Ocular Imaging for Eye Care

A special issue of Photonics (ISSN 2304-6732). This special issue belongs to the section "Biophotonics and Biomedical Optics".

Deadline for manuscript submissions: closed (31 January 2022) | Viewed by 33734

Special Issue Editors


E-Mail Website
Guest Editor
Applied Optics and Image Processing Group – GOAPI, Departament of Optics and Optometry, Universitat Politècnica de Catalunya – BarcelonaTech, c/ Violinista Vellsolà, 37, ES-08222, Spain
Interests: information optics; image processing; optical processors; biomedical optics and imaging; visual science; image-based industrial inspection; optical security

E-Mail Website
Guest Editor
Dept. of Mechanical and Mechatronics Engineering, Universidad Tecnológica de Bolívar, Cartagena, 130008, Km 1 via Turbaco, Colombia
Interests: biomedical imaging; optical metrology; digital image processing; computer vision; 3D imaging

Special Issue Information

Dear colleagues,

As we are still in the middle of an ongoing pandemic, the need for fast, inexpensive, and robust medical technology is now more present than ever. Ocular imaging is an integral part of the ophthalmic examination and is necessary for many clinical aspects in eye care.

The recent advances in imaging technologies and information processing have extended their applicability to eye care in continuous growth. The combination of several factors, such as multimodality, high-speed, high-resolution, 3D scanning, wide-field, in-vivo biometry, anterior and posterior eye examination, along with the development of smart sensors and innovative image analysis algorithms have provided physicians with new and powerful tools for diagnosis, understanding, screening and enhanced monitoring of a wide range of eye disorders. Other techniques such as on-bench optical simulation and testing, artificial intelligence, and deep learning contribute with added values to meet the emerging challenges in early diagnosis, treatment monitoring, and disease progression prediction. Many advanced imaging techniques, limited in the past to laboratory-based research methods, have already expanded the clinical abilities, but their potential still deserves to be explored. This Special Issue aims to show the progress in (but not limited to) the following ocular imaging technologies with applications to eye care, ranging from the tear film and the cornea to the retina:

  • Optical coherent tomography (OCT) to image both anterior and posterior eye structures;
  • Enface OCT, OCT-Angiography;
  • Imaging during ocular surgery;
  • Adaptive optics in ocular imaging;
  • Imaging and measuring in the eye: Multimodal imaging, ocular biometry, Scheimpflug imaging, corneal endothelial specular microscopy, (ultra)-wide field retinal imaging, scanning laser ophthalmoscopy, mydriatic and non-mydriatic retinal cameras, machine learning techniques and artificial intelligence applied to pattern recognition, image classification and ocular assessment.

Prof. Dr. Maria Millan
Dr. Andres G. Marrugo
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Photonics is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • In-vivo imaging and measuring the anterior segment of the eye
  • Imaging and measuring the posterior segment of the eye
  • Features for eye fundus and retinal image description
  • Adaptive optics and bio-microscopy applied to the eye
  • Multimodal optical inspection techniques of the eye
  • Optical instruments and innovative software for ocular image interpretation
  • Artificial-intelligence-driven techniques in ocular imaging
  • Applications of ocular imaging to Ophthalmology and clinical practice
  • Optical image quality of ocular implants and visual performance

Published Papers (13 papers)

Order results
Result details
Select all
Export citation of selected articles as:

Editorial

Jump to: Research

3 pages, 185 KiB  
Editorial
Editorial to the Special Issue on “Ocular Imaging for Eye Care”
by Maria S. Millan and Andres G. Marrugo
Photonics 2022, 9(7), 475; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9070475 - 08 Jul 2022
Viewed by 1367
Abstract
The need for fast, inexpensive, and robust medical technology is now more crucial than ever [...] Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)

