
Citation: Vandana Gulati, Mansi Dass Singh, Pankaj Gulati. Role of mushrooms in gestational diabetes mellitus[J]. AIMS Medical Science, 2019, 6(1): 49-66. doi: 10.3934/medsci.2019.1.49
[1] | Tej Bahadur Shahi, Cheng-Yuan Xu, Arjun Neupane, William Guo . Machine learning methods for precision agriculture with UAV imagery: a review. Electronic Research Archive, 2022, 30(12): 4277-4317. doi: 10.3934/era.2022218 |
[2] | Abul Bashar . Employing combined spatial and frequency domain image features for machine learning-based malware detection. Electronic Research Archive, 2024, 32(7): 4255-4290. doi: 10.3934/era.2024192 |
[3] | Xiaomeng An, Sen Xu . A selective evolutionary heterogeneous ensemble algorithm for classifying imbalanced data. Electronic Research Archive, 2023, 31(5): 2733-2757. doi: 10.3934/era.2023138 |
[4] | Yixin Sun, Lei Wu, Peng Chen, Feng Zhang, Lifeng Xu . Using deep learning in pathology image analysis: A novel active learning strategy based on latent representation. Electronic Research Archive, 2023, 31(9): 5340-5361. doi: 10.3934/era.2023271 |
[5] | Mashael S. Maashi, Yasser Ali Reyad Ali, Abdelwahed Motwakel, Amira Sayed A. Aziz, Manar Ahmed Hamza, Amgad Atta Abdelmageed . Anas platyrhynchos optimizer with deep transfer learning-based gastric cancer classification on endoscopic images. Electronic Research Archive, 2023, 31(6): 3200-3217. doi: 10.3934/era.2023162 |
[6] | Jiawang Li, Chongren Wang . A deep learning approach of financial distress recognition combining text. Electronic Research Archive, 2023, 31(8): 4683-4707. doi: 10.3934/era.2023240 |
[7] | Ju Wang, Leifeng Zhang, Sanqiang Yang, Shaoning Lian, Peng Wang, Lei Yu, Zhenyu Yang . Optimized LSTM based on improved whale algorithm for surface subsidence deformation prediction. Electronic Research Archive, 2023, 31(6): 3435-3452. doi: 10.3934/era.2023174 |
[8] | Shixiong Zhang, Jiao Li, Lu Yang . Survey on low-level controllable image synthesis with deep learning. Electronic Research Archive, 2023, 31(12): 7385-7426. doi: 10.3934/era.2023374 |
[9] | Xiao Chen, Fuxiang Li, Hairong Lian, Peiguang Wang . A deep learning framework for predicting the spread of diffusion diseases. Electronic Research Archive, 2025, 33(4): 2475-2502. doi: 10.3934/era.2025110 |
[10] | Jianting Gong, Yingwei Zhao, Xiantao Heng, Yongbing Chen, Pingping Sun, Fei He, Zhiqiang Ma, Zilin Ren . Deciphering and identifying pan-cancer RAS pathway activation based on graph autoencoder and ClassifierChain. Electronic Research Archive, 2023, 31(8): 4951-4967. doi: 10.3934/era.2023253 |
Common macular and vascular diseases include age-related macular degeneration (ARMD), diabetic macular edema (DME), branch retinal vein occlusion (BRVO), central retinal vein occlusion (CRVO), and central serous chorioretinopathy (CSCR), which are the leading causes of visual impairment and blindness worldwide [1,2,3]. According to the World Health Organization (WHO), DME, which primarily affects working-age adults, affected 425 million people worldwide in 2017 and is expected to affect 629 million people by 2045 [4]. The WHO also estimates that 196 million people had ARMD in 2020; this number is expected to rise to 288 million by 2040 [5]. The prevalence of ARMD in elderly people is 40% at the age of 70 years, rising to 70% at the age of 80 years. Rogers et al. [6] discovered that BRVO and CRVO affected 13.9 million and 2.5 million of the world's population aged 30 years and older, respectively, in 2008. Men have a higher prevalence of CSCR compared to women [7]. A large population is afflicted by these diseases, and projections suggest that this number will escalate in the future. However, the first stage of these diseases can be treated, and patients can recover their vision loss through early detection and treatment [8,9,10].
Optical coherence tomography (OCT) is a noninvasive imaging modality that provides high-resolution information within a cross sectional area. OCT retinal imaging enables visualization of the thickness, structure, and detail of various layers of the retina. In addition, when the retina develops a disease, OCT enables the visualization of abnormal features and damaged retinal structures [11]. Therefore, retinal OCT images are widely used in the medical field to monitor information in medical images prior to treatment or for the diagnosis of various diseases.
For several years, ophthalmologists have analyzed the comprehensive information inside the retina for retinal care services, treatment, and diagnosis using retinal OCT images in clinical settings. The clinician performs these tasks manually and wait for each process. As a result, manual analysis is time consuming when there are numerous OCT images. Even if the clinician has great expertise, this analysis may not be accurate [12]. An automated technique based on deep learning (DL) or machine learning using artificial intelligence has been proposed as a solution to overcome this limitation.
Recently, computer algorithms based on artificial intelligence, DL, and machine learning have been proposed for the automatic diagnosis of various retinal diseases and have been applied in clinical health care. Han et al. [13] modified three well known convolutional neural network (CNN) models to gain access to normal and three subtypes of neovascular age-related macular degeneration (nAMD). The classification layers of the original CNN models were replaced by new layers: four fully connected layers and three dropout layers, along with a Leaky rectified linear activation unit (ReLU) as an activation function. The modified models were trained using the transfer learning technique and tested on 920 OCT images; the VGG-16 model achieved an accuracy of 87.4%. Sotoudeh-Paima et al. [14] classified OCT images to identify normal, AMD, and choroidal neovascularization (CNV) using a multiscale CNN. This CNN was evaluated and achieved a classification accuracy of 93.40% on the public dataset. Elaziz et al. [15] developed a four-class classification method for accessing retinal diseases from OCT images based on an ensemble DL model and machine learning. First, the features are extracted from the two models, MobileNet and DenseNet, and were concatenated as full features of the input images. Then, feature selection was performed to remove irrelevant features and to input the useful features into machine learning to classify the input data. A total of 968 OCT images were used to evaluate classification performance, and an accuracy of 94.31% was achieved. Another study by Liu et al. [16] used a DL model to extract attention features from OCT images. It used the extracted features as guiding features for CNV, DME, drusen, and normal. The classification performance was assessed using public datasets, and an average accuracy of 95.10% was achieved. Minagi et al. [17] used transfer learning with universal adversarial perturbations (UAPs) for classification with a limited dataset. Three types of medical images, including OCT images, were used to assess diseases, and DL models were trained using the ImageNet dataset. The UAPs algorithm was used to generate a training set based on the data provided to train the DL model. There were 11,200 OCT images utilized in training and assessing the model's performance, and a classification accuracy of 95.3% was achieved for the four classes: CNV, DME, drusen, and normal. Tayal et al. [18] presented four ocular disease classifications based on three CNN models using OCT images. Images were enhanced before being fed to CNN models. To assess the performance of the presented method, 6,678 publicly available OCT images were evaluated. The method achieved an accuracy of 96.50% with a CNN model which compressed nine layers. The performance of the CNN models with nine layers outperformed the experimented CNN models with five and seven layers. Adversarial retraining is an algorithm used to improve the performance of DL models based on classification.
According to the literature, retinal OCT classification was developed using DL and DL based methods such as transfer learning, smoothing generative adversarial networks, adversarial retraining, and multi-scale CNN. This method is used to improve the model's performance by fine-tuning previous task knowledge using the OCT image problem, increasing the dataset size for training, applying the technique of inputting data for the training model, and changing the training input image sizes. However, the classification methods can achieve an accuracy of less than 97.00%, indicating their potential for further improvement. Moreover, these studies classify retinal diseases into fewer than five classes. This study aims to improve the classification accuracy and detect five classes of retinal diseases, which are more than the previous studies highlighted in the literature.
In this study, we propose an automatic method based on a hybrid of deep learning and ensemble machine learning for screening five different retinal diseases from OCT images to improve the performance of OCT image classification. The proposed method improves classification accuracy, outperforming standalone classifiers without a hybrid. In addition, it can be trained using a smaller dataset from our hospital, which has been strictly labelled by experts. Moreover, the proposed method enables deployment with a web server for open access to test the evaluation performance within seconds.
