Accuracy of a deep convolutional neural network in the detection of myopic macular diseases using swept-source optical coherence tomography.

This study examined and compared outcomes of deep learning (DL) in identifying swept-source optical coherence tomography (OCT) images without myopic macular lesions [i.e., no high myopia (nHM) vs. high myopia (HM)], and OCT images with myopic macular lesions [e.g., myopic choroidal neovascularization (mCNV) and retinoschisis (RS)]. A total of 910 SS-OCT images were included in the study as follows and analyzed by k-fold cross-validation (k = 5) using DL's renowned model, Visual Geometry Group-16: nHM, 146 images; HM, 531 images; mCNV, 122 images; and RS, 111 images (n = 910). The binary classification of OCT images with or without myopic macular lesions; the binary classification of HM images and images with myopic macular lesions (i.e., mCNV and RS images); and the ternary classification of HM, mCNV, and RS images were examined. Additionally, sensitivity, specificity, and the area under the curve (AUC) for the binary classifications as well as the correct answer rate for ternary classification were examined. The classification results of OCT images with or without myopic macular lesions were as follows: AUC, 0.970; sensitivity, 90.6%; specificity, 94.2%. The classification results of HM images and images with myopic macular lesions were as follows: AUC, 1.000; sensitivity, 100.0%; specificity, 100.0%. The correct answer rate in the ternary classification of HM images, mCNV images, and RS images were as follows: HM images, 96.5%; mCNV images, 77.9%; and RS, 67.6% with mean, 88.9%.Using noninvasive, easy-to-obtain swept-source OCT images, the DL model was able to classify OCT images without myopic macular lesions and OCT images with myopic macular lesions such as mCNV and RS with high accuracy. The study results suggest the possibility of conducting highly accurate screening of ocular diseases using artificial intelligence, which may improve the prevention of blindness and reduce workloads for ophthalmologists.

Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles.

Conjunctival hyperaemia is a common clinical ophthalmological finding and can be a symptom of various ocular disorders. Although several severity classification criteria have been proposed, none include objective severity criteria. Neural networks and deep learning have been utilised in ophthalmology, but not for the purpose of classifying the severity of conjunctival hyperaemia objectively. To develop a conjunctival hyperaemia grading software, we used 3700 images as the training data and 923 images as the validation test data. We trained the nine neural network models and validated the performance of these networks. We finally chose the best combination of these networks. The DenseNet201 model was the best individual model. The combination of the DenseNet201, DenseNet121, VGG19, and ResNet50 were the best model. The correlation between the multimodel responses, and the vessel-area occupied was 0.737 (p < 0.01). This system could be as accurate and comprehensive as specialists but would be significantly faster and consistent with objective values.

Accuracy of a deep convolutional neural network in detection of retinitis pigmentosa on ultrawide-field images.

Evaluating the discrimination ability of a deep convolution neural network for ultrawide-field pseudocolor imaging and ultrawide-field autofluorescence of retinitis pigmentosa. In total, the 373 ultrawide-field pseudocolor and ultrawide-field autofluorescence images (150, retinitis pigmentosa; 223, normal) obtained from the patients who visited the Department of Ophthalmology, Tsukazaki Hospital were used. Training with a convolutional neural network on these learning data objects was conducted. We examined the K-fold cross validation (K = 5). The mean area under the curve of the ultrawide-field pseudocolor group was 0.998 (95% confidence interval (CI) [0.9953-1.0]) and that of the ultrawide-field autofluorescence group was 1.0 (95% CI [0.9994-1.0]). The sensitivity and specificity of the ultrawide-field pseudocolor group were 99.3% (95% CI [96.3%-100.0%]) and 99.1% (95% CI [96.1%-99.7%]), and those of the ultrawide-field autofluorescence group were 100% (95% CI [97.6%-100%]) and 99.5% (95% CI [96.8%-99.9%]), respectively. Heatmaps were in accordance with the clinician's observations. Using the proposed deep neural network model, retinitis pigmentosa can be distinguished from healthy eyes with high sensitivity and specificity on ultrawide-field pseudocolor and ultrawide-field autofluorescence images.

