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.

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.

The present study aims to describe the use of machine learning (ML) in predicting the occurrence of postoperative refraction after cataract surgery and compares the accuracy of this method to conventional intraocular lens (IOL) power calculation formulas. In total, 3331 eyes from 2010 patients were assessed. The objects were divided into training data and test data. The constants for the IOL power calculation formulas and model training for ML were optimized using training data. Then, the occurrence of postoperative refraction was predicted using conventional formulas, or ML models were calculated using the test data. We evaluated the SRK/T formula, Haigis formula, Holladay 1 formula, Hoffer Q formula, and Barrett Universal II formula (BU-II); similar to ML methods, we assessed support vector regression (SVR), random forest regression (RFR), gradient boosting regression (GBR), and neural network (NN). Among the conventional formulas, BU-II had the lowest mean and median absolute error of prediction. Therefore, we compared the accuracy of our method with that of BU-II. The absolute errors of some ML methods were lower than those of BU-II. However, no statistically significant difference was observed. Thus, the accuracy of our method was not inferior to that of BU-II.

Surgical skill levels of young ophthalmologists tend to be instinctively judged by ophthalmologists in practice, and hence a stable evaluation is not always made for a single ophthalmologist. Although it has been said that standardizing skill levels presents difficulty as surgical methods vary greatly, approaches based on machine learning seem to be promising for this objective. In this study, we propose a method for displaying the information necessary to quantify the surgical techniques of cataract surgery in real-time. The proposed method consists of two steps. First, we use InceptionV3, an image classification network, to extract important surgical phases and to detect surgical problems. Next, one of the segmentation networks, scSE-FC-DenseNet, is used to detect the cornea and the tip of the surgical instrument and the incisional site in the continuous curvilinear capsulorrhexis, a particularly important phase in cataract surgery. The first and second steps are evaluated in terms of the area under curve (i.e., AUC) of the figure of the true positive rate versus (1—false positive rate) and the intersection over union (i.e., IoU) obtained by the ground truth and prediction associated with the region of interest. As a result, in the first step, the network was able to detect surgical problems with an AUC of 0.97. In the second step, the detection rate of the cornea was 99.7% when the IoU was 0.8 or more, and the detection rates of the tips of the forceps and the incisional site were 86.9% and 94.9% when the IoU was 0.1 or more, respectively. It was thus expected that the proposed method is one of the basic techniques to achieve the standardization of surgical skill levels.

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 796 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, 107 images; HM, 456 images; mCNV, 122 images; and RS, 111 images (n = 796). 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, 0.983; sensitivity, 0.953; specificity, 0.940. The classification results of HM images and images with myopic macular lesions were as follows: AUC, 0.976; sensitivity, 0.940; specificity, 0.941. The correct answer rate in the ternary classification of HM images, mCNV images, and RS images were as follows: HM images, 93.7%; mCNV images, 82.4%; and RS, 92.3% with mean, 91.4%. 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.

In this paper, a checking system for eye lotion instillations of ophthalmology patients is proposed, using a neural network. It first estimates tilt angles of the eye dropper bottle from acceleration values measured by a triaxial sensor. A sequence of such estimated values is next presented to a discrimination model after network learning, and its output value is considered to be a certainty degree indicating whether an eye lotion is applied at the time zone corresponding to the sequence. The final judgement for the instillation is made by conducting thresholding of the certainty degree. The proposed method employs either a long short-term memory (LSTM for short) or a bidirectional long short-term memory (B_LSTM for short) to construct the discrimination model. Experimental results using practical data reveal that the B_LSTM-based model achieves favorable metric values compared to the LSTM-based model. In addition, it is explored whether the metric values depend on the interval between the time points when the tilt values are measured.