Research

Jump to: Editorial

13 pages, 2058 KiB  
Article
The Effect of Accommodation on Peripheral Refraction under Two Illumination Conditions
by Raquel van Ginkel, María Mechó, Genis Cardona and José M. González-Méijome
Photonics 2022, 9(5), 364; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9050364 - 23 May 2022
Cited by 4 | Viewed by 2542
Abstract
The clinical importance of peripheral refraction as a function of accommodation has become increasingly evident in the last years with special attention given to myopia control. Low order ocular aberrations were measured with a Hartmann–Shack aberrometer in a sample of 28 young emmetropic [...] Read more.
The clinical importance of peripheral refraction as a function of accommodation has become increasingly evident in the last years with special attention given to myopia control. Low order ocular aberrations were measured with a Hartmann–Shack aberrometer in a sample of 28 young emmetropic subjects. A stationary Maltese cross was presented at 2.5 D and 5.0 D of accommodative demand and at 0°, 10° and 20° of eccentricity in the horizontal visual field under two different illumination conditions (white and red light). Wavefront data for a 3 mm pupil diameter were analyzed in terms of the vector components of refraction (M, J0 and J45) and the relative peripheral refractive error (RPRE) was calculated. M was myopic at both accommodative demands and showed a statistically significant myopic increase with red illumination. No significant change in J0 and J45 was found with accommodation nor between illumination conditions. However, J0 increased significantly with eccentricity, exhibiting a nasal-temporal asymmetry. The RPRE was myopic at both accommodation demands and showed a statistically significant hyperopic shift at 20° in the nasal retina. The use of red light introduced statistically and clinically significant changes in M, explained by the variation of the ocular focal length under a higher wavelength illumination, increasing the experimental accommodative demand. These findings may be of relevance for research exploring peripheral refraction under accommodation, as the choice of target illumination is not trivial. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

13 pages, 6405 KiB  
Article
Retinal and Choroidal Thickness in Myopic Young Adults
by Enrique J. Fernández, José A. Villa-Carpes, Rosa M. Martínez-Ojeda, Francisco J. Ávila and Juan M. Bueno
Photonics 2022, 9(5), 328; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9050328 - 10 May 2022
Cited by 3 | Viewed by 2426
Abstract
The retinal and the choroidal thickness were measured at four locations along the horizontal direction (foveola, one nasal to the fovea and two temporal) in a group of 43 young adults (mean age: 27.1 ± 3.9 years), with ocular refraction ranging from emmetropia [...] Read more.
The retinal and the choroidal thickness were measured at four locations along the horizontal direction (foveola, one nasal to the fovea and two temporal) in a group of 43 young adults (mean age: 27.1 ± 3.9 years), with ocular refraction ranging from emmetropia to high myopia (0 to −10D). Thickness values were obtained from OCT images centered at the foveal depression. The retinal thickness exhibited a correlation with refraction at all eccentricities but not at the fovea. When different subgroups of refraction were considered, the analysis of such correlations indicated that only the retinal thickness in the group of high myopia (refraction ≤ −6D) was statistically different from the other two groups (emmetropes: [−0.5, 0] D, and myopes: (−6, −0.5) D). No significant differences were found between emmetropic and myopic groups. In contrast to the retina, the choroidal thickness exhibited a significant correlation with refraction at the fovea, although such dependency only stood for high myopes (the choroid of myopes and emmetropes exhibited similar thickness). Correlation with refraction was also found at the nasal location, arising between emmetropic and high myopia groups. Other choroidal locations among groups did not exhibit relationship with the refraction. It is concluded that the differences in the choroid and retina thickness along the horizontal meridian as a function of refraction do not characterize the onset and progression of myopia at early stages, since they only manifest in the group of high myopia. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