All OCT images were collected from Soonchunhyang University's Bucheon Hospital. The OCT images were collected and normalized after approval by the Bucheon Hospital's Institutional Review Board (IRB). OCT images were captured using DRI-OCT (Topcon Medical System, Inc., Oakland, NJ, USA). The scan range was 3–12 mm in the horizontal and vertical directions, with a lateral resolution of 20 μm and an in-depth resolution of 8 μm. The shooting speed was 100,000 A-scans per second. The OCT images utilized were collected twice; the first comprised 2,000 images that were captured between April and September 2021, while the second consisted of 998 images, and took place over a period of approximately five months from September 2021 to January 2022. Therefore, the total number of OCT images collected twice was 2,998; these were labeled by ophthalmologists for five retinal diseases (ARMD:740, BRVO:450, CRVO:299, CSCR:749, and DME:760) as the ground truth.
This study was approved by the Institutional Review Board (IRB) from Soonchunhyang University Bucheon Hospital, Bucheon, Republic of Korea (IRB approval number: 2021-05-001). All methods were performed in accordance with relevant guidelines and regulations. Informed consent was obtained from all subjects.
Image processing is a technique for performing various operations on the original images to convert it into a format suitable for DL models or to extract useful features. In image classification based on deep learning, image processing is an essential initial process to change an image before feeding it to the CNN model. The CNN model requires a unique size for the image input, and higher-resolution images demand longer computing times. To shorten the operating time and the suitable size required by the CNN models, all OCT images were downsized to 300 pixels in height and 500 pixels in width. The OCT image dataset was split into an 80% training set and 20% testing set. The training set was used to train the deep learning model and the testing set was used to assess performance.
The size of the dataset has a significant impact on the DL performance. Therefore, a larger dataset may enable a better performance. However, in the medical field, most medical dataset has size limits. Data augmentation is a technique developed to overcome the limitations of a dataset by performing different operations on the data provided and creating new data, thereby enhancing the dataset size. Additionally, data augmentation is used to enhance performance [19], generalize the model [20], and avoid overfitting [21]. We utilize data augmentation techniques from the Python library imgaug including like vertical flip, rotation, scale, brightness, saturation, contrast, enhance and contrast, and equalization. The OCT images were augmented at angles of 170,175,185, and 190. The selected angle is suitable for rectangle shape representation without loss of information from the original OCT images; scale image with a random range between 0.01 to 0.12; the level of brightness from 1 to 3; the saturate operation, which ranges from 1 to 5, increases by one with each level; random contrast with contrast values ranging from 0.2 to 3; enhance and contrast with levels ranging from 1 to 1.5; and image equalization with levels ranging from 0.9 to 1.4. At the end of the data augmentation process, one OCT image can serve as the basic for generating 29 augmented images. Therefore, the training set comprised a total of 69,455 OCT images, including samples. The acquired OCT and augmented images are shown in Figure 1. Applying data augmentation, only the training set is used for training the proposed method. After finishing the augmentation operation, the OCT images are passed through the 10-fold cross-validation technique to partition the data into folds for the training model (training data) and to test the model after finishing every epoch (validation data).
Figure 2 shows the architecture of the proposed method that comprises three significant blocks: feature extraction, classification, and boosting performance. First, transfer learning based on CNN models extracts one thousand features from the OCT images. Second, various machine learning algorithms are used to classify the OCT images based on the features extracted by the CNN model. Finally, the ensemble algorithm fuses the distribution probabilities of the same class and predicts the retinal disease class based on probability. Each block of the proposed architecture is described in detail in the following subsections.
Transfer learning is a technique used to transform the knowledge of a related task that has already been studied to improve the learning of a new task. Training a CNN model from scratch is computationally expensive and time consuming; moreover, an extensive dataset is required to achieve a better performance. Therefore, transfer learning has been developed to overcome DL's drawbacks [22]. To retrain the model with new tasks based on prior knowledge, pretrain was refined, small top layers were trained, and the final layers were frozen. In this study, the transfer learning CNN (TL-CNN) models EfficientNetB0 [23], InceptionResNetV2 [24], InceptionV3 [25], ResNet50 [26], VGG16 [27], and VGG19 [28] are selected and updated. The modification names of the CNN models start with TL, indicating transfer learning, and ends with the original names of the CNN models, including TL-EfficientNetB0, TL -InceptionResNetV2, TL-InceptionV3, TL-ResNet50, TL-VGG16, and TL-VGG19. The original CNN models were created for generic image classification tasks. They were trained and tested on a large dataset (ImageNet) to categorize 1000 different types of images. To use a CNN model with the transfer learning technique and classify retinal OCT images, each CNN model must modify its classification layers to adapt to the target classes. One specific problem is the categorization of OCT images. The new classification layer is modified with continued stacking of GlobalAveragePooling2D, one Normalization layer, and two Dense layers. The first Dense layer consists of 1,024 with the ReLU activation function and the final dense layer with a five output- vectors. Finally, the updated model is pretrained, and the pretrain model is retrained to fine-tune the previous feature representation in the base model to make it more relevant for OCT image classification. The output consists of five vectors representing the distribution class probabilities using the Softmax activation function. As mentioned previously, a CNN model based on transfer learning is used to extract convolutional features from the OCT images. Therefore, the convolutional features were extracted from the TL-CNNs models where the GlobalAveragePooling2D layers of the classification layer. These features are one-dimensional. Different models provide various features and numbers based on the structure and convolution filters of the model.
Six TL-CNN models independently extracted the features. At the GlobalAverage-Pooling2D layers, the features were extracted (TL-EfficientNetB0: 1,280 features, TL-InceptionResnetV2: 1,536 features, TL-InceptionV3: 2,048 features, TL-ResNet50: 2,048 features, TL-VGG16: 512 features, TL-VGG19: 512 features). Then, the extracted features of each TL-CNN model were used as the input to five popular machine learning classifiers: support vector machine (SVM) [29], k-nearest neighbors (k-NN) [30], decision tree (DT) [31], Random Forests (RF) [32], Naïve Bayes [33], and XGBoost [34]. Various machine learning classifiers use different techniques for learning and distinguishing the different classes of data.
Individual machine learning classifiers provide different identification accuracies. This is because each classifier has its own learning ability to identify classes based on the given features. Therefore, an ensemble method is used to aggregate the distribution probabilities of the two classifiers. The proposed method selects two higher prediction classifiers (k-NN and XGBoost) based on an experiment to perform aggregation. An ensemble is a type of soft voting that performs better than other models [35]. Soft voting predicts the final class label as the class label most frequently predicted by classifiers. In soft voting, class labels are predicted by averaging the probability p of the class. Table 1 presents the proposed algorithm, which includes image processing, splitting data, data augmentation, feature extraction, classification, and an ensemble of classifiers:
$ {\boldsymbol{y}}_{\boldsymbol{F}\boldsymbol{C}} = {\boldsymbol{a}\boldsymbol{r}\boldsymbol{g}\boldsymbol{m}\boldsymbol{a}\boldsymbol{x}}_{\boldsymbol{i}}\sum \limits_{\boldsymbol{k} = 1}^{\boldsymbol{m}}{\boldsymbol{w}}_{\boldsymbol{k}}{\boldsymbol{p}}_{\boldsymbol{i}\boldsymbol{k}} $ | (1) |
Algorithm 1: Proposed OCT images Classification |
1: procedure OCT IMAGES PROCESSING 2: return preprocessed-images 3: procedure SPLIT-DATA (OCT-data) 4: train-data, test-data, train-labels, test-labels = split (OCT-images, labels) 5: procedure DATA AUGMENTATION (train-data) 6: augmented images = augmentation (vertical flip, rotation, scale, brightness, saturation, contrast, enhance and 7: contrast, and equalization) 8: return augmented-images 9: procedure 10-FOLD_CROSS_VALIDATION (augmented images, labels) 10: Fold1, Fold2, ……Fold10 = train_test_split(augmented images, labels) 11: return Fold1-10 12: procedure FEATURE_EXTRACTION (Fold1-10, test-data, test-labels) 13: TL-CNN models = modify the convolutional neural network (CNN) models 14: pre-train the TL-CNN models, small top layers are trained, and the final layers are frozen. 15: extracted features = TL-CNN model at GlobalAveragePooling2D layers 16: return extracted features saved in csv format 17: procedure CLASSIFICATION (extracted features, labels) 18: classifiers = [svm, k-NN, DT, RF, Naïve-Bayes, and XGBoost] 19: for clsf in range (0, 6): 20: predicted-labels = classifiers[clsf]. fit (extracted-features) 21: training-accuracy = accuracy (predicted-labels, labels) 22: save_train_weight 23: voting = "soft" 24: ML1 = k-NN (train-data, train-labels, test-data) 25: ML2 = XGBoost (train-data, train-labels, test-data) 26: procedure ENSMEBLE_CLASSIFIERS (train-data, train-labels, test-data) 27: ensemble-classifiers = concadenate(ML1, ML2) 28: ensemble-classifers.fit (train-data, train-labels) 29: predictions = ensemble-classifers.predict(test-data) 30: save_training_weights, results_visualization |
where w_k is the weight of the machine learning classifiers, which can be either k-NN or XGBoost; it automatically learns from disease features in OCT images and then identifies the type of disease based on the input data; i represents the class label of the retinal diseases, where i ∈{0: ARMD, 1: BRVO, 2: CRVO, 3: CSCR, 4: DME}; and p_ik represents the probability of machine-learning weight k for class i.