Accuracy of ultrawide-field fundus ophthalmoscopy-assisted deep learning for detecting treatment-naïve proliferative diabetic retinopathy.

PURPOSE: We investigated using ultrawide-field fundus images with a deep convolutional neural network (DCNN), which is a machine learning technology, to detect treatment-naïve proliferative diabetic retinopathy (PDR). METHODS: We conducted training with the DCNN using 378 photographic images (132 PDR and 246 non-PDR) and constructed a deep learning model. The area under the curve (AUC), sensitivity, and specificity were examined. RESULT: The constructed deep learning model demonstrated a high sensitivity of 94.7% and a high specificity of 97.2%, with an AUC of 0.969. CONCLUSION: Our findings suggested that PDR could be diagnosed using wide-angle camera images and deep learning.

Deep-learning classifier with ultrawide-field fundus ophthalmoscopy for detecting branch retinal vein occlusion.

AIM: To investigate and compare the efficacy of two machine-learning technologies with deep-learning (DL) and support vector machine (SVM) for the detection of branch retinal vein occlusion (BRVO) using ultrawide-field fundus images. METHODS: This study included 237 images from 236 patients with BRVO with a mean±standard deviation of age 66.3±10.6y and 229 images from 176 non-BRVO healthy subjects with a mean age of 64.9±9.4y. Training was conducted using a deep convolutional neural network using ultrawide-field fundus images to construct the DL model. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the curve (AUC) were calculated to compare the diagnostic abilities of the DL and SVM models. RESULTS: For the DL model, the sensitivity, specificity, PPV, NPV and AUC for diagnosing BRVO was 94.0% (95%CI: 93.8%-98.8%), 97.0% (95%CI: 89.7%-96.4%), 96.5% (95%CI: 94.3%-98.7%), 93.2% (95%CI: 90.5%-96.0%) and 0.976 (95%CI: 0.960-0.993), respectively. In contrast, for the SVM model, these values were 80.5% (95%CI: 77.8%-87.9%), 84.3% (95%CI: 75.8%-86.1%), 83.5% (95%CI: 78.4%-88.6%), 75.2% (95%CI: 72.1%-78.3%) and 0.857 (95%CI: 0.811-0.903), respectively. The DL model outperformed the SVM model in all the aforementioned parameters (P<0.001). CONCLUSION: These results indicate that the combination of the DL model and ultrawide-field fundus ophthalmoscopy may distinguish between healthy and BRVO eyes with a high level of accuracy. The proposed combination may be used for automatically diagnosing BRVO in patients residing in remote areas lacking access to an ophthalmic medical center.

MORPHOLOGIC CHARACTERISTICS OF MACULAR HOLE AND MACULAR HOLE RETINAL DETACHMENT ASSOCIATED WITH EXTREME MYOPIA.

PURPOSE: To investigate the surgical results and morphologic characteristics of macular hole (MH) and macular hole retinal detachment (MHRD) associated with extreme myopia. METHODS: We retrospectively reviewed consecutive cases with axial length ≥28 mm who were treated with pars plana vitrectomy for MH or MHRD. The choroidal and scleral thickness at the fovea, presence of dome-shaped macula, and the height of posterior staphyloma 3 mm from the fovea were measured from postoperative optical coherence tomography images. RESULTS: Significant improvement in visual acuity was obtained postoperatively in both MH (16 eyes; 15 patients) and MHRD (19 eyes; 18 patients) groups (P < 0.05). Final MH closure rate was not significantly different between the groups (MH: 15/16, MHRD: 14/19, P = 0.19). Axial length was not significantly different between the groups (MH: 30.5 ± 1.5 mm, MHRD: 29.6 ± 1.3 mm, P = 0.098). Eyes with MH had significantly greater choroidal thickness (MH: 61.9 ± 66.0 µm, MHRD: 24.1 ± 19.8 µm, P = 0.045), greater scleral thickness (MH: 294 ± 77 µm, MHRD: 232 ± 89 µm, P = 0.008), higher frequency of dome-shaped macula (MH: 6/16, MHRD: 1/19, P = 0.032), and lower staphyloma height (MH: 190 ± 113 µm, MHRD: 401 ± 156 µm, P < 0.001). CONCLUSION: Surgical outcomes were generally favorable. The pathogenetic differences between the two conditions may be attributable to differences with respect to eye morphology.