12 pages, 25681 KiB  
Article
Improvement of Retinal Images Affected by Cataracts
by Enrique Gonzalez-Amador, Justo Arines, Pablo Charlón, Nery Garcia-Porta, Maximino J. Abraldes and Eva Acosta
Photonics 2022, 9(4), 251; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9040251 - 10 Apr 2022
Cited by 3 | Viewed by 2405
Abstract
Eye fundus images are used in clinical diagnosis for the detection and assessment of retinal disorders. When retinal images are degraded by scattering due to opacities of the eye tissues, the precise detection of abnormalities is complicated depending on the grading of the [...] Read more.
Eye fundus images are used in clinical diagnosis for the detection and assessment of retinal disorders. When retinal images are degraded by scattering due to opacities of the eye tissues, the precise detection of abnormalities is complicated depending on the grading of the opacity. This paper presents a concept proof study on the use of the contrast limited adaptive histogram equalization (CLAHE) technique for better visualization of eye fundus images for different levels of blurring due to different stages of cataracts. Processing is performed in three different color spaces: RGB, CIELAB and HSV, with the aim of finding which one better enhances the missed diagnostic features due to blur. The experimental results show that some fundus features not observable by naked eye can be detected in some of the space color processed with the proposed method. In this work, we also develop and provide an online image process, which allows clinicians to tune the default parameters of the algorithm for a better visualization of the characteristics of fundus images. It also allows the choice of a region of interest (ROI) within the images that provide better visualization of some features than those enhanced by the processing of the full picture. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

16 pages, 1776 KiB  
Article
Understanding In Vivo Chromatic Aberrations in Pseudophakic Eyes Using on Bench and Computational Approaches
by Maria Vinas-Pena, Alberto de Castro, Carlos Dorronsoro, Ana Gonzalez-Ramos, Suad Redzovic, Nicolas Willet, Nuria Garzon and Susana Marcos
Photonics 2022, 9(4), 226; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9040226 - 30 Mar 2022
Cited by 4 | Viewed by 1941
Abstract
Diffractive multifocal intraocular lenses (IOLs) modulate chromatic aberration and reduce it at certain distances due to interactions between the refractive and diffractive chromatic components. However, the extent to which computer modeling and on bench measurements of IOL chromatic aberration translate to chromatic aberration [...] Read more.
Diffractive multifocal intraocular lenses (IOLs) modulate chromatic aberration and reduce it at certain distances due to interactions between the refractive and diffractive chromatic components. However, the extent to which computer modeling and on bench measurements of IOL chromatic aberration translate to chromatic aberration in patients implanted with these multifocal IOLs (MIOLs) is not yet fully understood. In this study, we compare the chromatic difference of focus and longitudinal chromatic aberrations in pseudophakic patients implanted with different IOL designs (monofocal and trifocal IOLs) and materials (hydrophobic and hydrophilic), and compared them with predictions from computer eye models and on bench measurements with the same IOLs. Patient data consisted of results from 63 pseudophakic eyes reported in four different studies and obtained psychophysically in the visual testing channel of a custom-developed polychromatic adaptive optics system. Computational predictions were obtained using ray tracing on computer eye models, and modulation transfer function (MTF) on bench measurements on physical eye models. We found that LCA (in vivo/simulated) for far vision was 1.37 ± 0.08 D/1.19 D for monofocal hydrophobic, 1.21 ± 0.08 D/0.88 D for monofocal hydrophilic, 0.99 ± 0.06 D/1.19 D for MIOL hydrophobic, and 0.82 ± 0.05 D/0.88 D for MIOL hydrophilic. For intermediate and near vision, LCA (in vivo/simulated) was 0.67 ± 0.10 D/0.75 D and 0.23 ± 0.08 D/0.19 D for MIOL hydrophobic and 0.27 ± 0.15 D/0.38 D and 0.15 ± 0.15 D/−0.13 D for MIOL hydrophilic, respectively. In conclusion, computational ray tracing and on bench measurements allowed for evaluating in vivo chromatic aberration with different materials and designs for multifocal diffractive intraocular lenses. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