The proposed OCT image classification method was developed using Python 3.7 and TensorFlow 2.6.0. In addition, Scikit Learn was operated on a personal computer running the Windows 10 operating system powered by an Intel(R) Xeon (R) Silver 4114 @ 2.20GHz CPU, 192GB RAM, and an NVIDIA TITAN RTX 119GB GPU.
The proposed OCT image classification method was trained using augmented OCT images and evaluated using a test set. There were two types of training. First, six TL-CNN models were trained to perform feature extraction from OCT images.
Six TL-CNN models were separately trained using a combination of the training set and the augmented images of the training set. The combination data were split using a 10-fold cross-validation algorithm to separate the images for training, validate the model during training, and prevent overfitting. Furthermore, the TL-CNN models were individually trained with a fixed batch size of 64, epochs of size 100, and an Adam optimizer with a learning rate of 0.0001. The learning rate was selected based on the standard learning rate provided by the TensorFlow library. For example, with a setting of 100 epochs, each model must be trained 100 times on the same data. Therefore, the performance is improved by updating the weight based on the information lost through repetitions of a training session. The weights of each TL-CNN model were recorded in a separate file after training and were utilized to extract features from the training and testing sets. Then, the machine learning models were trained with the convolution features extracted by the TL-CNN models to access the class probabilities. Six machine learning models were separately trained, and the weights were recorded after the training completed. Finally, an ensemble method based on soft voting was applied to the average class probabilities of the two classifiers to obtain an effective final class prediction.
The results of the proposed OCT image classification are divided into three parts: classification results, deployment of the classification results to web services, and a comparison of the results with similar studies in terms of classification accuracy.
A test set was used to evaluate the performance of the proposed method after training the model. The same preprocessing was performed on both the test dataset and the training dataset without data augmentation. The test set contained 601 OCT images, which were used to assess the classification performance. Six TL-CNN models were individually trained to extract features from the OCT images and store the extracted features in pickle format. Six machine learning classifiers were utilized to discriminate the classes of OCT images based on the features extracted by the TL-CNN. Statistical theories were analyzed to measure the classification ability among the classes, sensitivity, specificity, precision, and accuracy. The relationship between the sensitivity and specificity of various categories was shown through a receiver operating characteristic (ROC) curve. Moreover, the confusion matrix was analyzed, which indicated the correct and incorrect class predictions. Table 2 lists the test results of using TL-EfficientNetB0 as an extractor and seven types of classifiers, including an ensemble classifier, the classification result outperformed the ensemble classifier with a sensitivity, specificity, precision, and accuracy of 96.17, 98.92, 95.89 and 95.85%, respectively. The second highest performance was achieved with the k-NN classifier, which achieved a sensitivity, specificity, precision, and accuracy of 87.37, 96.95, 88.82 and 88.89%, respectively. The classification results for the other machine learning classifiers are unstable, both increasing and decreasing randomly.
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-EfficientNetB0 | SVM | 86.79% | 96.78% | 88.64% | 88.39% |
k-NN | 87.37% | 96.95% | 88.82% | 88.89% | |
DT | 85.80% | 96.41% | 84.95% | 86.90% | |
RF | 66.20% | 92.60% | 81.08% | 75.95% | |
Naive Bayes | 86.11% | 96.40% | 86.58% | 86.90% | |
XGBoost | 85.86% | 96.45% | 86.38% | 87.23% | |
Ensemble | 96.17% | 98.92% | 95.89% | 95.85% |
Table 3 shows the classification results when using TL-InceptionResnetV2 as an extractor and seven classifiers, showing that the result outperforms the ensemble classifier with a sensitivity, specificity, precision, and accuracy of 97.42, 99.40, 97.49 and 97.68%, respectively. The second highest performance was achieved with the k-NN classifier, with a sensitivity, specificity, precision, and accuracy of 87.37, 96.48, 88.19 and 87.56%, respectively. In addition, with the same extractor, the classification performance of XGBoost was similar to that of the k-NN classifier. Table 4 lists the evaluation results when using the TL-InceptionV3 extractor and seven machine learning classifiers, including the ensemble classifier, which outperformed other methods with a sensitivity, specificity, precision, and accuracy of 91.34, 97.59, 91.03 and 91.04%, respectively. The second highest performance was achieved by XGBoost, with a sensitivity, specificity, precision, and accuracy of 84.42, 95.10, 82.88, and 82.91%, respectively. Table 5 lists the classification results when using the TL-ResNet50 model as a feature extractor and classifying those features by seven different classifiers, which indicates that using ensemble classifiers outperforms the obtained a sensitivity, specificity, precision, and accuracy of 96.46, 99.14, 96.76 and 96.68%, respectively. The second highest performance was achieved by XGBoost, with a sensitivity, specificity, precision, and accuracy of 87.63, 96.59%, 88.27 and 87.73%, respectively. The performances of the other two classifiers, SVM and k-NN, were comparable and better than those of the three classifiers in the experiments. Table 6 lists the test results of the proposed classification with VGG-16 as a feature extractor and seven machine learning classifiers, the ensemble classifier exhibited the best performance, with a sensitivity, specificity, precision, and classification accuracy of 92.07, 98.00, 92.60 and 92.54%, respectively. The XGBoost classifier had the second highest performance for TL-VGG16 as a feature extractor; it obtained a sensitivity, specificity, precision, and accuracy of 80.48, 94.91, 81.44 and 82.26%, respectively. A similar performance was observed for SVM and k-NN. Table 7 lists the classification test results of the TL-VGG19 model for feature extraction and classification using these features by various classifiers. Ensemble classifiers algorithm outperformed the five other classifiers; its sensitivity, specificity, precision, and accuracy are 93.86, 93.40, 93.44 and 93.86%, respectively. The second- and third-highest performances were achieved by XGBoost and SVM, respectively.
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-InceptionResnetV2 | SVM | 86.27% | 96.13% | 86.86% | 86.40% |
k-NN | 87.37% | 96.48% | 88.19% | 87.56% | |
DT | 83.77% | 95.54% | 83.37% | 84.41% | |
RF | 72.93% | 93.79% | 80.67% | 79.27% | |
Naive Bayes | 79.66% | 93.41% | 78.25% | 77.78% | |
XGBoost | 87.29% | 96.47% | 88.05% | 87.56% | |
Ensemble | 97.42% | 99.40% | 97.49% | 97.68% |
TL-CNN models | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-InceptionV3 | SVM | 82.42% | 94.50% | 80.85% | 81.09% |
k-NN | 83.05% | 94.61% | 80.94% | 81.43% | |
DT | 81.54% | 94.18% | 79.65% | 80.09% | |
RF | 79.50% | 94.01% | 80.52% | 79.77% | |
Naive Bayes | 65.72% | 86.52% | 2.58% | 61.33% | |
XGBoost | 84.42% | 95.10% | 82.88% | 82.91% | |
Ensemble | 91.34% | 97.59% | 91.03% | 91.04% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-ResNet50 | SVM | 86.25% | 96.02% | 85.12% | 85.95% |
k-NN | 85.75% | 96.00% | 86.26% | 85.74% | |
DT | 82.04% | 94.94% | 82.34% | 82.59% | |
RF | 52.90% | 88063% | 71.77% | 65.67% | |
Naive Bayes | 67.71% | 89.82% | 72.49% | 64.68% | |
XGBoost | 87.63% | 96.59% | 88.27% | 87.73% | |
Ensemble | 96.46% | 99.14% | 96.76% | 96.68% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-VGG16 | SVM | 76.39% | 93.49% | 76.96% | 78.28% |
k-NN | 74.55% | 92.33% | 75.70% | 74.96% | |
DT | 57.42% | 84.86% | 55.72% | 58.37% | |
RF | 50.55% | 86.58% | 38.39% | 63.18% | |
Naive Bayes | 59.53% | 85.08% | 58.84% | 59.20% | |
XGBoost | 80.48% | 94.91% | 81.44% | 82.26% | |
Ensemble | 92.07% | 98.00% | 92.60% | 92.54% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-VGG19 | SVM | 79.90% | 93.74% | 78.77% | 78.82% |
k-NN | 69.16% | 90.99% | 70.56% | 71.64% | |
DT | 53.23% | 82.94% | 53.41% | 54.73% | |
RF | 48.29% | 85.62% | 37.71% | 60.36% | |
Naive Bayes | 56.41% | 82.96% | 54.70% | 54.89% | |
XGBoost | 82.44% | 95.30% | 81.90% | 83.58% | |
Ensemble | 93.86% | 93.40% | 93.44% | 93.86% |
Six TL-CNN models were compared, and TL-InceptionResNetV2 achieved a better performance than the other five models used in this study, with a sensitivity, specificity, precision, and accuracy of 97.42, 99.40, 97.49 and 97.68 %, respectively. The ensemble algorithm always outperformed all the TL-CNN models. The individual k-NN and XGBoost classifiers performed better than the three individual classifiers. Thus, ensembled k-NN and XGBoost also achieved better performance than k-NN and XGBoost.
