Deep learning in ophthalmology: a review.

Deep learning is an emerging technology with numerous potential applications in Ophthalmology. Deep learning tools have been applied to different diagnostic modalities including digital photographs, optical coherence tomography, and visual fields. These tools have demonstrated utility in assessment of various disease processes including cataracts, glaucoma, age-related macular degeneration, and diabetic retinopathy. Deep learning techniques are evolving rapidly, and will become more integrated into ophthalmic care. This article reviews the current evidence for deep learning in ophthalmology, and discusses future applications, as well as potential drawbacks.

Computer-aided diagnosis of retinopathy in retinal fundus images of preterm infants via quantification of vascular tortuosity.

Retinopathy of prematurity (ROP), a disorder of the retina occurring in preterm infants, is the leading cause of preventable childhood blindness. An active phase of ROP that requires treatment is associated with the presence of plus disease, which is diagnosed clinically in a qualitative manner by visual assessment of the existence of a certain level of increase in the thickness and tortuosity of retinal vessels. The present study performs computer-aided diagnosis (CAD) of plus disease via quantitative measurement of tortuosity in retinal fundus images of preterm infants. Digital image processing techniques were developed for the detection of retinal vessels and measurement of their tortuosity. The total lengths of abnormally tortuous vessels in each quadrant and the entire image were then computed. A minimum-length diagnostic-decision-making criterion was developed to assess the diagnostic sensitivity and specificity of the values obtained. The area ([Formula: see text]) under the receiver operating characteristic curve was used to assess the overall diagnostic accuracy of the methods. Using a set of 19 retinal fundus images of preterm infants with plus disease and 91 without plus disease, the proposed methods provided an overall diagnostic accuracy of [Formula: see text]. Using the total length of all abnormally tortuous vessel segments in an image, our techniques are capable of CAD of plus disease with high accuracy without the need for manual selection of vessels to analyze. The proposed methods may be used in a clinical or teleophthalmological setting.

Computer-aided diagnosis of plus disease via measurement of vessel thickness in retinal fundus images of preterm infants.

Changes in the characteristics of retinal vessels such as width and tortuosity can be signs of the presence of several diseases such retinopathy of prematurity (ROP) and diabetic retinopathy. Plus disease is an indicator of ROP which requires treatment and is signified by an increase in posterior venular width. In this work, we present image processing techniques for the detection, segmentation, tracking, and measurement of the width of the major temporal arcade (MTA), which is the thickest venular branch in the retina. Several image processing techniques have been employed, including the use of Gabor filters to detect the MTA, morphological image processing to obtain its skeleton, Canny's method to detect and select MTA vessel-edge candidates, least-squares fitting to interpolate the MTA edges, and geometrical procedures to measure the width of the MTA. The results, obtained using 110 retinal fundus images of preterm infants, indicate a statistically highly significant difference in MTA width of normal cases as compared to cases with plus disease (p<0.01). The results provide good accuracy in computer-aided diagnosis (CAD) of plus disease with an area under the receiver operating characteristic curve of 0.76. The proposed methods may be used in CAD of plus disease and timely treatment of ROP in a clinical or teleophthalmological setting.

Computer-aided diagnosis of plus disease in retinal fundus images of preterm infants via measurement of vessel tortuosity.

An increase in retinal vessel tortuosity can be indicative of the presence of various diseases including retinopathy of prematurity (ROP). Accurate detection and measurement of such changes could help in computer-aided diagnosis of plus disease, which warrants treatment of ROP. We present image processing methods for detection and segmentation of retinal vessels, quantification of vessel tortuosity, and diagnostic-decision-making criteria that incorporate the clinical definition of plus-diagnosis. The obtained results using 110 retinal fundus images of preterm infants (91 without plus and 19 with plus) provide high sensitivity = 0.89 (17/19) and excellent specificity = 0.95 (86/91) in the diagnosis of plus disease.

Quantification of the changes in the openness of the major temporal arcade in retinal fundus images of preterm infants with plus disease.

PURPOSE: We tested the hypothesis that the openness of the major temporal arcade (MTA) changes in the presence of plus disease, by quantification via parabolic modeling of the MTA, as well as measurement of an arcade angle for comparative analysis. Such analysis could assist in the detection and treatment of progressive retinopathy of prematurity. METHODS: Digital image processing techniques were applied for the detection and modeling of the MTA via a graphical user interface (GUI) to quantify the openness of the MTA. An arcade angle measure, based on a previously proposed method, also was obtained via the GUI for comparative analysis. The statistical significance of the differences between the plus and no-plus cases for each parameter was analyzed using the P value. The area (Az) under the receiver operating characteristic curve was used to assess the diagnostic performance of each feature. RESULTS: The temporal arcade angle measure and the openness parameter of the parabolic model were used to perform discrimination of plus versus no-plus cases. Using a set of 19 cases with plus and 91 with no plus disease, Az=0.70 was obtained using the results of dual-parabolic modeling in screening for plus disease. The arcade angle measure provided comparable results with Az=0.73. CONCLUSIONS: Using our proposed image analysis techniques and software, this study demonstrates, for the first time to our knowledge, that the openness of the MTA decreases in the presence of plus disease.

