Corneal Topography Raw Data Classification Using a Convolutional Neural Network.

PURPOSE: We investigated the efficiency of a convolutional neural network applied to corneal topography raw data to classify examinations of 3 categories: normal, keratoconus (KC), and history of refractive surgery (RS). DESIGN: Retrospective machine-learning experimental study. METHODS: A total of 3,000 Orbscan examinations (1,000 of each class) of different patients of our institution were selected for model training and validation. One hundred examinations of each class were randomly assigned to the test set. For each examination, the raw numerical data from "elevation against the anterior best fit sphere (BFS)," "elevation against the posterior BFS" "axial anterior curvature," and "pachymetry" maps were used. Each map was a square matrix of 2,500 values. The 4 maps were stacked and used as if they were 4 channels of a single image.A convolutional neural network was built and trained on the training set. Classification accuracy and class wise sensitivity and specificity were calculated for the validation set. RESULTS: Overall classification accuracy of the validation set (n = 300) was 99.3% (98.3%-100%). Sensitivity and specificity were, respectively, 100% and 100% for KC, 100% and 99% (94.9%-100%) for normal examinations, and 98% (97.4%-100%) and 100% for RS examinations. CONCLUSION: Using combined corneal topography raw data with a convolutional neural network is an effective way to classify examinations and probably the most thorough way to automatically analyze corneal topography. It should be considered for other routine tasks performed on corneal topography, such as refractive surgery screening.

Corneal topography patterns in the Tehran eye study: warning about the high prevalence of patterns with a skewed radial axis.

PURPOSE: The purpose of this study is to determine the distribution of corneal topography patterns in Tehran. MATERIALS AND METHODS: In this population-based study, a total of 442 individuals were randomly selected by cluster sampling for complete ophthalmologic examination. A total of 404 (788 eyes) met the inclusion criteria and were enrolled in this study. Orbscan II (Bausch and Lomb Surgical, Salt Lake City, USA) was used to determine the anterior segment indices and axial power maps for each eye. RESULTS: On the basis of the axial power maps, the symmetric bowtie (SB) (29.0%) and asymmetric bowtie with inferior steepening (16.7%) patterns were the most prevalent and the irregular (3.3%) and superior steepening (1.5%) patterns the least prevalent. Asymmetric bowtie pattern with a skewed radial axis (AB-SRAX) was seen in 7.6% of eyes. These cases had both thinner and steeper corneas compared with round and SB ones (P < 0.014 and P < 0.006). Maximum anterior elevation in 5 mm zone, of AB-SRAX corneas were higher than other patterns (P < 0.01) except for superior steepened and inferior steepened ones. CONCLUSIONS: The distribution of corneal topographic patterns in Iranians seems like other Asian population on the whole. The 7.6% prevalence of AB-SRAX patterns may be a warning on keratoconus prevalence in our population. This is in line with recent regional reports.

Characteristics of corneal topographic and pachymetric patterns in patients with pellucid marginal corneal degeneration.

PURPOSE: To determine the characteristics of the shape of the cornea in patients with pellucid marginal corneal degeneration (PMD) and to compare these characteristics to those of eyes with keratoconus and eyes of normal subjects. METHODS: This was a retrospective, cross-sectional case-series in which 49 eyes of 33 patients with PMD, 51 eyes of 51 patients with keratoconus and 53 eyes of 53 subjects with normal corneas (controls) were examined and compared. For all eyes, we obtained the topographic patterns of the axial power maps, anterior and posterior elevation maps and pachymetric maps using a rotating Scheimpflug camera. The eyes were classified into the respective patterns by visual inspection of these maps. RESULTS: In eyes with PMD, the most common axial power map pattern was the crab claw pattern (78 %) followed by the inferior steepening pattern (18 %). In eyes with keratoconus, the most common pattern was the inferior steepening pattern (67 %). The most common pattern in the elevation maps for both surfaces was the asymmetric island in eyes with PMD and keratoconus. Although the decentered pattern, including the decentered oval (27 %) and decentered round (20 %) pattern, on pachymetric map was specific to eyes with PMD, the incidence of these patterns was relatively low. CONCLUSIONS: The similarity in the topographic and pachymetric patterns in eyes with PMD and keratoconus suggests that they may be a continuity of the same disorder with different phenotypes.

Variability of subjective classifications of corneal topography maps from LASIK candidates.