Figure 3 shows the ROC result of the proposed classification method, which outperforms TL- InceptionResnetV2 with ensemble classifiers (k-NN and XGBoost). The ROC among each class ARMD, BRVO, CRVO, CSCR, and DME is 0.99, 0.96, 0.99, 0.99, and 0.98, respectively. The relationship between sensitivity and specificity of the five classes is most important. The confusion matrix is implemented by using the Sklearn library in Python. The size of test data is essential to present the robustness of classification. The confusion matrix shows the number of correct and incorrect predictions among all classes. Figure 4 shows the confusion matrix of the proposed method which exhibited best performance; 148 of 149 OCT images of ARMD class are correctly predicted, 85 of 91 images of BRVO class are correctly predicted (ARMD:3 and DME:3 are incorrect predictions), 59 of 60 images are correctly predicted as CRVO and one image that is incorrectly predicted as BRVO, 148 of 150 images are correctly predicted and two images are incorrectly predicted as ARMD, and 149 of 153 are correctly predicted, and four are incorrectly predicted (ARMD: 1, BRVO:1, and CRVO:2 are incorrect prediction).
To render the proposed method applicable and accessible from outside through an Internet connection, we deployed the proposed OCT image classification to a web server using the Flask framework. The web server receives one image input at a time and inputs it into the proposed classification method to predict retinal diseases. The input image is an OCT image consisting of three channels with a resolution of 300 pixels in height and 500 pixels in width. When inputting an OCT image through a web service user interface (UI), the image is transferred to a computer server that runs a DL classification model. First, the computer server performs image processing which is the same to the processes used in both the train and test sets. Second, the preprocessed image is inputted into the proposed classification weights for prediction. Finally, the predicted results are forwarded to the web service using the Flask framework. The prediction results consist of the image input, distribution probabilities among the five classes, the retinal disease diagnosis class, and prediction times of an image. The prediction time is the time taken to input an image to a web service to predict and return the prediction result. Figure 5 shows the initial UI of the web server. The prediction results obtained after inputting the OCT images are shown in Figure 6.
The higher accuracy of the proposed OCT image classification method is compared with that of the recent studies reviewed in the literature review section, as listed in Table 8. These studies focused on transfer learning, developing new models, and combining well known CNN models with machine learning. All the listed studies used either different OCT databases or a combination of these datasets. Moreover, the number and type of classification classes were different, with at most four classes. We classify retinal diseases into five classes using a dataset obtained from a hospital. An additional number of classes can affect the performance of the classification methods. Table 8 lists the methods and algorithms that have been presented, including the suggested model with transfer learning, the multiscale DL model, and transfer learning using existing CNN models. However, the results as listed in the literature review have shown an accuracy of < 97%. Instead of focusing on a single classifier, this study combines two machine-learning classifiers and the DL as a feature extractor. Our study exhibits an accuracy of 97.68%, which is greater than the accuracy of the aforementioned studies. In addition, the number of classification classes is higher than that of the studies reviewed.
Author | Year | Method | Disease type | Dataset size | Accuracy |
Han et al. [13] | 2022 | Transfer learning with a modification of the well-known CNN models | 4-class: PCV, RAP, nAMD, and NORMAL | 4749 | 87.4% |
Sotoudeh-Paima et al. [14] | 2022 | Deep learning: multi-scale convolutional neural network | 3-class: AMD, CNV, NORMAL | 120,961 | 93.4% |
Elaziz et al. [15] | 2022 | Ensemble deep learning model for feature extraction, features selection, machine learning as classifier. | 4-class: DME, CNV, DRUSEN, and NORMAL | 84,484 | 94.32% |
Liu et al. [16] | 2022 | Deep learning based on method and lesions segmentation model. | 4-class: CNV, DME, DRUSEN, and NORMAL | 86,134 | 95.10% |
Minagi et al. [17] | 2022 | Transfer learning with DNN models | 4-class: CNV, DME, DRUSEN, and NORMAL | 11,200 | 95.3% |
Tayal et al. [18] | 2022 | Deep learning-based method | 4-class: DME, CNV, DRUSEN, NORMAL | 84,484 | 96.5% |
Proposed method | − | Hybrid of deep learning and machine learning + ensemble machine learning classifiers. | 5-class: ARMD, BRVO, CRVO, CSCR, DME | 2,998 | 97.68% |
Our study classifies retinal OCT images with disease classes that differ from the reviewed studies and are not available in the public dataset. We hope that these retinal diseases will become available in the future, and we will evaluate the proposed OCT image classification system using a public dataset.
This study presents a hybrid ensemble OCT image classification method for the diagnosis of five classes of retinal diseases. The proposed method employs an ensemble machine learning classifier as the classifier and a hybrid deep learning model as the feature extractor. We identified the deep learning model and ensemble classifiers that were most suitable for OCT image classification. The proposed model outperformed an individual classifier. With an accuracy of 97.68%, the best deep learning model and ensemble machine learning classifiers of the proposed method were TL- InceptionResnetV2 and the aggregation of KNN and XGBoost. This classification can be deployed to web services for convenient access to diagnose retinal disease from outside the Internet. Moreover, the prediction time in seconds was short, reducing the time required for prediction. This study contributes to the development of accurate multiclass OCT image classification. In the future, we aim to improve the classification performance. If datasets with the same class as in our study are made public, we will assess the proposed method on these datasets to broaden their applicability. In the medical field, improved performance can be used to automatically classify OCT images and eliminate time-consuming tasks, and this classification can also aid in the prevention of vision loss.
The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.
The data used to support this study have not been made available access because they are real clinical data from Soonchunhyang Bucheon Hospital, and patient's privacy should be protected, it enables to detect people through this data, but they are available from the corresponding author on reasonable request.
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2021R1A2C1010362) and the Soonchunhyang University Research Fund.
The authors declare no competing interests.
