Detection Of Keratoconus Research Articles

Keratoconus: exploring fundamentals and future perspectives - a comprehensive systematic review.

New developments in artificial intelligence, particularly with promising results in early detection and management of keratoconus, have favorably altered the natural history of the disease over the last few decades. Features of artificial intelligence in different machine such as anterior segment optical coherence tomography, and femtosecond laser technique have improved safety, precision, effectiveness, and predictability of treatment modalities of keratoconus (from contact lenses to keratoplasty techniques). These options ingrained in artificial intelligence are already underway and allow ophthalmologist to approach disease in the most non-invasive way. This study comprehensively describes all of the treatment modalities of keratoconus considering machine learning strategies. A multidimensional comprehensive systematic narrative review. A comprehensive search was done in the five main electronic databases (PubMed, Scopus, Web of Science, Embase, and Cochrane), without language and time or type of study restrictions. Afterward, eligible articles were selected by screening the titles and abstracts based on main mesh keywords. For potentially eligible articles, the full text was also reviewed. Artificial intelligence demonstrates promise in keratoconus diagnosis and clinical management, spanning early detection (especially in subclinical cases), preoperative screening, postoperative ectasia prediction after keratorefractive surgery, and guiding surgical decisions. The majority of studies employed a solitary machine learning algorithm, whereas minor studies assessed multiple algorithms that evaluated the association of various keratoconus staging and management strategies. Last but not least, AI has proven effective in guiding the implantation of intracorneal ring segments in keratoconus corneas and predicting surgical outcomes. The efficient and widespread clinical translation of machine learning models in keratoconus management is a crucial goal of potential future approaches to better visual performance in keratoconus patients. The article has been registered through PROSPERO, an international database of prospectively registered systematic reviews, with the ID: CRD42022319338.

Deep learning in early keratoconus/corneal ectasia

Machine learning algorithms have been gaining tremendous relevance in the last decade in more or less all disciplines in medicine. In most cases supervised learning techniques are used to perform classification or graduation of diseases. Early detection of keratoconus in a screening environment is crucial as in this stage several therapeutic options such as collagen crosslinking or intracorneal ring implantation are available which could prevent further progression of the disease. In this talk we focus on new techniques of automatic identification of ectatic corneal diseases such as keratoconus in an early stage based on anterior segment optical coherence measurements with the Casia 2 tomographer. A neural network approach is implemented and fed with data of corneal front surface height and curvature, back surface height and curvature, and thickness data (in total 5 data layers) and the similarity of the measurement to normal or abnormal situation is provided. Performance data are shown for our training, validation, and test dataset.

Detecting Keratoconus in Adolescents with Anterior Segment Optical Coherence Tomography.

Assessing the applicability of an algorithm developed for keratoconus detection in adolescents. This algorithm relies on optical coherence tomography (OCT) and incorporates features related to corneal pachymetric and epithelial thickness alterations. We retrospectively reviewed charts of patients under the age of 18 and divided them into four groups according to the Belin-Ambrosio display (Pentacam): normal, manifest, and subclinical keratoconus, as well as very asymmetric eye with normal topography and tomography (VAE-NTT). Corneal and epithelial thickness maps (Cirrus 5000 HD-OCT, Carl Zeiss Meditec, Germany) were evaluated by a human grader. In the first step, if at least one of four parameters (pachymetry minimum (pachy min), pachy minimum-median (min-med), pachy superonasal-inferotemporal (SN-IT), or epithelial (epi SN-IT)) exceeded its cut-off value, the eye was considered as suspect. In the second step, the combined presence of coincident thinning of total cornea and epithelium as well as concentric epithelial thinning lead to the diagnosis of keratoconus. Receiver operating characteristic (ROC) curves were generated to determine area under the curve (AUC), sensitivity, and specificity for the parameters. The study involved 19 pediatric patients diagnosed with keratoconus, comprising 29 manifest keratoconic eyes, 3 eyes with subclinical keratoconus, and 5 VAE-NTT eyes. In addition, 22 eyes from 11 normal adolescents were included in the analysis. The AUC values of parameters in step 1 were 0.889 for pachy min, 0.997 for pachy min-med, 0.893 for pachy SN-IT, and 0.998 for epi SN-IT. When both steps were performed, this algorithm captured all manifest and subclinical pediatric keratoconic eyes. When all eyes of the keratoconus patients were combined, step 1 had 97.3% sensitivity and step 2 had 100% specificity. Using this OCT-based approach in adolescents yielded a high level of agreement with the current gold standard, tomography. Using them together, potentially also with other examinations may improve the diagnostic accuracy of KC in the pediatric population. Integration of this approach into the software of the device to facilitate automated evaluations is desired.