Assessment of vessel tortuosity in retinal images of preterm infants.

Diagnosis of plus disease is crucial for timely treatment and management of retinopathy of prematurity. An indicator of the presence of plus disease is an increase in the tortuosity of blood vessels in the retina. In this work, we propose a new angle-variation-based measure for quantification of tortuosity in retinal fundus images of preterm infants. The methods include the use of Gabor filters to detect vessels as well as to obtain their orientation at each pixel. Morphological image processing methods are used to obtain a skeleton image of the vessels for measurement of tortuosity. Out of 11 vessel segments, marked by an expert ophthalmologist as showing high levels of tortuosity due to plus disease, all were correctly identified using the proposed methods.

Computer-aided diagnosis of proliferative diabetic retinopathy via modeling of the major temporal arcade in retinal fundus images.

Monitoring the openness of the major temporal arcade (MTA) and how it changes over time could facilitate diagnosis and treatment of proliferative diabetic retinopathy (PDR). We present methods for user-guided semiautomated modeling and measurement of the openness of the MTA based on Gabor filters for the detection of retinal vessels, morphological image processing, and a form of the generalized Hough transform for the detection of parabolas. The methods, implemented via a graphical user interface, were tested with retinal fundus images of 11 normal individuals and 11 patients with PDR in the present pilot study on potential clinical application. A method of arcade angle measurement was used for comparative analysis. The results using the openness parameters of single- and dual-parabolic models as well as the arcade angle measurements indicate areas under the receiver operating characteristics of A z = 0.87, 0.82, and 0.80, respectively. The proposed methods are expected to facilitate quantitative analysis of the architecture of the MTA, as well as assist in detection and diagnosis of PDR.

Computer-aided diagnosis of proliferative diabetic retinopathy.

Monitoring the openness of the major temporal arcade (MTA) and how it changes over time could facilitate improved diagnosis and timely treatment of proliferative diabetic retinopathy (PDR). We present methods for user-guided modeling and measurement of the openness of the MTA based on a form of the generalized Hough transform for the detection of parabolas, and to compare it with a method of arcade angle measurement. The methods, implemented via a graphical user interface, were tested with retinal fundus images of 10 normal individuals and 15 patients with PDR. The results using the openness parameters of single- and dual-parabolic models as well as the arcade angle measurements indicate areas under the receiver operating characteristics of A(z)= 0.94, 0.87, and 0.84, respectively. The proposed methods should facilitate improved quantitative analysis of the architecture of the MTA, as well as assist in detection, diagnosis, and improved treatment of PDR.

Fractal analysis of contours of breast masses in mammograms via the power spectra of their signatures.

Contours of benign breast masses and malignant tumors in mammograms differ substantially in their shape and complexity; the former are usually round and smooth, whereas the latter are typically spiculated and irregular. We demonstrate the usefulness of fractal analysis via a frequency domain approach applied to one-dimensional signatures of the two-dimensional contours of breast masses. The 1/ƒ model was applied to power spectra of signatures to estimate the fractal dimension. Tests with a dataset of 111 contours, including those of 65 benign masses and 46 malignant tumors, indicated a high classification performance of 0.89 in terms of the area under the receiver operating characteristic curve.

Detection of the temporal arcade in fundus images of the retina using the Hough transform.

Quantitative analysis of the vascular architecture of the retina can help in monitoring the effects of retinopathy on the visual system. Retinopathy affects the blood vessels in the retina through modification of the shape, width, tortuosity, and the angle of insertion of the temporal arcade. Monitoring the openness of the temporal arcade and changes with treatment can facilitate improved diagnosis and optimized treatment. We propose methods for the detection and parametric modeling of the temporal arcade, including gradient operators and Gabor functions to detect retinal vessels, and the Hough transform to detect parabolic forms. Results obtained with 40 images of the retina indicate accurate to acceptable results for 24 images and partial fits of the parabolic models for 11 images.

Detection of blood vessels in fundus images of the retina using Gabor wavelets.

The monitoring of the effects of diabetes, hypertension, and premature birth on the visual system can be assisted by quantitative analysis of the vascular architecture of the retina. The application of image analysis techniques in ophthalmology becomes possible after the desired features have been detected through the use of an appropriate method. We propose image processing techniques for the detection of blood vessels in the retina. The methods include the design of a bank of directionally sensitive Gabor filters for several values of the scale and elongation parameters. Forty images of the retina from the DRIVE database were used to evaluate the performance of the methods. High efficiency in the detection of blood vessels with the area under the receiver operating characteristics curve of up to 0.96 was achieved.