[1] |
Buchanan TA, Xiang AH (2005) Gestational diabetes mellitus. J Clin Invest 115: 485–491. doi: 10.1172/JCI200524531
![]() |
[2] |
Konstanze M, Holger S, Mathias F (2012) Leptin, adiponectin and other adipokines in gestational diabetes mellitus and pre-eclampsia. Clin Endocrinol 76: 2–11. doi: 10.1111/j.1365-2265.2011.04234.x
![]() |
[3] |
Bellamy L, Casas JP, Hingorani AD, et al. (2009) Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis. Lancet 373: 1773–1779. doi: 10.1016/S0140-6736(09)60731-5
![]() |
[4] | Bener A, Saleh NM, Al-Hamaq A (2011) Prevalence of gestational diabetes and associated maternal and neonatal complications in a fast-developing community: global comparisons. Int J Women's Health 3: 367–373. |
[5] |
Reece EA, Leguizamón G, Wiznitzer A (2009) Gestational diabetes: the need for a common ground. Lancet 373: 1789–1797. doi: 10.1016/S0140-6736(09)60515-8
![]() |
[6] |
Hod M, Kapur A, Sacks DA, et al. (2015) The International Federation of Gynecology and Obstetrics (FIGO) Initiative on gestational diabetes mellitus: A pragmatic guide for diagnosis, management, and care. Int J Gynecol Obstet 131: S173–S211. doi: 10.1016/S0020-7292(15)30033-3
![]() |
[7] |
Craig WJ (2010) Nutrition Concerns and Health Effects of Vegetarian Diets. Nutr Clin Pract 25: 613–620. doi: 10.1177/0884533610385707
![]() |
[8] |
De Silva DD, Rapior S, Hyde KD, et al. (2012) Medicinal mushrooms in prevention and control of diabetes mellitus. Fungal Divers 56: 1–29. doi: 10.1007/s13225-012-0187-4
![]() |
[9] |
Martel J, Ojcius DM, Chang CJ, et al. (2017) Anti-obesogenic and antidiabetic effects of plants and mushrooms. Nat Rev Endocrinol 13: 149–160. doi: 10.1038/nrendo.2016.142
![]() |
[10] | Royse DJ, Singh M (2014) A global perspective on the high five: Agaricus, Pleurotus, Lentinula, Auricularia & Flammulina, 1–6. |
[11] | Valverde ME, Hernndez-Prez T, Paredes-Lopez O (2015) Edible Mushrooms: Improving Human Health and Promoting Quality Life. Int J Microbiol 2015: 376387. |
[12] |
Horowitz S (2011) Medicinal Mushrooms: Research Support for Modern Applications of Traditional Uses. Altern Complem Ther 17: 323–329. doi: 10.1089/act.2011.17602
![]() |
[13] | Mohamed M, Nassef D, Waly E, et al. (2012) Earliness, Biological efficiency and basidiocarp yield of Pleurotus ostreatus and P. columbinus oyster mushrooms in response to different sole and mixed substrates. Assiut J Agric Sci 43: 91–114. |
[14] |
Gargano ML, van Griensven LJ, Isikhuemhen OS, et al. (2017) Medicinal mushrooms: Valuable biological resources of high exploitation potential. Plant Biosys 151: 548–565. doi: 10.1080/11263504.2017.1301590
![]() |
[15] | Deepalakshmi K, Mirunalini S (2011) Therapeutic properties and current medical usage of medicinal mushroom: Ganoderma lucidum. Inter J Pharm Sci Res 2: 1922–1929. |
[16] |
Klupp NL, Kiat H, Bensoussan A, et al. (2016) A double-blind, randomised, placebo-controlled trial of Ganoderma lucidum for the treatment of cardiovascular risk factors of metabolic syndrome. Sci Rep 6: 29540. doi: 10.1038/srep29540
![]() |
[17] |
Holliday JC, Cleaver MP (2008) Medicinal Value of the Caterpillar Fungi Species of the Genus Cordyceps (Fr.) Link (Ascomycetes). A Review. Int J Med Mushrooms 10: 219–234. doi: 10.1615/IntJMedMushr.v10.i3.30
![]() |
[18] |
Firenzuoli F, Gori L, Lombardo G (2008) The Medicinal Mushroom Agaricus blazei Murrill: Review of Literature and Pharmaco-Toxicological Problems. Evid Based Complement Alternat Med 5: 3–15. doi: 10.1093/ecam/nem007
![]() |
[19] |
Vitak T, Yurkiv B, Wasser S, et al. (2017) Effect of medicinal mushrooms on blood cells under conditions of diabetes mellitus. World J Diabetes 8: 187–201. doi: 10.4239/wjd.v8.i5.187
![]() |
[20] |
Lei H, Guo S, Han J, et al. (2012) Hypoglycemic and hypolipidemic activities of MT-α-glucan and its effect on immune function of diabetic mice. Carbohydr Polym 89: 245–250. doi: 10.1016/j.carbpol.2012.03.003
![]() |
[21] |
Khan MA, Tania M (2012) Nutritional and medicinal importance of Pleurotus mushrooms: An overview. Food Rev Int 28: 313–329. doi: 10.1080/87559129.2011.637267
![]() |
[22] |
Vitak TY, Wasser SP, Nevo E, et al. (2015) Structural Changes of Erythrocyte Surface Glycoconjugates after Treatment with Medicinal Mushrooms. Int J Med Mushrooms 17: 867–878. doi: 10.1615/IntJMedMushrooms.v17.i9.70
![]() |
[23] |
Maschio BH, Gentil BC, Caetano ELA, et al. (2017) Characterization of the Effects of the Shiitake Culinary-Medicinal Mushroom, Lentinus edodes (Agaricomycetes), on Severe Gestational Diabetes Mellitus in Rats. Int J Med Mushrooms 19: 991–1000. doi: 10.1615/IntJMedMushrooms.2017024498
![]() |
[24] |
Chen YH, Lee CH, Hsu TH, et al. (2015) Submerged-Culture Mycelia and Broth of the Maitake Medicinal Mushroom Grifola frondosa (Higher Basidiomycetes) Alleviate Type 2 Diabetes-Induced Alterations in Immunocytic Function. Int J Med Mushrooms 17: 541–556. doi: 10.1615/IntJMedMushrooms.v17.i6.50
![]() |
[25] |
Rony KA, Ajith TA, Janardhanan KK (2015) Hypoglycemic and Hypolipidemic Effects of the Cracked-Cap Medicinal Mushroom Phellinus rimosus (Higher Basidiomycetes) in Streptozotocin-Induced Diabetic Rats. Int J Med Mushrooms 17: 521–531. doi: 10.1615/IntJMedMushrooms.v17.i6.30
![]() |
[26] |
Yurkiv B, Wasser SP, Nevo E, et al. (2015) The Effect of Agaricus brasiliensis and Ganoderma lucidum Medicinal Mushroom Administration on the L-arginine/Nitric Oxide System and Rat Leukocyte Apoptosis in Experimental Type 1 Diabetes Mellitus. Int J Med Mushrooms 17: 339–350. doi: 10.1615/IntJMedMushrooms.v17.i4.30
![]() |
[27] | Jayasuriya WJ, Suresh TS, Abeytunga D, et al. (2012) Oral hypoglycemic activity of culinary-medicinal mushrooms Pleurotus ostreatus and P. cystidiosus (higher basidiomycetes) in normal and alloxan-induced diabetic Wistar rats. Int J Med Mushrooms 14: 347–355. |
[28] |
Ganeshpurkar A, Kohli S, Rai G (2014) Antidiabetic potential of polysaccharides from the white oyster culinary-medicinal mushroom Pleurotus florida (higher Basidiomycetes). Int J Med Mushrooms 16: 207–217. doi: 10.1615/IntJMedMushr.v16.i3.10
![]() |
[29] |
Lei H, Guo S, Han J, et al. (2012) Hypoglycemic and hypolipidemic activities of MT-alpha-glucan and its effect on immune function of diabetic mice. Carbohydr Polym 89: 245–250. doi: 10.1016/j.carbpol.2012.03.003
![]() |
[30] |
Zhang Y, Hu T, Zhou H, et al. (2016) Antidiabetic effect of polysaccharides from Pleurotus ostreatus in streptozotocin-induced diabetic rats. Int J Biol Macromol 83: 126–132. doi: 10.1016/j.ijbiomac.2015.11.045
![]() |
[31] |
Zhou S, Liu Y, Yang Y, et al. (2015) Hypoglycemic Activity of Polysaccharide from Fruiting Bodies of the Shaggy Ink Cap Medicinal Mushroom, Coprinus comatus (Higher Basidiomycetes), on Mice Induced by Alloxan and Its Potential Mechanism. Int J Med Mushrooms 17: 957–964. doi: 10.1615/IntJMedMushrooms.v17.i10.50
![]() |
[32] |