Artificial intelligence in cataract and refractive surgery

Artificial intelligence (AI) technology is proving to be promising in the field of healthcare, especially in the areas relying on big data and image‐based analysis. In Ophthalmology, AI was already tested in majority of subspecialities, including cataract and refractive surgery. There were interesting studies showing the benefits of using AI algorithms in early keratoconus detection, to predict the outcomes of laser refractive surgery, and to cataract surgery training and monitoring. Moreover, AI models using fundus images to diagnose and grade cataract have been developed. New intraocular lens (IOL) formulas are applying big data and computational methodologies to achieve better prediction of targeted refraction and simplify the IOL selection process. Recent developments in AI have shown their potential possibility for the diagnosis and management of paediatric cataract using slit‐lamp images.

Deep Learning Algorithm for Keratoconus Detection from Tomographic Maps and Corneal Biomechanics: A Diagnostic Study

Abstract Purpose: To develop an artificial intelligence (AI) approach for differentiating between normal cornea, subclinical, and keratoconus (KC) using tomographic maps from Pentacam (Oculus) and corneal biomechanics from Corvis ST (Oculus). Methods: A total of 1,668 tomographic (769 patients) and 611 biomechanical (307 patients) images from the Chula Refractive Surgery Center, King Chulalongkorn Memorial Hospital were included. The sample size was divided into the Pentacam and combined Pentacam-Corvis groups. Different convolutional neural network approaches were used to enhance the KC and subclinical KC detection performance. Results: AI model 1, which obtained refractive maps from Pentacam, achieved an area under the receiver operating characteristic curve (AUC) of 0.938 and accuracy of 0.947 (sensitivity, 90.8% and specificity, 96.9%). AI model 2, which added dynamic corneal response and the Vinciguerra screening report from Corvis ST to AI Model 1, achieved an AUC of 0.985 and accuracy of 0.956 (sensitivity, 93.0% and specificity, 94.3%). AI model 3, which added the corneal biomechanical index to AI Model 2, reached an AUC of 0.991 and accuracy of 0.956 (sensitivity, 93.0% and specificity, 94.3%). Conclusions: Our study showed that AI models using either anterior corneal curvature alone or combined with corneal biomechanics could help classify normal and keratoconic corneas, which would make diagnosis more accurate and would be helpful in decision-making for the treatment.

Revisiting Pentacam Parameters in the Diagnosis of Subclinical and Mild Keratoconus Based on Different Grading System Definitions

To retest the performance of Pentacam parameters in the detection of eyes with subclinical keratoconus (KC) and mild KC based on different definitions from the Amsler-Krumeich (AK), Collaborative Longitudinal Evaluation of Keratoconus (CLEK), and ABCD systems. This cross-sectional university-based study comprised 24 eyes with subclinical KC, 144 eyes with mild KC (based on AK in 101 eyes, CLEK in 28 eyes, and ABCD in 15 eyes), and 70 controls. Diagnostic ability of the thinnest point (TP) pachymetry, KISA% index, inferior-superior asymmetry, corneal aberrations, Pentacam indices, front/back elevations, pachymetric progression index, Ambrósio-Relational Thickness (ARTmax), and Belin/Ambrósio Enhanced Ectasia Display scores (Df, Db, Dp, Dt, Da, and D-final) were evaluated. ARTmax (83.3% sensitivity/74.3% specificity) had the highest ability in distinguishing subclinical KC from normal, followed by TP pachymetry, Dt, and Da. D-final showed excellent sensitivity/specificity in mild KC diagnosis based on AK (98%/100%) and CLEK (97.4%/100%) descriptions. In the mild KC-ABCD group, index of vertical asymmetry accurately detected all eyes with mild KC and 97.1% of the controls. This study points out the gray zone in the detection of eyes with subclinical and mild KC due to overlapping terminology and grading criteria. Pentacam parameters seem to have modest capability in subclinical KC detection, indicating the necessity for additional diagnostic modalities. However, eyes with mild KC can be diagnosed with high accuracy using Pentacam parameters, although the strongest parameters may vary according to the definition of “mild KC.” Nevertheless, uniform and definitive criteria for subclinical and clinical KC classification are required for a diagnostic and therapeutic consensus in KC.