Accuracy of ultra-wide-field fundus ophthalmoscopy-assisted deep learning, a machine-learning technology, for detecting age-related macular degeneration.

PURPOSE: To predict exudative age-related macular degeneration (AMD), we combined a deep convolutional neural network (DCNN), a machine-learning algorithm, with Optos, an ultra-wide-field fundus imaging system. METHODS: First, to evaluate the diagnostic accuracy of DCNN, 364 photographic images (AMD: 137) were amplified and the area under the curve (AUC), sensitivity and specificity were examined. Furthermore, in order to compare the diagnostic abilities between DCNN and six ophthalmologists, we prepared yield 84 sheets comprising 50% of normal and wet-AMD data each, and calculated the correct answer rate, specificity, sensitivity, and response times. RESULTS: DCNN exhibited 100% sensitivity and 97.31% specificity for wet-AMD images, with an average AUC of 99.76%. Moreover, comparing the diagnostic abilities of DCNN versus six ophthalmologists, the average accuracy of the DCNN was 100%. On the other hand, the accuracy of ophthalmologists, determined only by Optos images without a fundus examination, was 81.9%. CONCLUSION: A combination of DCNN with Optos images is not better than a medical examination; however, it can identify exudative AMD with a high level of accuracy. Our system is considered useful for screening and telemedicine.

Comparison between support vector machine and deep learning, machine-learning technologies for detecting epiretinal membrane using 3D-OCT.

PURPOSE: In this study, we compared deep learning (DL) with support vector machine (SVM), both of which use three-dimensional optical coherence tomography (3D-OCT) images for detecting epiretinal membrane (ERM). METHODS: In total, 529 3D-OCT images from the Tsukazaki hospital ophthalmology database (184 non-ERM subjects and 205 ERM patients) were assessed; 80% of the images were divided for training, and 20% for test as follows: 423 training (non-ERM 245, ERM 178) and 106 test (non-ERM 59, ERM 47) images. Using the 423 training images, a model was created with deep convolutional neural network and SVM, and the test data were evaluated. RESULTS: The DL model's sensitivity was 97.6% [95% confidence interval (CI), 87.7-99.9%] and specificity was 98.0% (95% CI, 89.7-99.9%), and the area under the curve (AUC) was 0.993 (95% CI, 0.993-0.994). In contrast, the SVM model's sensitivity was 97.6% (95% CI, 87.7-99.9%), specificity was 94.2% (95% CI, 84.0-98.7%), and AUC was 0.988 (95% CI, 0.987-0.988). CONCLUSION: DL model is better than SVM model in detecting ERM by using 3D-OCT images.

Deep Neural Network-Based Method for Detecting Central Retinal Vein Occlusion Using Ultrawide-Field Fundus Ophthalmoscopy.

The aim of this study is to assess the performance of two machine-learning technologies, namely, deep learning (DL) and support vector machine (SVM) algorithms, for detecting central retinal vein occlusion (CRVO) in ultrawide-field fundus images. Images from 125 CRVO patients (n=125 images) and 202 non-CRVO normal subjects (n=238 images) were included in this study. Training to construct the DL model using deep convolutional neural network algorithms was provided using ultrawide-field fundus images. The SVM uses scikit-learn library with a radial basis function kernel. The diagnostic abilities of DL and the SVM were compared by assessing their sensitivity, specificity, and area under the curve (AUC) of the receiver operating characteristic curve for CRVO. For diagnosing CRVO, the DL model had a sensitivity of 98.4% (95% confidence interval (CI), 94.3-99.8%) and a specificity of 97.9% (95% CI, 94.6-99.1%) with an AUC of 0.989 (95% CI, 0.980-0.999). In contrast, the SVM model had a sensitivity of 84.0% (95% CI, 76.3-89.3%) and a specificity of 87.5% (95% CI, 82.7-91.1%) with an AUC of 0.895 (95% CI, 0.859-0.931). Thus, the DL model outperformed the SVM model in all indices assessed (P < 0.001 for all). Our data suggest that a DL model derived using ultrawide-field fundus images could distinguish between normal and CRVO images with a high level of accuracy and that automatic CRVO detection in ultrawide-field fundus ophthalmoscopy is possible. This proposed DL-based model can also be used in ultrawide-field fundus ophthalmoscopy to accurately diagnose CRVO and improve medical care in remote locations where it is difficult for patients to attend an ophthalmic medical center.

