin entering the anterior chamber were measured fluorophotometrically.

RESULTS

The measurement results of the riboflavin dilution series in the artificial anterior chamber showed a very high concordance with the results obtained in a glass cuvette (Pitman test P = 0.329). In the CXL simulation, the mean riboflavin concentration measured in the anterior chamber increased within 15 minutes from 5 (±1) to 903 (±204) ng/mL and stood at 1089 (±56) ng/mL after 30 minutes.

CONCLUSIONS

Fluorophotometry is able to measure riboflavin in an artificial anterior chamber across human corneas over a wide range of concentrations and it reliably detects the increasing riboflavin signal in simulated CXL procedures.

TRANSLATIONAL RELEVANCE

The replacement of the subjective riboflavin detection by a technically straightforward, objective detection method might increase patient safety and treatment efficiency in CXL.

Effects and adverse events after CXL for keratoconus are independent of age: a 1-year follow-up study

PURPOSE

To compare the effect, failure rate and the risks of corneal cross-linking (CXL) in keratoconus patients aged ≥35 years to patients <35 years.

METHODS

In 141 eyes of 116 keratoconus patients we compared the changes in best phoropter-corrected visual acuity (BCVA) and maximum keratometry values (Kmax) before and 12 months after CLX in patients aged ≥35 years (n=34, 38 eyes) to the cohort of patients below 35 years of age.

RESULTS

Overall, CXL significantly improved BCVA from 0.487 logMAR (95% confidence interval (CI) 0.426-0.548) by -0.197 logMAR (95% CI -0.243 to -0.150; P<0.001) and reduced Kmax from 48.96 diopter (Dpt) by -1.33 Dpt (95% CI -1.85 to -0.81: P<0.001). Age ≥35 years had no effect on the changes of BCVA (-0.02 (95% CI -0.13 to 0.09); P=0.757) or Kmax (0.58 (95%CI -0.51 to 1.68); P=0.294) as compared with younger patients. In 54 patients (55 eyes, 38.5%) aged <35 years and in 18 patients (18 eyes, 47.4%) aged ≥35 years, BCVA increased by ≥2 Snellen lines. Failure (increase in Kmax ≥1 Dpt) was observed in 17 eyes (16.5%) of patients aged <35 years and in 3 eyes (7.9%) of patients aged ≥35 years during the 12-month follow-up period. Adverse outcomes (loss of ≥2 Snellen lines) occurred in 4 (3.9%) eyes of patients aged <35 years and 1 (2.6%) eye of a patient aged ≥35 years.

CONCLUSION

Effects and adverse events of CXL treatment do not seem to differ between subjects younger or older than 35 years.

Superiority of small islets in human islet transplantation

Many factors influence the outcome of islet transplantation. As islets in the early posttransplant setting are supplied with oxygen by diffusion only and are in a hypoxic state in the portal system, we tested whether small human islets are superior to large islets both in vitro and in vivo. We assessed insulin secretion of large and small islets and quantified cell death during hypoxic conditions simulating the intraportal transplant environment. In the clinical setting, we analyzed the influence of transplanted islet size on insulin production in patients with type 1 diabetes. Our results provide evidence that small islets are superior to large islets with regard to in vitro insulin secretion and show a higher survival rate during both normoxic and hypoxic culture. Islet volume after 48 h of hypoxic culture decreased to 25% compared with normoxic culture at 24 h due to a preferential loss of large islets. In human islet transplantation, the isolation index (islet volume as expressed in islet equivalents/islet number), or more simply the islet number, proved to be more reliable to predict stimulated C-peptide response compared with islet volume. Thus, islet size seems to be a key factor determining human islet transplantation outcome.