PURPOSE: To evaluate the variability of subjective corneal topography map classification between different experienced examiners and the impact of changing from an absolute to a normative scale on the classifications. METHODS: Preoperative axial curvature maps using Scheimpflug imaging obtained with the Pentacam HR (Oculus Optikgeräte, Wetzlar, Germany) and clinical parameters were sent to 11 corneal topography specialists for subjective classification according to the Ectasia Risk Scoring System. The study population included two groups: 11 eyes that developed ectasia after LASIK and 14 eyes that had successful and stable LASIK outcomes. Each case was first reviewed using the absolute scale masked to the patient group. After 3 months, the same cases were represented using a normative scale and reviewed again by the same examiners for new classifications masked to the patient group. RESULTS: Using the absolute scale, 17 of 25 (68%) cases had variations on the classifications from 0 to 4 for the same eye across examiners, and the overall agreement with the mode was 60%. Using the normative scale, the classifications from 11 of 25 (44%) cases varied from 0 to 4 for the same eye across examiners, and the overall agreement with the mode was 61%. Eight examiners (73%) reported statistically higher scores (P < .05) when using the normative scale. Considering all 550 topographic analyses (25 cases, 11 examiners, and two scales), the same classification from the two scales was reported for 121 case pairs (44%). CONCLUSION: There was significant inter-observer variability in the subjective classifications using the same scale, and significant intra-observer variability between scales. Changing from an absolute to a normative scale increased the scores on the classifications by the same examiner, but significant inter-observer variability in the subjective interpretation of the maps still persisted.

Incidence and characteristics of unilateral keratoconus classified on corneal topography.

PURPOSE: To evaluate the characteristics of unilateral keratoconus defined on the basis of corneal topography and analyze videokeratography parameters between fellow eyes and normal controls. METHODS: A total of 111 patients with clinical keratoconus were prospectively enrolled. Both eyes were evaluated with Tomey (Tomey Corp) and Orbscan II (Bausch & Lomb) corneal topography systems. The patient was classified as having unilateral keratoconus if one eye had clinical keratoconus and the other eye did not have any topographic signs of keratoconus such as asymmetric videokeratographic pattern, positive result in Tomey keratoconus screening, maximum posterior elevation >40 µm, or corneal thinnest pachymetry <500 µm. Clinical characteristics and 13 Orbscan II quantitative indices between keratoconic and fellow eyes and normal control eyes were evaluated. RESULTS: Five (4.5%) of 111 patients with keratoconus had no topographic evidence of keratoconus in the fellow eye. All clinically normal fellow eyes had symmetric bowtie patterns. Statistically significant differences were noted in maximum posterior elevation, corneal irregularity, and corneal thinnest values between keratoconic eyes and fellow eyes, and between keratoconic eyes and control eyes. Only 3-mm irregularity was significantly higher in the fellow eyes compared with control eyes (P<.05). An increased trend for corneal 5-mm irregularity was found in fellow eyes compared to control eyes. CONCLUSIONS: The incidence of unilateral keratoconus was 4.5%. A trend of higher irregularity was found in fellow eyes compared with control eyes. This finding indicates that fellow eyes may show a certain low-expressivity morphologic feature of keratoconus.

Relación entre los componentes ópticos oculares e implicaciones en el proceso de emetropización / The realtionships between ocular optical components and implication sin the process of emmetropization

Objetivo: Determinar la relación entre los diferentes componentes ópticos oculares y la emetropización en función de la longitud axial (LA). Métodos: Participaron 109 jóvenes universitarios divididos en cinco grupos según el error refractivo: emétropes, hipermétropes, miopes bajos, miopes moderados y miopes altos. La medida de los parámetros intraoculares y el análisis de la topografía corneal se realizó mediante ultrasonografía y videoqueratoscopia respectivamente. Resultados: Se encontró una correlación estadísticamente significativa entre la profundidad de la cámara anterior (PCA) y la LA en ojos con LA menor de 24 mm (r = 0,441; p < 0,001). Sin embargo en ojos con LA mayor no se encontró tal relación (r = 0,098; p = 0,527). El espesor del cristalino (EC) mostró una correlación inversa con la LA para los ojos con LA < 24 mm (r = 0,391; p < 0,001), siendo nula en el caso de ojos con LA ≥ 24 mm. La curvatura corneal (RC) sólo mostró una débil correlación con la profundidad de la cámara vítrea (PCV) en ojos con LA < 24 mm (r = 0,363; p < 0,003). Conclusión: Los parámetros ópticos del ojo humano se relacionan de distinta forma según el tamaño del ojo, encontrándose en los ojos más miopes una ausencia de coordinación entre los componentes ópticos oculares, que según los resultados de este estudio, es a partir de 2,00 D

