研究目的
To design and implement deep convolutional neural networks to identify the presence of an exudate, and thereby classify it into Diabetic Retinopathy, Glaucoma, and/or Cataract.
研究成果
The paper proposes a system for detecting the presence and probability of Diabetic Retinopathy, Glaucoma, and Cataract using a Convolutional Neural Network model. The system's architecture and implementation details are discussed, along with the quadratic kappa metric for evaluating accuracy. Future work could expand the analysis to include more diseases and use larger datasets.
研究不足
The dataset's heterogeneity in lens quality and camera model might affect the appearance of retinal images. Some images are through the lens of a microscope, varying from camera quality images. The system's accuracy could be impacted by these variations.
1：Experimental Design and Method Selection:
The system uses deep convolutional neural networks (CNN) with three convolution layers, each having a depth of 32, followed by a Max Pooling layer, a dense layer, and an output layer with softmax nodes.
2：Sample Selection and Data Sources:
The dataset consists of about 10,00,000 labeled images of left and right eyes, collected from multiple sources, with around 45,000 images used for testing.
3：List of Experimental Equipment and Materials:
Intel Core i7 processor with 12GB RAM hosting NVIDIA GeForce 950M GTX series GPU, TensorFlow for backend, and ImageMagick for image preprocessing.
4：Experimental Procedures and Operational Workflow:
Images were preprocessed to standardize resolution, cleaned, segmented, and augmented. The CNN model was trained using stochastic gradient descent with a momentum of
5：9, batch size of 100, weight decay of 0005, and 250 epochs. Data Analysis Methods:
The accuracy of the system was evaluated using sensitivity, specificity, accuracy, and precision metrics.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