11 pages, 946 KiB  
Article
Gullstrand Intracapsular Accommodation Mechanism Revised
by Norberto López-Gil
Photonics 2022, 9(3), 152; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9030152 - 03 Mar 2022
Cited by 2 | Viewed by 2302
Abstract
More than a century has passed since Allvar Gullstrand proposed his so called “mechanism of intracapsular accommodation” based on the concept that the change in lens power during accommodation can be larger if the lens is modeled by two different refractive indexes (cortex [...] Read more.
More than a century has passed since Allvar Gullstrand proposed his so called “mechanism of intracapsular accommodation” based on the concept that the change in lens power during accommodation can be larger if the lens is modeled by two different refractive indexes (cortex and nucleus) than if it is homogenous. Intracapsular accommodation of the original eye model proposed by Gullstrand is compared with that of a six-surfaces model eye built based on actual experimental data obtained with precise imaging techniques of the lens change during accommodation. The results show that nearly half (42%) of the accommodation of the Gullstrand model eye is produced by said mechanism, while a model eye based on actual data produces a small intracapsular dis-accommodation. The main reason for the discrepancy is the larger change in the radii of curvature of the lens nucleus during accommodation proposed by Gullstrand, which is necessary because his model likely did not take into account the reduction in the change in lens power produced by depth-of-focus. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

8 pages, 3226 KiB  
Article
Assessment of a New Trifocal Diffractive Corneal Inlay for Presbyopia Correction Using an Adaptive Optics Visual Simulator
by Anabel Martínez-Espert, Diego Montagud-Martínez, Vicente Ferrando, Walter D. Furlan and Juan A. Monsoriu
Photonics 2022, 9(3), 135; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9030135 - 25 Feb 2022
Cited by 5 | Viewed by 1692
Abstract
In this work, we analyze a proposal of a new intracorneal diffractive lens for presbyopia correction that could allow good, distance, intermediate and near vision. By using an adaptive optics visual simulator, we study the influence of two factors in the inlay performance: [...] Read more.
In this work, we analyze a proposal of a new intracorneal diffractive lens for presbyopia correction that could allow good, distance, intermediate and near vision. By using an adaptive optics visual simulator, we study the influence of two factors in the inlay performance: the spherical aberration (SA) and the potential errors of in thickness, induced in the manufacturing process. We show that the inlay through-the-focus imaging performance can be customized with the SA value, favoring either distance–intermediate or intermediate–near vision. Moreover, we found that with thickness variations of 10%, the inlay still maintains its trifocal nature. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

15 pages, 6059 KiB  
Article
Trade-Off Asymmetric Profile for Extended-Depth-of-Focus Ocular Lens
by Lenny A. Romero, Andrés G. Marrugo and María S. Millán
Photonics 2022, 9(2), 119; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9020119 - 19 Feb 2022
Cited by 2 | Viewed by 1723
Abstract
We explore the possibility of extending the depth of focus of an imaging lens with an asymmetric quartic phase-mask, while keeping the aberration within a relatively low level. This can be intended, for instance, for ophthalmic applications, where no further digital processing can [...] Read more.
We explore the possibility of extending the depth of focus of an imaging lens with an asymmetric quartic phase-mask, while keeping the aberration within a relatively low level. This can be intended, for instance, for ophthalmic applications, where no further digital processing can take place, relying instead on the patient’s neural adaptation to their own aberrations. We propose a computational optimization method to derive the design-strength factor of the asymmetric profile. The numerical and experimental results are shown. The optical experiment was conducted by means of a modulo-2π phase-only spatial light modulator. The proposed combination of the asymmetric mask and the lens can be implemented in a single refractive element. An exemplary case of an extended-depth-of focus intraocular lens based on the proposed element is described and demonstrated with a numerical experiment. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