Accuracy of deep learning, a machine learning technology, using ultra-wide-field fundus ophthalmoscopy for detecting idiopathic macular holes.

We aimed to investigate the detection of idiopathic macular holes (MHs) using ultra-wide-field fundus images (Optos) with deep learning, which is a machine learning technology. The study included 910 Optos color images (715 normal images, 195 MH images). Of these 910 images, 637 were learning images (501 normal images, 136 MH images) and 273 were test images (214 normal images and 59 MH images). We conducted training with a deep convolutional neural network (CNN) using the images and constructed a deep-learning model. The CNN exhibited high sensitivity of 100% (95% confidence interval CI [93.5-100%]) and high specificity of 99.5% (95% CI [97.1-99.9%]). The area under the curve was 0.9993 (95% CI [0.9993-0.9994]). Our findings suggest that MHs could be diagnosed using an approach involving wide angle camera images and deep learning.

Accuracy of deep learning, a machine-learning technology, using ultra-wide-field fundus ophthalmoscopy for detecting rhegmatogenous retinal detachment.

Rhegmatogenous retinal detachment (RRD) is a serious condition that can lead to blindness; however, it is highly treatable with timely and appropriate treatment. Thus, early diagnosis and treatment of RRD is crucial. In this study, we applied deep learning, a machine-learning technology, to detect RRD using ultra-wide-field fundus images and investigated its performance. In total, 411 images (329 for training and 82 for grading) from 407 RRD patients and 420 images (336 for training and 84 for grading) from 238 non-RRD patients were used in this study. The deep learning model demonstrated a high sensitivity of 97.6% [95% confidence interval (CI), 94.2-100%] and a high specificity of 96.5% (95% CI, 90.2-100%), and the area under the curve was 0.988 (95% CI, 0.981-0.995). This model can improve medical care in remote areas where eye clinics are not available by using ultra-wide-field fundus ophthalmoscopy for the accurate diagnosis of RRD. Early diagnosis of RRD can prevent blindness.

Axial length changes in highly myopic eyes and influence of myopic macular complications in Japanese adults.

PURPOSE: To investigate changes of the axial length in normal eyes and highly myopic eyes and influence of myopic macular complications in Japanese adults. STUDY DESIGN: Retrospective longitudinal case series. METHODS: The changes in the axial length of 316 eyes from 316 patients (mean age, 63.8 ± 9.0 years; range, 34-82; 240 females) examined using IOLMaster with a follow-up period of at least 1 year were studied. This study included 85 non-highly myopic eyes (|refractive error| ≤ 5 diopters; 63 females; non-highly myopic group), 165 highly myopic eyes (refractive error ≤ -6 diopters or axial length ≥ 26 mm; 124 females) without macular complications (no complications group), 32 eyes (25 females) with myopic traction maculopathy (MTM group), and 34 eyes (28 females) with myopic choroidal neovascularization (CNV group). RESULTS: All groups showed a significant increase in the axial length during the follow-up period (mean follow-up, 28.7 ± 16.8 months; range, 12-78) (P < 0.01). Changes in the axial length/year in the no complications group (0.041 ± 0.05 mm) were significantly greater than those in the non-highly myopic group (0.007 ± 0.02 mm) (P < 0.0001). Furthermore, changes in the CNV group (0.081 ± 0.04 mm) were significantly greater than those in the no complications (P < 0.0001) and MTM (0.040 ± 0.05 mm) (P = 0.0059) groups, whereas no significant difference was found between the changes in the MTM and no complications groups (P = 0.91). Multiple regression analyses indicated that CNV eyes (P < 0.0001) and female patients' eyes (P = 0.04) showed greater changes in the axial length/year. CONCLUSIONS: All groups showed an increase in the axial length, which was greater for highly myopic eyes. In particular, CNV eyes showed greater increases, indicating that larger changes in the axial length may require careful follow-up.