Computer-aided diagnosis of keratoconus through VAE-augmented images using deep learning

Detecting clinical keratoconus (KCN) poses a challenging and time-consuming task. During the diagnostic process, ophthalmologists are required to review demographic and clinical ophthalmic examinations in order to make an accurate diagnosis. This study aims to develop and evaluate the accuracy of deep convolutional neural network (CNN) models for the detection of keratoconus (KCN) using corneal topographic maps. We retrospectively collected 1758 corneal images (978 normal and 780 keratoconus) from 1010 subjects of the KCN group with clinically evident keratoconus and the normal group with regular astigmatism. To expand the dataset, we developed a model using Variational Auto Encoder (VAE) to generate and augment images, resulting in a dataset of 4000 samples. Four deep learning models were used to extract and identify deep corneal features of original and synthesized images. We demonstrated that the utilization of synthesized images during training process increased classification performance. The overall average accuracy of the deep learning models ranged from 99% for VGG16 to 95% for EfficientNet-B0. All CNN models exhibited sensitivity and specificity above 0.94, with the VGG16 model achieving an AUC of 0.99. The customized CNN model achieved satisfactory results with an accuracy and AUC of 0.97 at a much faster processing speed compared to other models. In conclusion, the DL models showed high accuracy in screening for keratoconus based on corneal topography images. This is a development toward the potential clinical implementation of a more enhanced computer-aided diagnosis (CAD) system for KCN detection, which would aid ophthalmologists in validating the clinical decision and carrying out prompt and precise KCN treatment.

Diagnostic Validation of the Screening Corneal Objective Risk of Ectasia Analyzer Evaluated by Swept Source Optical Coherence Tomography for Keratoconus in an Asian Population.

We aimed to investigate the diagnostic accuracy of Screening Corneal Objective Risk of Ectasia (SCORE) Analyzer software using ANTERION, a swept-source optical coherence tomography device, for keratoconus diagnosis in an Asian population. A total of 151 eyes of 151 patients were included in this retrospective study as follows: 60, 45, and 46 keratoconus, keratoconus suspects, and normal control eyes, respectively. Parameters in the SCORE calculation, including six indices, were compared for the three groups. The receiver operating characteristic curve analysis and cut-off value were estimated to assess the diagnostic ability to differentiate keratoconus and keratoconus suspect eyes from the normal group. The SCORE value and six indices were significantly correlated-"AntK max" (R = 0.864), "AntK oppoK" (R = 0.866), "Ant inf supK" (R = 0.943), "Ant irre 3mm" (R = 0.741), "post elevation at the thinnest point" (R = 0.943), and "minimum corneal thickness" (R = -0.750). The SCORE value showed high explanatory power (98.1%), sensitivity of 81.9%, and specificity of 78.3% (cut-off value: 0.25) in diagnosing normal eyes from the keratoconus suspect and keratoconus eyes. The SCORE Analyzer was found to be valid and consistent, showing good sensitivity and specificity for keratoconus detection in an Asian population.