12 pages, 2033 KiB  
Article
Visual Acuity at Various Distances and Defocus Curve: A Good Match
by Laura Clavé, Aurora Torrents and María S. Millán
Photonics 2022, 9(2), 85; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics9020085 - 01 Feb 2022
Cited by 4 | Viewed by 2235
Abstract
The defocus curve (DC) has become the gold standard method for assessing the visual performance at different distances, and hence, to evaluate the depth of focus that is provided by modern presbyopia-compensating intraocular lenses that are used in cataract surgery. A former study, [...] Read more.
The defocus curve (DC) has become the gold standard method for assessing the visual performance at different distances, and hence, to evaluate the depth of focus that is provided by modern presbyopia-compensating intraocular lenses that are used in cataract surgery. A former study, carried out by other authors, reported that the DC method, based on placing the chart at far distance and using trial negative lenses, underestimated the visual acuity that was measured by varying the test distance (VD) in pseudophakic patients. We have revisited the issue with pseudophakic subjects and have extended the study to include young phakic observers with their natural accommodation preserved. Our results show that both the VD and DC methods are equivalent in precision and accuracy and can be used indistinctly. However, the DC method is more practical in the clinics because it does not require further control of the chart positioning and illuminance. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

12 pages, 4430 KiB  
Article
Separate Detection of Stromal and Epithelial Corneal Edema on Optical Coherence Tomography Using a Deep Learning Pipeline and Transfer Learning
by Pierre Zéboulon, Wassim Ghazal, Karen Bitton and Damien Gatinel
Photonics 2021, 8(11), 483; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics8110483 - 28 Oct 2021
Cited by 3 | Viewed by 2204
Abstract
The accurate detection of corneal edema has become a topic of growing interest with the generalization of endothelial keratoplasty. Despite recent advances in deep learning for corneal edema detection, the problem of minimal edema remains challenging. Using transfer learning and a limited training [...] Read more.
The accurate detection of corneal edema has become a topic of growing interest with the generalization of endothelial keratoplasty. Despite recent advances in deep learning for corneal edema detection, the problem of minimal edema remains challenging. Using transfer learning and a limited training set of 11 images, we built a model to segment the corneal epithelium, which is part of a three-model pipeline to detect corneal edema. A second and a third model are used to detect edema on the stroma alone and on the epithelium. A validation set of 233 images from 30 patients consisting of three groups (Normal, Minimal Edema and important Edema) was used to compare the results of our new pipeline to our previous model. The mean edema fraction (EF), defined as the number of pixels detected as edema divided by the total number of pixels of the cornea, was calculated for each image. With our previous model, the mean EF was not statistically different between the Normal and Minimal Edema groups (p = 0.24). With the current pipeline, the mean EF was higher in the Minimal Edema group compared to the Normal group (p < 0.01). The described pipeline constitutes an adjustable framework for the detection of corneal edema based on optical coherence tomography and yields better performances in cases of minimal or localized edema. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

9 pages, 8943 KiB  
Article
Simulations of Decentration and Tilt of a Supplementary Sulcus-Fixated Intraocular Lens in a Polypseudophakic Combination Using Ray-Tracing Software
by Grzegorz Łabuz, Gerd U. Auffarth, Weijia Yan, Timur M. Yildirim and Ramin Khoramnia
Photonics 2021, 8(8), 309; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics8080309 - 02 Aug 2021
Cited by 7 | Viewed by 2735
Abstract
This study aimed to assess image quality after the tilt and decentration of supplementary intraocular lenses (IOLs) in a two-lens configuration. One was designed for sulcus fixation with a nominal power range of 1D–10D and was combined with a capsular fixation 20D IOL. [...] Read more.
This study aimed to assess image quality after the tilt and decentration of supplementary intraocular lenses (IOLs) in a two-lens configuration. One was designed for sulcus fixation with a nominal power range of 1D–10D and was combined with a capsular fixation 20D IOL. The optical performance of a ray-tracing model was tested under IOL misalignment through the area under the modulation transfer function (MTFa) and wave aberrations. Tilting by 10° resulted in a 4% reduction of the MTFa for a 10D IOL as compared to 9% for the 20D lens. The two models demonstrated good tolerance to a 1 mm decentration; as for the 10D sulcus-fixated lens, the MTFa loss was 2%, and 4% for the capsular fixation lens. Coma and astigmatism increased three- and four-fold, respectively, after a 10° tilt compared to the aberration level induced by the 1 mm decentration. Both analyses showed a trend towards a lower MTF impact and fewer optical errors with decreasing nominal power. In conclusion, when misaligned, low-power sulcus-fixated IOLs might retain their good optical quality. An extreme tilt of 10° has a more detrimental effect on the IOL performance than a 1 mm decentration. The proper alignment of a high-power capsular fixation lens is important in achieving a desirable postoperative outcome. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