Jeong SC, Jeong YT, Yang BK, et al. (2010) White button mushroom (Agaricus bisporus) lowers blood glucose and cholesterol levels in diabetic and hypercholesterolemic rats. Nutr Res 30: 49–56. doi: 10.1016/j.nutres.2009.12.003
![]() |
[33] |
Kiho T, Sobue S, Ukai S (1994) Structural features and hypoglycemic activities of two polysaccharides from a hot-water extract of Agrocybe cylindracea. Carbohydr Res 251: 81–87. doi: 10.1016/0008-6215(94)84277-9
![]() |
[34] |
Gray AM, Flatt PR (1998) Insulin-releasing and insulin-like activity of Agaricus campestris (mushroom). J Endocrinol 157: 259–266. doi: 10.1677/joe.0.1570259
![]() |
[35] |
Wisitrassameewong K, Karunarathna SC, Thongklang N, et al. (2012) Agaricus subrufescens: A review. Saudi J Biol Sci 19: 131–146. doi: 10.1016/j.sjbs.2012.01.003
![]() |
[36] |
Kerrigan RW (2005) Agaricus subrufescens, a cultivated edible and medicinal mushroom, and its synonyms. Mycologia 97: 12–24. doi: 10.1080/15572536.2006.11832834
![]() |
[37] |
Niwa A, Tajiri T, Higashino H (2011) Ipomoea batatas and Agarics blazei ameliorate diabetic disorders with therapeutic antioxidant potential in streptozotocin-induced diabetic rats. J Clin Biochem Nutr 48: 194–202. doi: 10.3164/jcbn.10-78
![]() |
[38] |
Vincent HK, Innes KE, Vincent KR (2007) Oxidative stress and potential interventions to reduce oxidative stress in overweight and obesity. Diabetes Obes Metab 9: 813–839. doi: 10.1111/j.1463-1326.2007.00692.x
![]() |
[39] |
Hu XY, Liu CG, Wang X, et al. (2017) Hpyerglycemic and anti-diabetic nephritis activities of polysaccharides separated from Auricularia auricular in diet-streptozotocin-induced diabetic rats. Exp Ther Med 13: 352–358. doi: 10.3892/etm.2016.3943
![]() |
[40] |
Ding ZY, Lu YJ, Lu ZX, et al. (2010) Hypoglycaemic effect of comatin, an antidiabetic substance separated from Coprinus comatus broth, on alloxan-induced-diabetic rats. Food Chem 121: 39–43. doi: 10.1016/j.foodchem.2009.12.001
![]() |
[41] |
Lv YT, Han LN, Yuan C, et al. (2009) Comparison of Hypoglycemic Activity of Trace Elements Absorbed in Fermented Mushroom of Coprinus comatus. Biol Trace Elem Res 131: 177–185. doi: 10.1007/s12011-009-8352-7
![]() |
[42] |
Guo JY, Han CC, Liu YM (2010) A Contemporary Treatment Approach to Both Diabetes and Depression by Cordyceps sinensis, Rich in Vanadium. Evid Based Complement Alternat Med: 7: 387–389. doi: 10.1093/ecam/nep201
![]() |
[43] |
Nie S, Cui SW, Xie MY, et al. (2013) Bioactive polysaccharides from Cordyceps sinensis: Isolation, structure features and bioactivities. Bioact Carbohydrates Dietary Fibre 1: 38–52. doi: 10.1016/j.bcdf.2012.12.002
![]() |
[44] |
Pan D, Zhang D, Wu JS, et al. (2013) Antidiabetic, Antihyperlipidemic and Antioxidant Activities of a Novel Proteoglycan from Ganoderma Lucidum Fruiting Bodies on db/db Mice and the Possible Mechanism. PLoS One 8: e68332. doi: 10.1371/journal.pone.0068332
![]() |
[45] |
Hong L, Xun M, Wutong W (2007) Anti-diabetic effect of an alpha-glucan from fruit body of maitake (Grifola frondosa) on KK-Ay mice. J Pharm Pharmacol 59: 575–582. doi: 10.1211/jpp.59.4.0013
![]() |
[46] |
Chaiyasut C, Sivamaruthi BS (2017) Anti-hyperglycemic property of Hericium erinaceus – A mini review. Asian Pac J Trop Biomed 7: 1036–1040. doi: 10.1016/j.apjtb.2017.09.024
![]() |
[47] |
Liang B, Guo ZD, Xie F, et al. (2013) Antihyperglycemic and antihyperlipidemic activities of aqueous extract of Hericium erinaceus in experimental diabetic rats. BMC Complement Altern Med 13: 253. doi: 10.1186/1472-6882-13-253
![]() |
[48] |
Geng Y, Lu ZM, Huang W, et al. (2013) Bioassay-Guided Isolation of DPP-4 Inhibitory Fractions from Extracts of Submerged Cultured of Inonotus obliquus. Molecules 18: 1150–1161. doi: 10.3390/molecules18011150
![]() |
[49] |
Wang J, Wang C, Li S, et al. (2017) Anti-diabetic effects of Inonotus obliquus polysaccharides in streptozotocin-induced type 2 diabetic mice and potential mechanism via PI3K-Akt signal pathway. Biomed Pharmacother 95: 1669–1677. doi: 10.1016/j.biopha.2017.09.104
![]() |
[50] |
Bisen P, Baghel RK, Sanodiya BS, et al. (2010) Lentinus edodes: A macrofungus with pharmacological activities. Curr Med Chem 17: 2419–2430. doi: 10.2174/092986710791698495
![]() |
[51] | Wahab NAA, Abdullah N, Aminudin N (2014) Characterisation of Potential Antidiabetic-Related Proteins from Pleurotus pulmonarius (Fr.) Quél. (Grey Oyster Mushroom) by MALDI-TOF/TOF Mass Spectrometry. Biomed Res Int 2014: 131607. |
[52] | Badole SL, Patel NM, Thakurdesai PA, et al. (2008) Interaction of Aqueous Extract of Pleurotus pulmonarius (Fr.) Quel-Champ. with Glyburide in Alloxan Induced Diabetic Mice. Evid Based Complement Alternat Med 5: 159–164. |
[53] |
Kiho T, Morimoto H, Kobayashi T, et al. (2000) Effect of a polysaccharide (TAP) from the fruiting bodies of Tremella aurantia on glucose metabolism in mouse liver. Biosci Biotechnol Biochem 64: 417–419. doi: 10.1271/bbb.64.417
![]() |
[54] |
Kiho T, Kochi M, Usui S, et al. (2001) Antidiabetic effect of an acidic polysaccharide (TAP) from Tremella aurantia and its degradation product (TAP-H). Biol Pharm Bull 24: 1400–1403. doi: 10.1248/bpb.24.1400
![]() |
[55] |
Cho EJ, Hwang HJ, Kim SW, et al. (2007) Hypoglycemic effects of exopolysaccharides produced by mycelial cultures of two different mushrooms Tremella fuciformis and Phellinus baumii in ob/ob mice. Appl Microbiol Biotechnol 75: 1257–1265. doi: 10.1007/s00253-007-0972-2
![]() |
[56] |
Fu M, Wang L, Wang XY, et al. (2018) Determination of the Five Main Terpenoids in Different Tissues of Wolfiporia cocos. Molecules 23: 1839. doi: 10.3390/molecules23081839
![]() |
[57] |
Esteban CI (2009) Medicinal interest of Poria cocos (Wolfiporia extensa). Rev Iberoam Micol 26: 103–107. doi: 10.1016/S1130-1406(09)70019-1
![]() |
[58] |
Li Y, Zhang J, Li T, et al. (2016) A Comprehensive and Comparative Study of Wolfiporia extensa Cultivation Regions by Fourier Transform Infrared Spectroscopy and Ultra-Fast Liquid Chromatography. PLoS One 11: e0168998. doi: 10.1371/journal.pone.0168998
![]() |
[59] |
Shafrir E, Spielman S, Nachliel I, et al. (2001) Treatment of diabetes with vanadium salts: general overview and amelioration of nutritionally induced diabetes in the Psammomys obesus gerbil. Diabetes Metab Res Rev 17: 55–66. doi: 10.1002/1520-7560(2000)9999:9999<::AID-DMRR165>3.0.CO;2-J
![]() |
[60] |
Clark TA, Deniset JF, Heyliger CE, et al. (2014) Alternative therapies for diabetes and its cardiac complications: role of vanadium. Heart Fail Rev 19: 123–132. doi: 10.1007/s10741-013-9380-0
![]() |
[61] | Gruzewska K, Michno A, Pawelczyk T, et al. (2014) Essentiality and toxicity of vanadium supplements in health and pathology. J Physiol Pharmacol 65: 603–611. |
[62] |
Halberstam M, Cohen N, Shlimovich P, et al. (1996) Oral vanadyl sulfate improves insulin sensitivity in NIDDM but not in obese nondiabetic subjects. Diabetes 45: 659–666. doi: 10.2337/diab.45.5.659
![]() |