Purpose: To report the relationship between different optical ocular components and the influence of axial length on emmetropization. Methods: 109 young university students, divided into five groups, were enrolled in this study: emmetropes, hyperopes, low myopes, moderate myopes and high myopes. Intraocular parameters and topographic corneal analyses were performed by ultrasonography and videokeratoscopy respectively. Results: Anterior chamber depth and axial length were found to correlate significantly in eyes with axial lengths less than 24 mm (r = 0.441; p < 0.001) However this correlation was not found in eyes with longer axial lengths (r = 0.098; p = 0.527). Lens thickness showed an inverse correlation with axial length for shorter eyes (r = 0.391; p < 0.001), whereas any correlation in longer eyes was associated with moderate to high levels of myopia. Anterior corneal curvature only correlated, although weakly, with vitreous chamber depth for shorter eyes (r = 0.363; p < 0.003). Conclusion: Differences in optical parameters of the human eye displayed different associations in longer eyes. A lack of correlation was seen between ocular components in eyes with higher myopia, corresponding to myopia in excess of 2.00 D (Arch Soc Esp Oftalmol 2008; 83: 307-316)

Topographic patterns in refractive surgery candidates.

PURPOSE: To evaluate the prevalence of different topographic patterns in refractive surgery candidates and to estimate the extent of candidate rejection based on topography alone. METHODS: We performed a retrospective evaluation of videokeratographies of previously unoperated refractive surgery candidates. The defined topographic patterns of their corneas and the indices computed by the videokeratographer software were registered. The data were evaluated by SPSS/PC statistical software. RESULTS: One hundred candidate electronic topography records were evaluated. The candidates included 41 women and 59 men whose average age was 32 years (range 17.5-63.5). Their topographic patterns were: spherical (36/200 evaluated eyes), spherocylindrical (60), upper steep (32), lower steep (43), irregular astigmatism (9), decentered (3), suspected keratoconus (11), and probable keratoconus (6 eyes). Rejection from undergoing the procedure was based on topography alone in 27 of the 200 eyes. It was difficult to decide what to do with the 43 eyes with a lower steep pattern, and they were eventually rejected. Of the nine software-calculated indices, only one, "keratoconus index," could distinguish between the spherocylindrical and the lower steep patterns. CONCLUSIONS: More than one-half of corneal topographies of refractive surgery candidates did not comply with the assumed "normal" spherical or spherocylindrical patterns. The possible continuum of keratoconus-suspected keratoconus-lower steep pattern raises the question of where to draw the line between reasonable and risky when considering corneal refractive surgery.

Videokeratography after excimer laser in situ keratomileusis for myopia.

OBJECTIVE: The purpose of this study was to define qualitative patterns of videokeratography after excimer laser in situ keratomileusis (LASIK) for myopia and to identify associations of videokeratography patterns with clinical variables. DESIGN: Single-center, retrospective, cohort study. PARTICIPANTS: One hundred twenty-six eyes of 68 patients with myopia who desired surgical correction of myopia ranging from -1.50 to -8.80 diopters. INTERVENTION: Myopia was corrected with LASIK using a Summit OmniMed/Apex excimer laser. Computer-assisted videokeratography data were acquired at 3 months after primary LASIK, and associations of videokeratography pattern with clinical variable were assessed. MAIN OUTCOME MEASURES: Seven subgroup ablation zone patterns were characterized using tangential curvature maps; these subgroups were combined into pooled symmetric and asymmetric groups. Associations with age, attempted correction, single-zone or multizone ablation, postoperative uncorrected and spectacle-corrected visual acuity, predictability, astigmatism, corneal uniformity, glare disability, and contrast sensitivity were analyzed. RESULTS: Thirty-two eyes (25.4%) showed a uniform pattern, 14 eyes (11.1%) showed a bow-tie pattern, 22 eyes (17.5%) showed a peninsula pattern, 22 eyes (17.5%) showed a semicircular pattern, 22 eyes (17.5%) showed an off-center blue spot pattern, 10 eyes (7.9%) showed an irregular pattern, and 4 eyes (3.2%) showed a central island pattern. Spectacle-corrected visual acuity, astigmatism, corneal uniformity, glare disability, and contrast sensitivity were significantly related to videokeratography pattern, although the difference among videokeratography groups was clinically small. Age, attempted correction, single- or multizone ablation, and postoperative uncorrected visual acuity and predictability were not predictive of videokeratography pattern. CONCLUSIONS: Most eyes in this study had an asymmetric videokeratography pattern 3 months after LASIK. Small but statistically significant differences were identified among the videokeratography patterns that potentially may affect postoperative quality of vision.