Deep learning models based on CNN architecture for early keratoconus detection using corneal topographic maps

Deep learning models based on CNN architecture for early keratoconus detection using corneal topographic maps

Advanced Corneal Imaging in Keratoconus: A Report by the American Academy of Ophthalmology

To review the published literature on the diagnostic capabilities of the newest generation of corneal imaging devices for the identification of keratoconus. Corneal imaging devices studied included tomographic platforms (Scheimpflug photography, OCT) and functional biomechanical devices (imaging an air impulse on the cornea). A literature search in the PubMed database for English language studies was last conducted in February 2023. The search yielded 469 citations, which were reviewed in abstract form. Of these, 147 were relevant to the assessment objectives and underwent full-text review. Forty-five articles met the criteria for inclusion and were assigned a level of evidence rating by the panel methodologist. Twenty-six articles were rated level II, and 19 articles were rated level III. There were no level I evidence studies of corneal imaging for the diagnosis of keratoconus found in the literature. To provide a common cross-study outcome measure, diagnostic sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were extracted. (A perfect diagnostic test that identifies all cases properly has an AUC of 1.0.) RESULTS: For the detection of keratoconus, sensitivities for all devices and parameters (e.g., anterior or posterior corneal curvature, corneal thickness) ranged from 65% to 100%. The majority of studies and parameters had sensitivities greater than 90%. The AUCs ranged from 0.82 to 1.00, with the majority greater than 0.90. Combined indices that integrated multiple parameters had an AUC in the mid-0.90 range. Keratoconus suspect detection performance was lower with AUCs ranging from 0.66 to 0.99, but most devices and parameters had sensitivities less than 90%. Modern corneal imaging devices provide improved characterization of the cornea and are accurate in detecting keratoconus with high AUCs ranging from 0.82 to 1.00. The detection of keratoconus suspects is less accurate with AUCs ranging from 0.66 to 0.99. Parameters based on single anatomic locations had a wide range of AUCs. Studies with combined indices using more data and parameters consistently reported high AUCs. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Enhanced morphological assessment based on interocular asymmetry analysis for keratoconus detection.

To clarify the interocular asymmetry of corneal morphological descriptors and evaluate its discriminant ability of keratoconus (KC). This retrospective study recruited 344 normal participants and 290 KC patients, randomized to training and validation datasets. Interocular correlation and agreement were evaluated on 44 corneal morphological descriptors derived from Schiempflug tomography. Logistic regression models were constructed using binocular data and of which diagnostic performance was evaluated using the area under receiver operating characteristics curve (AUC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Interocular agreement of corneal descriptors is better in the normal than in KC except for dimensions of cornea and anterior chamber. The interocular asymmetry increases along with the severity of KC. Interocular asymmetry in maximum anterior keratometry, mean anterior keratometry and higher-order aberrations of anterior surface show high AUC above 0.950. Binocular logistic regression index reaches an AUC of 0.963 with high specificity (95.2%) and brings gain to monocular parameters in distinguishing the normal eyes from KC (NRI = 0.080 (0.042 ~ 0.118), P < 0.001) and IDI = 0.071 (0.049 ~ 0.092), P < 0.001). Interocular asymmetry benefits even more in subclinical keratoconus (SKC) detection reflected by NRI (0.4784 (0.2703-0.6865), P < 0.001) and IDI (0.2680 (0.1495-0.3866), P < 0.001) measures. Interocular asymmetry is a well-characterized feature of KC and related to the severity. It is feasible to apply the interocular asymmetry in diagnosis of KC and SKC as a replenishment of monocular parameters and in progression tracking.

Keratoconus Diagnosis and Treatment: Recent Advances and Future Directions.

Keratoconus is a disorder characterized by progressive corneal thinning and steepening that may result in significant visual impairment secondary to high astigmatism, corneal scarring, or even corneal perforation. Early detection and screening of keratoconus are essential for effective management and treatment. Several screening methods, such as corneal topography and tomography, corneal biomechanics, and genetic testing, are being developed to detect keratoconus at an early stage. Once detected, prevention of progression is the mainstay of keratoconus management. Corneal collagen cross-linking is a minimally invasive treatment option that can slow or halt the progression of keratoconus. Additionally, recent studies have investigated the potential use of copper sulfate eye drops (IVMED-80) and extracellular vesicles to prevent the progression of keratoconus as non-invasive treatment options. For visual rehabilitation, currently available treatments include scleral lenses, intracorneal ring segments, corneal allogenic intrastromal ring segments, and deep anterior lamellar keratoplasty. The safety and efficacy of these emerging treatment options for keratoconus are currently being investigated.