Morphologic characteristics of macular complications of a dome-shaped macula determined by swept-source optical coherence tomography.

PURPOSE: To investigate the morphologic characteristics of macular complications of dome-shaped maculas using swept-source optical coherence tomography (OCT). DESIGN: Retrospective observational case series. METHODS: Axial length measurements and swept-source OCT were performed in 49 highly myopic eyes (in 5 male and 30 female subjects) with dome-shaped maculas. We classified the dome patterns and measured the central retinal thickness, central choroidal thickness, central scleral thickness, and the macular bulge height, and assessed the associations of these parameters with macular complications. RESULTS: The central scleral thickness was significantly negatively correlated with age and the axial length. We classified the eyes into 3 groups: 6 with choroidal neovascularization (CNV group), 8 with retinal pigment epithelial detachment (PED group; 5 with serous retinal detachment), and 35 with no complications (no complications group). Nine eyes had a round dome and 40 had horizontally oriented oval-shaped domes. There were no significant differences in the frequency of macular complications between these patterns. The CNV group was significantly older and had a longer axial length than the other groups. The PED group had significantly larger values for both the central scleral thickness and bulge height than the other groups. The central choroidal thickness was significantly thinner in the CNV group than in the no complications group. CONCLUSION: A dome-shaped macula results from relative thickening of the macular sclera, and this may lead to PED. Thinning of the sclera owing to long-term changes and elongation of the axis may develop CNV and cause visual impairment.

Changes in choroidal thickness after cataract surgery.

PURPOSE: To evaluate changes in choroidal thickness before and after cataract surgery and factors affecting the changes. SETTING: Tsukazaki Hospital, Himeji, Japan. DESIGN: Prospective interventional study. METHODS: Patients having cataract surgery without other eye pathology were studied. The corrected distance visual acuity (CDVA), intraocular pressure (IOP), axial length (AL), and enhanced-depth-imaging optical coherence tomography (OCT) were measured preoperatively. The choroidal thickness was measured at 5 points (subfoveal and 1.5 mm nasal, temporal, superior, and inferior to the fovea) using the OCT device's software. Enhanced-depth-imaging OCT and IOP measurements were obtained 3 days, 1 and 3 weeks, and 3 and 6 months postoperatively and compared with the baseline values. Stepwise analysis determined which factors (ie, age, CDVA, preoperative IOP, AL, operative time, changes in IOP) were associated with changes in choroidal thickness. RESULTS: One hundred eyes were analyzed. The postoperative IOP significantly decreased at 3 weeks, 3 months, and 6 months. The postoperative choroidal thickness significantly increased at the foveal and inferior regions throughout the follow-up; at the nasal region at 3 days, 1 week, and 6 months; at the temporal region at 1 week; and at the superior region at 6 months. These changes negatively correlated with those in IOP early after surgery. The changes in choroidal thickness later negatively correlated with the AL in all regions. CONCLUSION: Cataract surgery caused changes in choroidal thickness. The AL and changes in the IOP are critical for evaluating the changes in choroidal thickness. FINANCIAL DISCLOSURE: No author has a financial or proprietary interest in any material or method mentioned.

Comparison of visual performance of multifocal intraocular lenses with same material monofocal intraocular lenses.

