click here to download project abstract
click here to download project abstract
ABSTRACT
Retinal Disease Diagnosis using Deep Learning Is a system to process medical images using Deep Learning. It is a subset of Artificial intelligence (AI) which has the potential to revolutionize the disease diagnosis and management by performing classification difficult for human experts and by rapidly reviewing immense amounts of images. Despite its potential, the preparation and implementation of Artificial Intelligence algorithms is always challenging. In this Identifying medical diagnoses and treatable diseases by Image-based Deep Learning system, we sought to develop an effective transfer learning algorithm to process medical images to provide accurate predictions for diagnosis of the respective patient.
The retinal disease diagnosing process involves optical coherence tomography (OCT) images of the retina. The model (neural network) is trained on the OCT images to recognize the distinguishing features of specific classes of images of the human eye, faster and with fewer training examples and less computational power. Applying a conventional approach to a dataset of optical coherence tomography images, we can demonstrate performance comparable to that of human experts in classifying age related macular degeneration and diabetic macular edema. We can also provide a more transparent and interpretable diagnosis by highlighting the regions recognized by the neural network.
Deep learning is a branch of machine learning which is completely based on artificial neural networks, as neural networks are going to mimic the human brain so deep learning is also a kind of mimic of the human brain. It is a multidisciplinary field and aims to develop automated systems that are able to extract information from the unstructured data. Our aim is to classify a given Retinal OCT(Optical Coherence Tomography) image into one of the four different classes of Eye Diseases(CNV, DME, DRUSEN, NORMAL RETINA).
Keywords: Optical Coherence Tomography, Artificial Intelligence, Deep Learning, Neural Networks, Choroidal Neovascularization(CNV), Macular Degeneration, Diabetic macular edema.
INTRODUCTION
Introduction to Retinal disease:
Diabetes is a very common disease worldwide. It is the most common cause of blindness for many people. Many researchers acknowledged that 90 percent of diabetic patients could be saved from this disease through an early diagnosis. A person having diabetes is more prone to the risk of diabetic retinopathy (DR) and macular degeneration (MD). The blood supply towards all layers of the retina is done through micro blood vessels which are susceptible to unrestrained blood sugar level. As we know, for the people with diabetes, the glucose and fructose levels in the blood will be more or less than the actual required amount. The increased levels of sugars (glucose or fructose) get accumulated in the blood vessels of the eye and they start crumbling because the oxygen distribution to the cells is diminished.
Any blockage in these vessels results in a severe eye injury. As a result, metabolic rate slows down and leads to structural abnormality in vessels which intern DR and MD. DR normally has two different stages named as proliferative DR (PDR) and non – proliferative DR (NPDR). It is a main reason for blindness. The development is at different rates due to the 2 important vision pressuring difficulties: diabetic macular edema (DME) and proliferative retinopathy.
Ophthalmologists can check for diseased eyes as part of a dilated eye exam. If they find any symptoms of diabetic eye disease, they may do a test called a fluorescein angiogram. This test lets them see pictures of the blood vessels in the retina. They will take an Optical Coherence Tomography(OCT) image for the patient’s eye. OCT shows how thick the retina is and can identify accumulated fluid from abnormal blood vessels in the retina. This helps the ophthalmologist to classify and identify the disease.
In this system, we establish a diagnostic tool based on a deep-learning framework for the screening of patients with common treatable blinding retinal diseases. The objective of computer-aided diagnosis is to recognize diseased and normal images. It effectively classifies images for macular degeneration and diabetic retinopathy. In this system, we use a technique.
called Transfer learning which has proven to be a highly effective technique, particularly when faced with domains with limited data. This diagnostic tool has the potential to produce high accurate and precise diagnostic predictions on the Optical Coherence Tomography (OCT) image data of the retina.
Multilayer perceptron
Multi layer perceptron (MLP) is a supplement of feed forward neural network. It consists of three types of layers—the input layer, output layer and hidden layer. The input layer receives the input signal to be processed. The required task such as prediction and classification is performed by the output layer. An arbitrary number of hidden layers that are placed in between the input and output layer are the true computational engine of the MLP. Similar to a feed forward network in a MLP the data flows in the forward direction from input to output layer. The neurons in the MLP are trained with the back propagation learning algorithm. MLPs are designed to approximate any continuous function and can solve problems which are not linearly separable. The major use cases of MLP are pattern classification, recognition, prediction and approximation.
Fig : Multilayer Perceptron (MLP)
Convolutional Neural Networks (CNN)
A convolutional neural network, or CNN, is a deep learning neural network designed for processing structured arrays of data such as images. Convolutional neural networks are widely used in computer vision and have become the state of the art for many visual applications such as image classification, and have also found success in natural language processing for text classification.
Convolutional neural networks are very good at picking up on patterns in the input image, such as lines, gradients, circles, or even eyes and faces. It is this property that makes convolutional neural networks so powerful for computer vision. Unlike earlier computer vision algorithms, convolutional neural networks can operate directly on a raw image and do not need any preprocessing.
A convolutional neural network is a feed-forward neural network, often with up to 20 or 30 layers. The power of a convolutional neural network comes from a special kind of layer called the convolutional layer.
The architecture of a convolutional neural network is a multi-layered feed-forward neural network, made by stacking many hidden layers on top of each other in sequence. It is this sequential design that allows convolutional neural networks to learn hierarchical features. The hidden layers are typically convolutional layers followed by activation layers, some of them followed by pooling layers. A simple convolutional neural network that aids understanding of the 10core design principles is the early convolutional neural network LeNet-5, published by Yann LeCun in 1998. LeNet is capable of recognizing handwritten characters.
Fig -Sample CNN architecture