Contributed Session II: Detecting subclinical keratoconus by mapping corneal biomechanics using wave-based optical coherence elastography.

The detection of subclinical keratoconus in human corneas remains a challenging task. It is hypothesized that the focal reduction in biomechanical properties of the cornea would be the primary initiating event of keratoconus before changes in topography and pachymetry are manifested. We propose to use an air-couple ultrasonic optical coherence elastography system to map the elasticity of the corneas of 15 patients with a clinical diagnosis of keratoconus (KC) in one eye, and subclinical keratoconus (SK) in the fellow eye. The fully non-contact system excites the corneal apex to produce Lamb wave propagation and measures wave speed and average corneal thickness along 16 semi-meridians. Measurements in 30 human healthy subjects (control group: CG, N = 60 corneas) defined a baseline of normal biomechanics in two biomarkers: Spatial Anisotropy of Wave Speed (SAWS < 0.207), and the Speed-Thickness Index (STI > 0). SAWS was significantly higher in KC (0.353, p < 0.001) and SK (0.249, p < 0.001) corneas when compared to the CG. Moreover, STI maps show abnormal elasticity in SK (STI = -1.22) and KC (STI = -0.375) compared to the CG. Our results show important biomechanical differences in SAWS and STI between normal, subclinical, and advanced stages of keratoconus, suggesting those as potential biomarkers to identify "at-risk" corneas before changes in topography and pachymetry become evident.

Keratoconus Detection-based on Dynamic Corneal Deformation Videos Using Deep Learning

To assess the performance of convolutional neural networks (CNNs) for automated detection of keratoconus (KC) in standalone Scheimpflug-based dynamic corneal deformation videos. Retrospective cohort study. We retrospectively analyzed datasets with records of 734 nonconsecutive, refractive surgery candidates, and patients with unilateral or bilateral KC. We first developed a video preprocessing pipeline to translate dynamic corneal deformation videos into 3-dimensional pseudoimage representations and then trained a CNN to directly identify KC from pseudoimages. We calculated the model's KC probability score cut-off and evaluated the performance by subjective and objective accuracy metrics using 2 independent datasets. Area under the receiver operating characteristics curve (AUC), accuracy, specificity, sensitivity, and KC probability score. The model accuracy on the test subset was 0.89 with AUC of 0.94. Based on the external validation dataset, the AUC and accuracy of the CNN model for detecting KC were 0.93 and 0.88, respectively. Our deep learning-based approach was highly sensitive and specific in separating normal from keratoconic eyes using dynamic corneal deformation videos at levels that may prove useful in clinical practice. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Screening of sensitive in vivo characteristics for early keratoconus diagnosis: a multicenter study.

Purpose: To analyze and compare sensitive in vivo characteristics for screening early keratoconus. Methods: This multicenter, case-control study included 712 eyes, after matching for age and biomechanically corrected intraocular pressure, from three clinics in different cities. The keratoconus (n = 288), early keratoconus (n = 91), and normal cornea (n = 333) groups included eyes diagnosed with bilateral keratoconus, fellow eyes with relatively normal topography with unilateral keratoconus, and normal eyes before refractive surgery, respectively. After adjusting for central corneal thickness, differences in vivo characteristics were analyzed among the three groups. The in vivo characteristics were measured by Pentacam and Corvis ST. Fifty-four indices were evaluated to screen for a sensitive index for the detection of early keratoconus. Results: Significant differences were observed in 26 of the 36 corneal biomechanical indeces between the early keratoconus and normal corneas. The area under the receiver operating characteristic curve of tomographic and biomechanical index, Belin/Ambrósio deviation, and Da in differentiating keratoconus from normal cornea was 1.000. Among the top five indeces of the area under the receiver operating characteristic curve for detecting early keratoconus, the corneal biomechanical-related index accounted for 80% (4/5), including A1 dArc length, highest concavity radius, A2 time, and tomographic and biomechanical index, of which the area under the receiver operating characteristic curve of A1 dArc length was 0.901. Conclusion: A1 dArc length and several corneal biomechanical indices are highly sensitive for the detection of early keratoconus, even in the absence of topographic abnormalities. Ophthalmologists should focus on the clinical application of corneal biomechanics and combine corneal tomography for the timely and accurate detection of early keratoconus.

