1. ARTIFICIAL INTELLIGENCE (AI) BASED GLAUCOMA DIAGNOSIS
DETECTION USING DEEP LEARNING
TEAM MEMBERS
• Jaiselvan R
• Hariharan T
• Pagutharivalan A
GUIDE
MS.S.Uma
2. ABSTRACT
´ Deep learning (DL), a subset of artificial intelligence (AI) based on deep neural networks, has made
significant breakthroughs in medical imaging, particularly for image classification and pattern
recognition. Glaucoma is a group of eye conditions that damage the optic nerve.
´ The optic nerve sends visual information from your eye to your brain and is vital for good vision.
Damage to the optic nerve is often related to high pressure in your eye. But glaucoma can happen even
with normal eye pressure. The increased pressure in your eye, called intraocular pressure, can damage
your optic nerve, which sends images to your brain.
´ If the damage worsens, glaucoma can cause permanent vision loss or even total blindness within a few
years. It's important to have regular eye tests so problems such as glaucoma can be diagnosed and treated
as early as possible.
3. ´ Early treatment can help stop your vision becoming severely affected. To overcome this problem, we
employed ImageNet-trained models VGG16 for automatic glaucoma assessment using fundus images.
´ Results from an extensive validation using cross-validation and cross-testing strategies were compared with
previous works. It is composed glaucomatous images and normal images, which means, the largest public
database for glaucoma diagnosis.
´ The high specificity and sensitivity obtained from the proposed approach are supported by an extensive
validation using not only the cross-validation strategy but also the cross-testing validation.
´ The proposed method achieved 99.8% accuracy and also better results for other metrics.
4. INTRODUCTION
´ Deep learning is a subset of machine learning, which is essentially a neural network with three or more
layers. These neural networks attempt to simulate the behaviour of the human brain albeit far from
matching its ability allowing it to “learn” from large amounts of data.
´ Glaucoma is a group of eye conditions that damage the optic nerve. The optic nerve sends visual
information from your eye to your brain and is vital for good vision. Damage to the optic nerve is often
related to high pressure in your eye. But glaucoma can happen even with normal eye pressure.
´ Visual Geometry Group (VGG) One of CNN's deep architectures. This model consists of a different
number of convolution layers to extract the features, followed by three fully connected layers.
´ To detect Glaucoma by the proposed transfer learning models, data augmentation on the training data is
first applied. Then, the transfer learning-model VGG16 is trained through forward and backward
propagation to obtain the best weights with the lowest loss for the best performance.
5. EXISTING SYSTEM
´ A trained multiclass least-squares-support vector machine (MC-LS-SVM) classifier has been
utilized for classification.
´ This approach has been tested on two different public glaucoma database. The diagnosis of
the patients who have glaucoma. The model with CNN architecture was used to learn from
training the Glaucoma image dataset.
´ Since the existing dataset has a small number of images, this study uses the data
augmentation techniques to increase the virtual number of images.
´ Our method achieved the lowest classification accuracy with tenfold cross-validation. The
experimental results show that the proposed approach performed far better as compared to
state-of-the-art approaches.
6. Proposed system
´ The proposed Visual Geometry Group (VGG) 16 is one of CNN's deep architectures. This model
consists of a different number of convolution layers to extract the features, followed by three fully
connected layers.
´ To detect Glaucoma by the proposed transfer learning models, data augmentation on the training data
is first applied. Then, the transfer learning-model VGG16 is trained through forward and backward
propagation to obtain the best weights with the lowest loss for the best performance.
´ Afterward, the loss is calculated to update the weights in the backward propagation. Finally, through a
fully connected layer, where the outputs of the previous layer are taken and connected to each neuron
in this layer, and then obtained the output of network through sigmoid function that determines only
one output Glaucoma, Non_Glaucoma.
7. ADVANTAGES
´ The proposed method easy to recognize the disease.
´ To implementation of proposed system provide accurate classification results.
´ It classifies the results with low time complexity compared with previous methods.
´ Detect features very efficiently which gives achieving better accuracy.
8. MODULES DESCRIPTION
Image Preprocessing:
´ Since the data typically come from various origins, a strategy to guide the complexity and accuracy is
essential. Image pre-processing ensures complexity reduction and better accuracy yielded from certain data.
´ This method standardizes the data through several stages in order to feed the network with a clean dataset. In
this architecture, data pre-processing is performed through the following stages: Image standardization:
neural networks that deal with images need unified aspect ratio images.
´ Therefore, the first step is to resize the images into unique dimensions and a square shape, which is the
typical shape used in neural networks.
´ Normalization: input pixels to any AI algorithm must have normalized data distribution to enhance the
convergence of the training phase. Normalization is the action of subtracting the mean of the distribution
from each pixel and dividing by standard deviation. To achieve positive values, scaling normalized data is
considered at the end of this step.
9. Feature Extraction:
´ The main goal is to extract the neural network with various diversities, which leads to a network
that distinguishes relevant from irrelevant characteristics in the dataset.
´ Image extraction can be done using several techniques. Augmentation techniques are used
efficiently when necessary according to data availability and quality. Our proposal integrates
multiple techniques, to support a large number of dataset for different conditions
Glaucoma find by VGG 16:
´ The transfer learning-model VGG16 is trained through forward and backward propagation to obtain
the best weights with the lowest loss for the best performance. A
´ forward, the loss is calculated to update the weights in the backward propagation. Finally, through a
fully connected layer, where the outputs of the previous layer are taken and connected to each
neuron in this layer.
´ VGG 16 to handle the problem of Glaucoma detection from fundus images. Transfer learning based
methods guarantee high performance with no requirement for huge training data.
10. BLOCK DIAGRAM
Input Image
Image pre processing
Feature Extraction
Glaucoma find by VGG 16
Image classification
Optimized Result
11. CONCLUSION
´ In this proposed work, an ensemble model was designed for the detection of glaucoma in the early
stage. In order to distinguish between normal and glaucomatous fundus pictures, the ensemble model
proposes using a convolutional neural network to extract feature information from the images.
´ The performance of the proposed method of ensemble architecture is compared with three CNN
architectures with VGG 16. The proposed approach is tested on various public and private data sets.
´ The performance of the proposed algorithm is better than the state-of-the-art technique. The proposed
ensemble model yields an accuracy, sensitivity and specificity using VGG 16 data set.
´ Experiments conducted on both public and private data sets show that the proposed model
outperforms traditional computer-aided diagnosis algorithms and the convolutional neural network
architecture