Proposed classification for topographic patterns seen after penetrating keratoplasty.

AIMS: To create a clinically useful classification for post-keratoplasty corneas based on corneal topography. METHODS: A total of 360 topographic maps obtained with the TMS-1, from 95 eyes that had undergone penetrating keratoplasty (PKP), were reviewed independently by two examiners in a masked fashion, and were categorised according to a proposed classification scheme. RESULTS: A high interobserver agreement (88% in the first categorisation) was achieved. At 12 months post-PKP, a regular astigmatic pattern was observed in 20/85 cases (24%). This was subclassified as oval in three cases (4%), oblate symmetric bow tie in six cases (7%), prolate asymmetric bow tie in six cases (7%), and oblate asymmetric bow tie in five cases (6%). An irregular astigmatic pattern was observed in 61/85 cases (72%), subclassified as prolate irregular in five cases (6%), oblate irregular in four cases (5%), mixed in seven cases (8%), steep/flat in 11 cases (13%), localised steepness in 16 cases (19%), and triple pattern in three cases (4%). Regular astigmatic patterns were associated with significantly higher astigmatism measurements. The surface asymmetry index was significantly lower in the regular astigmatic patterns. CONCLUSIONS: In post-PKP corneas, the prevalence of irregular astigmatism is about double that of regular astigmatism, with a trend for increase of the irregular patterns over time.

Corneal topography classification in myopic eyes based on axial, instantaneous, refractive, and profile difference maps.

PURPOSE: To refine and develop systems for classifying the topography of myopic corneas using axial, instantaneous, refractive, and profile difference maps. SETTING: Baylor College of Medicine, Cullen Eye Institute Houston, Texas, USA. METHODS: Using the EyeSys Corneal Analysis System, computerized videokeratographs of 153 corneas of 78 myopic patients were retrospectively analyzed. Patterns were defined with respect to the mid-dioptric green color. Relationships among pattern types, refractive power, corneal power, corneal uniformity index (CU index), and predicted corneal visual acuity (PC acuity) were calculated. RESULTS: Six types of patterns for axial, instantaneous, and refractive maps and 3 types of patterns for the profile difference map were defined. For a given cornea, there was a weak correlation among the patterns in the axial, instantaneous, and profile difference maps. The circular with central irregularity pattern in auto-scale refractive maps and the irregular pattern in profile difference maps and axial maps were correlated with lower CU index and PC acuity values. Clinical classification of instantaneous maps did not contribute substantially to the information provided by axial maps. CONCLUSION: Patterns of computerized videokeratographs varied according to the type of topographic map. The classification systems used provide a baseline for analyzing the response of the cornea to various interventions.

Estudo videoceratoscópico no pré e pós-operatório da cirurgia refrativa incisional / Computerized corneal topography and radial keratotomy: a clinical analysis

Objetivos: o objetivo deste estudo foi analisar a eficácia da videoceratoscopia computadorizada na detecçäo de ectasias corneanas nos candidatos à ceratotomia radial, avaliar a reprodutibilidade na análise destes mapas quando analisamos por diferentes examinadores e associar padröes topográficos pré-operatórios a padröes topográficos pós-operatórios. Métodos: foram analisados, através da videoceratoscopia computadorizada, 66 olhos submetidos a ceratotmoia radial no Departamento de Oftalmologia da UNIFESP-EPM no período entre maio de 1993 a setembro de 1994. A classificaçäo dos mapas pré e pós-operatórios foi realizada por 3 examinadores independentes. Resultados: os mapas topográficos do pós-operatório mantiveram-se estáveis ou evouiram para um padräo ..