[63] | Huang HY, Korivi M, Chaing YY, et al. (2012) Pleurotus tuber-regium Polysaccharides Attenuate Hyperglycemia and Oxidative Stress in Experimental Diabetic Rats. Evid Based Complement Alternat Med 2012: 856381. |
[64] |
Huang HY, Korivi M, Yang HT, et al. (2014) Effect of Pleurotus tuber-regium polysaccharides supplementation on the progression of diabetes complications in obese-diabetic rats. Chin J Physiol 57: 198–208. doi: 10.4077/CJP.2014.BAC245
![]() |
[65] |
Kobayashi M, Kawashima H, Takemori K, et al. (2012) Ternatin, a cyclic peptide isolated from mushroom, and its derivative suppress hyperglycemia and hepatic fatty acid synthesis in spontaneously diabetic KK-A(y) mice. Biochem Biophys Res Commun 427: 299–304. doi: 10.1016/j.bbrc.2012.09.045
![]() |
[66] | Laurino LF, Viroel FJM, Pickler TB, et al. (2017) Functional foods in gestational diabetes: Evaluation of the oral glucose tolerance test (OGTT) in pregnant rats treated with mushrooms. Reprod Toxicol 72: 36. |
[67] | Jayasuriya WJ, Wanigatunge CA, Fernando GH, et al. (2015) Hypoglycaemic activity of culinary Pleurotus ostreatus and P. cystidiosus mushrooms in healthy volunteers and type 2 diabetic patients on diet control and the possible mechanisms of action. Phytother Res 29: 303–309. |
[68] | Gao Y, Lan J, Dai X, et al. (2004) A Phase I/II Study of Ling Zhi Mushroom Ganoderma lucidum (W.Curt.:Fr.) Lloyd (Aphyllophoromycetideae) Extract in Patients with Type II Diabetes Mellitus. Int J Med Mushrooms 6: 327-334. |
[69] |
Friedman M (2016) Mushroom Polysaccharides: Chemistry and Antiobesity, Antidiabetes, Anticancer, and Antibiotic Properties in Cells, Rodents, and Humans. Foods 5: 80. doi: 10.3390/foods5040080
![]() |
[70] |
Lo HC, Wasser SP (2011) Medicinal mushrooms for glycemic control in diabetes mellitus: history, current status, future perspectives, and unsolved problems (review). Int J Med Mushrooms 13: 401–426. doi: 10.1615/IntJMedMushr.v13.i5.10
![]() |
1. | Mohammad Mahdi Azizi, Setareh Abhari, Hedieh Sajedi, Alan Marmorstein, Stitched vision transformer for age-related macular degeneration detection using retinal optical coherence tomography images, 2024, 19, 1932-6203, e0304943, 10.1371/journal.pone.0304943 | |
2. | S. Vishnu Priyan, R. Vinod Kumar, C. Moorthy, V.S. Nishok, A fusion of deep neural networks and game theory for retinal disease diagnosis with OCT images, 2024, 32, 08953996, 1011, 10.3233/XST-240027 | |
3. | P. T. Karule, Sujata B. Bhele, Prasanna Palsodkar, Poonam T. Agarkar, Hirendra R. Hajare, Prashant R. Patil, 2024, Detection of Multi-Class Multi-Label Ophthalmological Diseases in Retinal Fundus Images Using Machine Learning, 979-8-3503-1901-9, 1, 10.1109/ICICET59348.2024.10616291 | |
4. | Yash Mori, Nandini Modi, 2024, Enhancing OCT Image Classification for Retinal Disease Diagnosis: A Novel Approach Using Squeeze-and-Excitation ResNet, 979-8-3503-8459-8, 940, 10.1109/AIC61668.2024.10730829 | |
5. | Muhammed Enes Atik, İbrahim Kocak, Nihat Sayin, Sadik Etka Bayramoglu, Ahmet Ozyigit, Integration of Optical Coherence Tomography Images and Real‐Life Clinical Data for Deep Learning Modeling: A Unified Approach in Prognostication of Diabetic Macular Edema, 2024, 1864-063X, 10.1002/jbio.202400315 | |
6. | G. Jeyasri, R. Karthiyayini, Deep learning based retinal disease classification using an autoencoder and generative adversarial network, 2025, 108, 17468094, 107852, 10.1016/j.bspc.2025.107852 | |
7. | Pavithra Mani, Neelaveni Ramachandran, Palanichamy Naveen, Prasanna Venkatesh Ramesh, An enhanced lightweight transformer-based framework for accurate retinal disease classification from OCT images, 2025, 0972-8821, 10.1007/s12596-025-02793-6 |
Algorithm 1: Proposed OCT images Classification |
1: procedure OCT IMAGES PROCESSING 2: return preprocessed-images 3: procedure SPLIT-DATA (OCT-data) 4: train-data, test-data, train-labels, test-labels = split (OCT-images, labels) 5: procedure DATA AUGMENTATION (train-data) 6: augmented images = augmentation (vertical flip, rotation, scale, brightness, saturation, contrast, enhance and 7: contrast, and equalization) 8: return augmented-images 9: procedure 10-FOLD_CROSS_VALIDATION (augmented images, labels) 10: Fold1, Fold2, ……Fold10 = train_test_split(augmented images, labels) 11: return Fold1-10 12: procedure FEATURE_EXTRACTION (Fold1-10, test-data, test-labels) 13: TL-CNN models = modify the convolutional neural network (CNN) models 14: pre-train the TL-CNN models, small top layers are trained, and the final layers are frozen. 15: extracted features = TL-CNN model at GlobalAveragePooling2D layers 16: return extracted features saved in csv format 17: procedure CLASSIFICATION (extracted features, labels) 18: classifiers = [svm, k-NN, DT, RF, Naïve-Bayes, and XGBoost] 19: for clsf in range (0, 6): 20: predicted-labels = classifiers[clsf]. fit (extracted-features) 21: training-accuracy = accuracy (predicted-labels, labels) 22: save_train_weight 23: voting = "soft" 24: ML1 = k-NN (train-data, train-labels, test-data) 25: ML2 = XGBoost (train-data, train-labels, test-data) 26: procedure ENSMEBLE_CLASSIFIERS (train-data, train-labels, test-data) 27: ensemble-classifiers = concadenate(ML1, ML2) 28: ensemble-classifers.fit (train-data, train-labels) 29: predictions = ensemble-classifers.predict(test-data) 30: save_training_weights, results_visualization |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-EfficientNetB0 | SVM | 86.79% | 96.78% | 88.64% | 88.39% |
k-NN | 87.37% | 96.95% | 88.82% | 88.89% | |
DT | 85.80% | 96.41% | 84.95% | 86.90% | |
RF | 66.20% | 92.60% | 81.08% | 75.95% | |
Naive Bayes | 86.11% | 96.40% | 86.58% | 86.90% | |
XGBoost | 85.86% | 96.45% | 86.38% | 87.23% | |
Ensemble | 96.17% | 98.92% | 95.89% | 95.85% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-InceptionResnetV2 | SVM | 86.27% | 96.13% | 86.86% | 86.40% |
k-NN | 87.37% | 96.48% | 88.19% | 87.56% | |
DT | 83.77% | 95.54% | 83.37% | 84.41% | |
RF | 72.93% | 93.79% | 80.67% | 79.27% | |
Naive Bayes | 79.66% | 93.41% | 78.25% | 77.78% | |
XGBoost | 87.29% | 96.47% | 88.05% | 87.56% | |
Ensemble | 97.42% | 99.40% | 97.49% | 97.68% |
TL-CNN models | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-InceptionV3 | SVM | 82.42% | 94.50% | 80.85% | 81.09% |
k-NN | 83.05% | 94.61% | 80.94% | 81.43% | |
DT | 81.54% | 94.18% | 79.65% | 80.09% | |
RF | 79.50% | 94.01% | 80.52% | 79.77% | |
Naive Bayes | 65.72% | 86.52% | 2.58% | 61.33% | |
XGBoost | 84.42% | 95.10% | 82.88% | 82.91% | |
Ensemble | 91.34% | 97.59% | 91.03% | 91.04% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-ResNet50 | SVM | 86.25% | 96.02% | 85.12% | 85.95% |
k-NN | 85.75% | 96.00% | 86.26% | 85.74% | |
DT | 82.04% | 94.94% | 82.34% | 82.59% | |
RF | 52.90% | 88063% | 71.77% | 65.67% | |
Naive Bayes | 67.71% | 89.82% | 72.49% | 64.68% | |
XGBoost | 87.63% | 96.59% | 88.27% | 87.73% | |
Ensemble | 96.46% | 99.14% | 96.76% | 96.68% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-VGG16 | SVM | 76.39% | 93.49% | 76.96% | 78.28% |