Comparison of different corneal imaging modalities using artificial intelligence for diagnosis of keratoconus: a systematic review and meta-analysis.

This review was designed to compare different corneal imaging modalities using artificial intelligence (AI) for the diagnosis of keratoconus (KCN), subclinical KCN (SKCN), and forme fruste KCN (FFKCN). A comprehensive systematic search was conducted in scientific databases, including Web of Science, PubMed, Scopus, and Google Scholar based on the PRISMA statement. Two independent reviewers assessed all potential publications on AI and KCN up to March 2022. The Critical Appraisal Skills Program (CASP) 11-item checklist was used to evaluate the validity of the studies. Eligible articles were categorized into three groups (KCN, SKCN, and FFKCN) and included in the meta-analysis. The pooled estimate of accuracy (PEA) was calculated for all selected articles. The initial search yielded 575 relevant publications, of which 36 met the CASP quality criteria and were included in the analysis. Qualitative assessment showed that Scheimpflug and Placido combined with biomechanical and wavefront evaluations improved KCN detection (PEA, 99.2, and 99.0, respectively). The Scheimpflug system (92.25 PEA, 95% CI, 94.76-97.51) and a combination of Scheimpflug and Placido (96.44 PEA, 95% CI, 93.13-98.19) had the highest diagnostic accuracy for the detection of SKCN and FFKCN, respectively. The meta-analysis outcomes showed no significant difference between the CASP score and accuracy of the publications (all P > 0.05). Simultaneous Scheimpflug and Placido corneal imaging methods provide high diagnostic accuracy for early detection of keratoconus. The use of AI models improves the discrimination of keratoconic eyes from normal corneas.

Brillouin Spectroscopy in Ophthalmology.

Information about corneal biomechanics is crucial for achieving satisfactory outcomes after surgical corneal interventions, e.g., for astigmatic keratotomies, but also to identify corneas that are at risk for postoperative complications such as corneal ectasia. Hitherto, approaches to characterize corneal biomechanics in an in vivo setting have yielded only minor success, demonstrating the unmet medical need for a diagnostic technique to measure ocular biomechanics. This review shall explain the mechanism of Brillouin spectroscopy and summarize the current scientific knowledge for ocular tissue. PubMed research of relevant experimental and clinical publications, as well as reporting of own experience using Brillouin spectroscopy. Brillouin spectroscopy can measure different biomechanical moduli with a high spatial resolution. Currently, available devices are able to detect focal corneal weakening, e.g., in keratoconus, as well as stiffening after corneal cross-linking. Also, the mechanical properties of the crystalline can be measured. Corneal anisotropy and hydration, together with the dependence on the angle of the incident laser beam in Brillouin spectroscopy, are challenges in the precise interpretation of measured data. A clear advantage in the detection of subclinical keratoconus compared to corneal tomography, however, has not been shown yet. Brillouin spectroscopy is a technique to characterize biomechanical properties of ocular tissue in vivo. Published results confirm ex vivo data of ocular biomechanics; however, further improvements in the acquisition and interpretation of measured data are required until this technique can be used in a clinically viable setting.

Comparison of bilateral differential characteristics of corneal biomechanics between keratoconus and normal eyes.