PURPOSE: To compare the visual performance of multifocal intraocular lenses (IOLs) and monofocal IOLs made of the same material. METHODS: The subjects included patients implanted with either Tecnis® monofocal IOLs (ZA9003 or ZCB00) or Tecnis® multifocal IOLs (ZMA00 or ZMB00) bilaterally. We conducted a retrospective study comparing the two types of IOLs. The multifocal group included 46 patients who were implanted with Tecnis® multifocal IOLs bilaterally. The monofocal group was an age- and sex-matched control group, and included 85 patients who were implanted with Tecnis® monofocal IOLs bilaterally. Lens opacity grading, the radius of corneal curvature, corneal astigmatism, axial length and the refractive status were measured preoperatively. Pupil size, ocular aberrometry, distance, intermediate and near visual acuity, contrast sensitivity with and without glare and the responses to a quality-of-vision questionnaire were evaluated pre- and postoperatively. RESULTS: The uncorrected near visual acuity was significantly better in the multifocal group, whereas both the corrected intermediate and near visual acuity were better in the monofocal group. Contrast sensitivity (with and without glare) was significantly better in the monofocal group. The rate of spectacle dependency was significantly lower in the multifocal group. There were no significant differences between the two groups regarding most items of the postoperative quality-of-vision questionnaire (VFQ-25), with the exception that the patients in the monofocal group reported fewer problems with nighttime driving. CONCLUSIONS: The multifocal IOLs used in this study reduced spectacle dependency more so than monofocal IOLs and did not compromise the subjective visual function, with the exception of nighttime driving.

3-D choroidal thickness maps from EDI-OCT in highly myopic eyes.

PURPOSE: Myopic chorioretinal atrophy is a debilitating condition that causes severe loss of primary vision. However, its mechanisms and pathologic course are not well understood. We performed volumetric measurements of the posterior choroid via three-dimensional analysis of the choroid in patients with high myopia to understand its structure, and we compared the measurements with those of normal subjects. METHODS: Twenty-five highly myopic but otherwise normal eyes and 25 nonmyopic eyes were evaluated. Enhanced depth imaging optical coherence tomography (EDI-OCT) was performed using 20 × 20-degree raster scans consisting of 25 high-speed line scans. Three-dimensional retinal and choroidal thickness maps were produced from the EDI-OCT data. For the quantitative analyses, the macula was divided into nine regions, as defined by the Early Treatment Diabetic Retinopathy Study (ETDRS) layout, and the mean retinal and choroidal thicknesses of each region were obtained. RESULTS: The choroidal thicknesses at all regions in the high-myopia group were significantly smaller than those in the normal refractive group (p < 0.0001). The foveal choroidal thickness was the greatest in the normal group but not in the high-myopia group. In the high-myopia group, the choroidal thickness at the fovea was significantly greater than that at the outer nasal quadrants (p < 0.0001) but significantly smaller than that at the outer superior (p < 0.0001) quadrants. CONCLUSIONS: Three-dimensional choroidal thickness maps obtained via EDI-OCT are useful for quantifying choroid thickness in subjects with high myopia more accurately.

Effects of corneal thickness and axial length on intraocular pressure and ocular pulse amplitude before and after cataract surgery.

OBJECTIVE: To investigate the relationship between the biophysical properties of the cornea and eye on the intraocular pressure (IOP) and ocular pulse amplitude (OPA) before and after cataract surgery. DESIGN: Intervention study. PARTICIPANTS: The left eyes of 311 patients. METHODS: The left eyes of 338 patients undergoing cataract surgery without other eye pathology were studied. IOP and OPA were recorded by dynamic contour tonometry (DCT) 1 week before and 14 weeks after cataract surgery. The axial length, corneal curvature, central corneal thickness, anterior chamber depth, and anterior chamber angle were measured 1 week before cataract surgery. Multiple regression analyses of these factors to the preoperative OPA were performed. The difference between the pre- and postoperative IOP and OPA were investigated by paired t tests. RESULTS: Three hundred and eleven of 338 eyes were analyzed. The preoperative OPA was negatively correlated with axial length (ß = -0.24, p < 0.0001) and positively correlated with the preoperative IOP (ß = 0.13, p < 0.0001). The average OPA was significantly decreased after cataract surgery (p < 0.0001). The mean change in postoperative OPA was -0.45 ± 0.63 mm Hg (95% CI -0.52 to -0.38 mm Hg). CONCLUSIONS: The preoperative OPA was negatively correlated with axial length as reported. A significant decrease in OPA was observed after the cataract surgery.