k-NN | 74.55% | 92.33% | 75.70% | 74.96% | |
DT | 57.42% | 84.86% | 55.72% | 58.37% | |
RF | 50.55% | 86.58% | 38.39% | 63.18% | |
Naive Bayes | 59.53% | 85.08% | 58.84% | 59.20% | |
XGBoost | 80.48% | 94.91% | 81.44% | 82.26% | |
Ensemble | 92.07% | 98.00% | 92.60% | 92.54% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-VGG19 | SVM | 79.90% | 93.74% | 78.77% | 78.82% |
k-NN | 69.16% | 90.99% | 70.56% | 71.64% | |
DT | 53.23% | 82.94% | 53.41% | 54.73% | |
RF | 48.29% | 85.62% | 37.71% | 60.36% | |
Naive Bayes | 56.41% | 82.96% | 54.70% | 54.89% | |
XGBoost | 82.44% | 95.30% | 81.90% | 83.58% | |
Ensemble | 93.86% | 93.40% | 93.44% | 93.86% |
Author | Year | Method | Disease type | Dataset size | Accuracy |
Han et al. [13] | 2022 | Transfer learning with a modification of the well-known CNN models | 4-class: PCV, RAP, nAMD, and NORMAL | 4749 | 87.4% |
Sotoudeh-Paima et al. [14] | 2022 | Deep learning: multi-scale convolutional neural network | 3-class: AMD, CNV, NORMAL | 120,961 | 93.4% |
Elaziz et al. [15] | 2022 | Ensemble deep learning model for feature extraction, features selection, machine learning as classifier. | 4-class: DME, CNV, DRUSEN, and NORMAL | 84,484 | 94.32% |
Liu et al. [16] | 2022 | Deep learning based on method and lesions segmentation model. | 4-class: CNV, DME, DRUSEN, and NORMAL | 86,134 | 95.10% |
Minagi et al. [17] | 2022 | Transfer learning with DNN models | 4-class: CNV, DME, DRUSEN, and NORMAL | 11,200 | 95.3% |
Tayal et al. [18] | 2022 | Deep learning-based method | 4-class: DME, CNV, DRUSEN, NORMAL | 84,484 | 96.5% |
Proposed method | − | Hybrid of deep learning and machine learning + ensemble machine learning classifiers. | 5-class: ARMD, BRVO, CRVO, CSCR, DME | 2,998 | 97.68% |
Algorithm 1: Proposed OCT images Classification |
1: procedure OCT IMAGES PROCESSING 2: return preprocessed-images 3: procedure SPLIT-DATA (OCT-data) 4: train-data, test-data, train-labels, test-labels = split (OCT-images, labels) 5: procedure DATA AUGMENTATION (train-data) 6: augmented images = augmentation (vertical flip, rotation, scale, brightness, saturation, contrast, enhance and 7: contrast, and equalization) 8: return augmented-images 9: procedure 10-FOLD_CROSS_VALIDATION (augmented images, labels) 10: Fold1, Fold2, ……Fold10 = train_test_split(augmented images, labels) 11: return Fold1-10 12: procedure FEATURE_EXTRACTION (Fold1-10, test-data, test-labels) 13: TL-CNN models = modify the convolutional neural network (CNN) models 14: pre-train the TL-CNN models, small top layers are trained, and the final layers are frozen. 15: extracted features = TL-CNN model at GlobalAveragePooling2D layers 16: return extracted features saved in csv format 17: procedure CLASSIFICATION (extracted features, labels) 18: classifiers = [svm, k-NN, DT, RF, Naïve-Bayes, and XGBoost] 19: for clsf in range (0, 6): 20: predicted-labels = classifiers[clsf]. fit (extracted-features) 21: training-accuracy = accuracy (predicted-labels, labels) 22: save_train_weight 23: voting = "soft" 24: ML1 = k-NN (train-data, train-labels, test-data) 25: ML2 = XGBoost (train-data, train-labels, test-data) 26: procedure ENSMEBLE_CLASSIFIERS (train-data, train-labels, test-data) 27: ensemble-classifiers = concadenate(ML1, ML2) 28: ensemble-classifers.fit (train-data, train-labels) 29: predictions = ensemble-classifers.predict(test-data) 30: save_training_weights, results_visualization |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-EfficientNetB0 | SVM | 86.79% | 96.78% | 88.64% | 88.39% |
k-NN | 87.37% | 96.95% | 88.82% | 88.89% | |
DT | 85.80% | 96.41% | 84.95% | 86.90% | |
RF | 66.20% | 92.60% | 81.08% | 75.95% | |
Naive Bayes | 86.11% | 96.40% | 86.58% | 86.90% | |
XGBoost | 85.86% | 96.45% | 86.38% | 87.23% | |
Ensemble | 96.17% | 98.92% | 95.89% | 95.85% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-InceptionResnetV2 | SVM | 86.27% | 96.13% | 86.86% | 86.40% |
k-NN | 87.37% | 96.48% | 88.19% | 87.56% | |
DT | 83.77% | 95.54% | 83.37% | 84.41% | |
RF | 72.93% | 93.79% | 80.67% | 79.27% | |
Naive Bayes | 79.66% | 93.41% | 78.25% | 77.78% | |
XGBoost | 87.29% | 96.47% | 88.05% | 87.56% | |
Ensemble | 97.42% | 99.40% | 97.49% | 97.68% |
TL-CNN models | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-InceptionV3 | SVM | 82.42% | 94.50% | 80.85% | 81.09% |
k-NN | 83.05% | 94.61% | 80.94% | 81.43% | |
DT | 81.54% | 94.18% | 79.65% | 80.09% | |
RF | 79.50% | 94.01% | 80.52% | 79.77% | |
Naive Bayes | 65.72% | 86.52% | 2.58% | 61.33% | |
XGBoost | 84.42% | 95.10% | 82.88% | 82.91% | |
Ensemble | 91.34% | 97.59% | 91.03% | 91.04% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-ResNet50 | SVM | 86.25% | 96.02% | 85.12% | 85.95% |
k-NN | 85.75% | 96.00% | 86.26% | 85.74% | |
DT | 82.04% | 94.94% | 82.34% | 82.59% | |
RF | 52.90% | 88063% | 71.77% | 65.67% | |
Naive Bayes | 67.71% | 89.82% | 72.49% | 64.68% | |
XGBoost | 87.63% | 96.59% | 88.27% | 87.73% | |
Ensemble | 96.46% | 99.14% | 96.76% | 96.68% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-VGG16 | SVM | 76.39% | 93.49% | 76.96% | 78.28% |
k-NN | 74.55% | 92.33% | 75.70% | 74.96% | |
DT | 57.42% | 84.86% | 55.72% | 58.37% | |
RF | 50.55% | 86.58% | 38.39% | 63.18% | |
Naive Bayes | 59.53% | 85.08% | 58.84% | 59.20% | |
XGBoost | 80.48% | 94.91% | 81.44% | 82.26% | |
Ensemble | 92.07% | 98.00% | 92.60% | 92.54% |
TL-CNN model | Machine Learning | Sensitivity | Specificity | Precision | Accuracy |
TL-VGG19 | SVM | 79.90% | 93.74% | 78.77% | 78.82% |
k-NN | 69.16% | 90.99% | 70.56% | 71.64% | |
DT | 53.23% | 82.94% | 53.41% | 54.73% | |
RF | 48.29% | 85.62% | 37.71% | 60.36% | |
Naive Bayes | 56.41% | 82.96% | 54.70% | 54.89% | |
XGBoost | 82.44% | 95.30% | 81.90% | 83.58% | |
Ensemble | 93.86% | 93.40% | 93.44% | 93.86% |
Author | Year | Method | Disease type | Dataset size | Accuracy |
Han et al. [13] | 2022 | Transfer learning with a modification of the well-known CNN models | 4-class: PCV, RAP, nAMD, and NORMAL | 4749 | 87.4% |
Sotoudeh-Paima et al. [14] | 2022 | Deep learning: multi-scale convolutional neural network | 3-class: AMD, CNV, NORMAL | 120,961 | 93.4% |
Elaziz et al. [15] | 2022 | Ensemble deep learning model for feature extraction, features selection, machine learning as classifier. | 4-class: DME, CNV, DRUSEN, and NORMAL | 84,484 | 94.32% |
Liu et al. [16] | 2022 | Deep learning based on method and lesions segmentation model. | 4-class: CNV, DME, DRUSEN, and NORMAL | 86,134 | 95.10% |
Minagi et al. [17] | 2022 | Transfer learning with DNN models | 4-class: CNV, DME, DRUSEN, and NORMAL | 11,200 | 95.3% |
Tayal et al. [18] | 2022 | Deep learning-based method | 4-class: DME, CNV, DRUSEN, NORMAL | 84,484 | 96.5% |
Proposed method | − | Hybrid of deep learning and machine learning + ensemble machine learning classifiers. | 5-class: ARMD, BRVO, CRVO, CSCR, DME | 2,998 | 97.68% |