Purpose: To compare bilateral differences in corneal biomechanics between keratoconus and normal eyes. Methods: In this case-control study, 346 eyes of 173 patients (aged 22.1 ± 6.1years) with keratoconus (KC group) and 378 eyes of 189 patients (aged 26.7 ± 5.6years) with ametropia (control group) were enrolled. Corneal tomography and biomechanical properties were examined using Pentacam HR and Corvis ST, respectively. The corneal biomechanical parameters were compared between eyes with forme fruste keratoconus (FFKC) and normal eyes. Bilateral differences in corneal biomechanical parameters were compared between the KC and control groups. Receiver operating characteristic (ROC) analysis was used to assess discriminative efficacies. Results: The areas under the ROC curves (AUROCs) of stiffness parameter at the first applanation (SP-A1) and Tomographic and Biomechanical Index (TBI) for identifying FFKC were 0.641 and 0.694, respectively. The bilateral differential values of major corneal biomechanical parameters were significantly increased in the KC group (all p < 0.05), except for the Corvis Biomechanical Index (CBI). The AUROCs of the bilateral differential values of the deformation amplitude ratio at 2mm (ΔDAR2), Integrated Radius (ΔIR), SP-A1 (ΔSP-A1), and the maximum inverse concave radius (ΔMax ICR) for discriminating keratoconus were 0.889, 0.884, 0.826, and 0.805, respectively. The Logistic Regression Model-1 (comprising of ΔDAR2, ΔIR, and age) and the Logistic Regression Model-2 (comprising of ΔIR, ΔARTh, ΔBAD-D, and age) had AUROCs of 0.922 and 0.998, respectively, for discriminating keratoconus. Conclusion: The bilateral asymmetry of corneal biomechanics was significantly increased in keratoconus compared with normal eyes, which may be helpful for the early detection of keratoconus.

Detection of Keratoconus With a New Corvis Biomechanical Index Optimized for Chinese Populations

The aim of this study was to introduce an optimized version of the Corvis Biomechanical Index for Chinese population (cCBI). Retrospective, multicenter clinical validity-enhancement study. Patients were included from seven clinics (Beijing, Shenyang, Guangzhou, Shanghai, Wenzhou, Chongqing and Tianjin). Logistic regression was employed to optimize the values of the constants of the CBI, based on Database 1 as the development dataset (six out of 7 clinics), to create a new version of the index named cCBI. The factors of the CBI (A1Velocity, ARTh, Stiffness Parameter-A, DARatio2mm and Inverse Integrated Radius) and the cut-off value were kept the same (0.5). With the formation of cCBI determined, it was validated on Database 2 (1 out 7 of the clinics). Two thousand four hundred and seventy-three patients (healthy and keratoconus) were included. In the database 2, the Area Under the Curve (AUC) of the cCBI was 0.985 with a 93.4 % specificity and 95.5 % sensitivity. In the same Dataset the original CBI produced an AUC of 0.978 with a 68.1 % specificity and 97.7 % sensitivity. There was a statistically significant difference between the ROC curve of cCBI and CBI (De Long p=0.0009) CONCLUSION: :The new cCBI for Chinese patients was shown to be statistically significantly better when compared to CBI to separate healthy from keratoconic eyes. The presence of an external validation dataset confirms this finding and suggests the use of cCBI in everyday clinical practice to aid in the diagnosis of keratoconus in Chinese ethnicity.

Combinations of Scheimpflug tomography, ocular coherence tomography and air-puff tonometry improve the detection of keratoconus

PurposeTo determine whether combinations of devices with different measuring principles, supported by artificial intelligence (AI), can improve the diagnosis of keratoconus (KC). MethodsScheimpflug tomography, spectral-domain optical coherence tomography (SD-OCT), and air-puff tonometry were performed in all eyes. The most relevant machine-derived parameters to diagnose KC were determined using feature selection. The normal and forme fruste KC (FFKC) eyes were divided into training and validation datasets. The selected features from a single device or different combinations of devices were used to develop models based on random forest (RF) or neural networks (NN) trained to distinguish FFKC from normal eyes. The accuracy was determined using receiver operating characteristic (ROC) curves, area under the curve (AUC), sensitivity, and specificity. Results271 normal eyes, 84 FFKC eyes, 85 early KC eyes, and 159 advanced KC eyes were included. A total of 14 models were built. Air-puff tonometry had the highest AUC for detecting FFKC using a single device (AUC = 0.801). Among all two-device combinations, the highest AUC was accomplished using RF applied to selected features from SD-OCT and air-puff tonometry (AUC = 0.902), followed by the three-device combination with RF (AUC = 0.871) with the best accuracy. ConclusionExisting parameters can precisely diagnose early and advanced KC, but their diagnostic ability for FFKC could be optimized. Applying an AI algorithm to a combination of air-puff tonometry with Scheimpflug tomography or SD-OCT could improve FFKC diagnostic ability. The improvement in diagnostic ability by combining three devices is modest.