9 pages, 1874 KiB  
Article
Scaling the Retinal Image of the Wide-Angle Eye Using the Nodal Point
by Michael J. Simpson
Photonics 2021, 8(7), 284; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics8070284 - 17 Jul 2021
Cited by 9 | Viewed by 5151
Abstract
Angles subtended at the second nodal point of the eye (NP2) are approximately the same as input visual angles over a very large angular range, despite the nodal point being a paraxial lens property. Raytracing using an average pseudophakic eye showed that the [...] Read more.
Angles subtended at the second nodal point of the eye (NP2) are approximately the same as input visual angles over a very large angular range, despite the nodal point being a paraxial lens property. Raytracing using an average pseudophakic eye showed that the angular nodal point criterion was only valid up to about 10°, and that the linear relationship was due instead to the cornea and lens initially creating chief ray angles at the exit pupil that are about 0.83 times input values for this particular eye, and then by the retina curving around to meet the rays in a manner that compensates for increasing angle. This linear relationship is then also maintained when retinal intersections are calculated relative to other axial points, with angles rescaled approximately using the equation R/(R + delta), where delta is the axial distance from the center of a spherical retina of radius R. Angles at NP2 approximately match the input angles, but the terminology is misleading because this is not a paraxial property of the eye. Chief rays are used with finite raytracing to determine the actual behavior. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

14 pages, 2902 KiB  
Article
Digital Image Processing and Development of Machine Learning Models for the Discrimination of Corneal Pathology: An Experimental Model
by Andres Bustamante-Arias, Abbas Cheddad, Julio Cesar Jimenez-Perez and Alejandro Rodriguez-Garcia
Photonics 2021, 8(4), 118; https://0-doi-org.brum.beds.ac.uk/10.3390/photonics8040118 - 10 Apr 2021
Cited by 6 | Viewed by 3319
Abstract
Machine learning (ML) has an impressive capacity to learn and analyze a large volume of data. This study aimed to train different algorithms to discriminate between healthy and pathologic corneal images by evaluating digitally processed spectral-domain optical coherence tomography (SD-OCT) corneal images. A [...] Read more.
Machine learning (ML) has an impressive capacity to learn and analyze a large volume of data. This study aimed to train different algorithms to discriminate between healthy and pathologic corneal images by evaluating digitally processed spectral-domain optical coherence tomography (SD-OCT) corneal images. A set of 22 SD-OCT images belonging to a random set of corneal pathologies was compared to 71 healthy corneas (control group). A binary classification method was applied where three approaches of ML were explored. Once all images were analyzed, representative areas from every digital image were also extracted, processed and analyzed for a statistical feature comparison between healthy and pathologic corneas. The best performance was obtained from transfer learning—support vector machine (TL-SVM) (AUC = 0.94, SPE 88%, SEN 100%) and transfer learning—random forest (TL- RF) method (AUC = 0.92, SPE 84%, SEN 100%), followed by convolutional neural network (CNN) (AUC = 0.84, SPE 77%, SEN 91%) and random forest (AUC = 0.77, SPE 60%, SEN 95%). The highest diagnostic accuracy in classifying corneal images was achieved with the TL-SVM and the TL-RF models. In image classification, CNN was a strong predictor. This pilot experimental study developed a systematic mechanized system to discern pathologic from healthy corneas using a small sample. Full article
(This article belongs to the Special Issue Ocular Imaging for Eye Care)
Show Figures

Figure 1

Back to TopTop