Evaluation of Deep Learning Networks for Keratoconus Detection Using Corneal Topographic Images

Abstract

Keratoconus is an eye disease of 'deformation of corneal curvature' caused due to 'non-inflammatory progressive thinning' resulting into loss of elasticity in cornea and protrudes a cone shape formation that ultimately reduces visual acuity. For many years, researchers have worked towards accurate detection of keratoconus (KCN) as it is essential checkup before any refractive surgery demanding quick as well as precise clinical diagnosis and treatments of keratoconus prior to LASIK. In our study, we have firstly derived two variants of the original corneal topographies namely 'images with edges' and 'images with edges-and-mask', as data sets. The deep neural network techniques such as Artificial Neural Network (ANN), Convolutional Neural Network (CNN) and pertained VGG16 model are applied on original 'corneal topographies' as well as on the two of its variants and the results obtained are presented. Keywords: Keratoconus Á Corneal topography Á ATLAS 9000 Á ANN Á CNN Á VGG16 Á Canny edge detection Á Edges with mask

Artificial Intelligence Technique of Synthesis and Characterizations for Measurement of Optical Particles in Medical Devices

“…where ɳ p = y ρ /cos υ and ɳ s = y ρ cos υ and υ is the angle of incidence [13]. Nanoscience and nanotechnology basically deal with structure, characterization, exploration, and utilization of nanostructured materials.…”

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

“…The most common model was the adapted VGG16 model, which was evaluated by 8 studies. 12,[16][17][18]21,23,24,27 To train the model and evaluate the performance, 11 studies split the samples into training and validation and/or test sets. Six studies used crossvalidation, 5,15,17,21,23,26 and 2 studies used both crossvalidation and split sampling technique.…”

“…12,[16][17][18]21,23,24,27 To train the model and evaluate the performance, 11 studies split the samples into training and validation and/or test sets. Six studies used crossvalidation, 5,15,17,21,23,26 and 2 studies used both crossvalidation and split sampling technique. 7,20 The median sample size used to evaluate model performance was 179 images/eyes (ranging from 24 to 1104 images/eyes).…”

“…We included 19 studies in the systematic review that contained 96,549 images or scans from over 10,500 patients. 5,7,[12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28] Five studies assessed the diagnostic performance of DL algorithms using anterior segment OCT images, 5,[17][18][19]26 12 studies using corneal topography images (8 Scheimpflug and 4 Placido disk based), 7,[12][13][14][15][16][20][21][22][23][24]28 and 2 studies using smartphone images (topography 25 and lateral segment images 27 ). Keratoconus grading was reported in 11 studies, 4 of which used the Amsler-Krumeich classification.…”

Applications of Deep Learning on Topographic Images to Improve the Diagnosis for Dynamic Systems and Unconstrained Optimization

“…Analysis Method. Continuing with the use of the Fundamental Methodology for Data Science [10,11], the modelling stage is where the data sets defined in the previous stage will be used. That is why this research work will propose five designs with different neural network architectures as well as the dimension that each image will have to validate which of them will provide the highest level of precision concerning learning and testing of the neural network, the designs proposed for this research work are the following (Table 1).…”