A Novel Approach for Detection of Microaneurysms in Diabetic Retinopathy Disease from Retinal Fundus Images

Diabetic Retinopathy (DR) is a leading cause of blindness in human beings aged between 20 to 74 years. It has a great influence on the patient and society because it normally influences humans in their most gainful years. Early detection in DR is very difficult which is not detected by human beings. Many algorithms and techniques were established to detect DR. These techniques faced the problems such as increasing sensitivity, specificity and accuracy. To overcome those problems we have to introduce an effective image processing algorithms for increasing performances and also easily identify the DR diseases. One of the most challenging tasks in screening is automatic detection of Microaneurysms (MAs). This paper presents a new approach to detect MAs. Our proposed work consists of preprocessing, blood vessel segmentation (FPCM), fovea localization, fovea elimination, feature extraction and classification (Neuro-Fuzzy). Neuro-Fuzzy is a combined version of neural networks and fuzzy logical models. Experiments are conducted using MATLAB simulation tool. Using MESSIDOR database for our experiments which provides efficient and effective results in sensitivity, specificity, correct classification and detection rate (accuracy) and precision.


Introduction
Diabetic eye disease contains group of eye conditions that affects people with diabetes.These conditions include diseases such as diabetic retinopathy, diabetic macular edema, cataract and glaucoma these are very potential to cause severe vision loss problem and blindness problem in working age adults.Changes in blood vessels lead to the problem of diabetic retinopathy since it affects the lining of back of the eye.This will called as retina.DR has a significant impact on the world health organization systems.The number of people with DR will grow from 126.6 million in 2010 to 191.0 million by 2030.In general, DR is a silent disease (Rajan, 2015) , (Kanika Verma, Prakash Deep & Ramakrishnan, 2011) because this is identified by the patient when the level changes in the retina.Some common symptoms of diabetic retinopathy are given below: • Blurred vision

• Fluctuating vision
• Impaired color vision

• Floaters vision
• Flashers vision (dark spots) • Vision in dark and empty areas The effect of diabetic retinopathy on vision is varies widely, depending on the stage of the disease.Generally DR has two stages such as Proliferative Diabetic Retinopathy (PDR) and Non-Proliferative Diabetic Retinopathy (NPDR).PDR has the components of neovascularisation and vitreous fluid hemorrhage since new blood vessels grow on the surface of the retina and it can bleed.But in NPDR has no symptoms which are detect only by retinal photography.Diabetic NPDR stages are classified into three: Mild, moderate and severe (Latare & Patil, 2015): cis.ccsenet.

FPCM seg
Step 1: Ob Where u ik notations a point n.
Step 2: Ca Step 4: Co We have membersh applicable According
Where x, y denotes the NPV and RPV respectively and p(x,y) denotes the probability of maximum pixel value(NPV and RPV) (II) Contrast (Cn): (III) Entropy (Ep): IV) Angular Second Moment (Asm): VII) Mean (Mn): Where n is the number of rows and m is the number of columns.

F) Classification:
Classification is a final step of our DR disease identification process.After extracting the features we have to classify diseases into Microaneurysms and Non-Microaneurysms.For classification purpose, features as an input that is fed to a classifier based on neural network and fuzzy logical models.Fuzzy logics and neural networks are natural complementary tools in building for classification.Neural networks and fuzzy logic are two approaches that are widely used to solve classification problem.While networks are low-level computational structures that perform well when dealing with raw data, fuzzy logic deals with reasoning on a higher level.A fuzzy logical system is a non-linear mapping of features into a scalar output.The fuzzy model is used for giving more accurate results when adding more number of features into fuzzy models.The main advantage of neural networks is their learning capabilities and their ease of implementation.When combined neural networks and fuzzy logic it exploits more number of advantages.
Feature sets such as Mp, Em, Ep, Asm, Hm, Ds, Mn and Cr are extracted from the input images (fundus image) using GLCM.The fundus images are classified using the Neuro-Fuzzy classifier.Extracted features are given as the input to the Neuro-Fuzzy Classifier which is classified by all the given fundus images into 2 classes    ) and to the ber of rated input max ble to classify whether the given input fundus image is Microaneurysms or Non-Microaneurysms (Bhanumurthy & Koteswararaa, 2014).

Experimental Analysis
Our proposed Microaneurysms detection and classification of retinal disease should be processed by segmentation and classification steps.Initially pre-processing fundus image by reduces the noise level and also performed with contrast enhancement.Blood vessel segmentation done with the help of Fuzzy Possibilistic C-Means should contain segmentation information about every retinal image.In classification of diabetic retinopathy we have to use Neuro-Fuzzy classifier which increases sensitivity and accuracy.Comparison shall be made for classification.For classification we have to compare previous classifiers with our Neuro-Fuzzy.

Database
MESSIDOR database (Methods for Evaluating Segmentation and Indexing techniques for Dedicated to Retinal Ophthalmology) is used in our proposed system.Generally, MESSIDOR database consists of two sets of data such as training set and evaluation set.Here, we evaluate this database for identifying Microneurysms and Non-Microneurysms.MESSIDOR database contains 1200 retinal images, which is the largest database publicly available on the internet.The images will be saved as uncompressed TIFF format with a 1440 * 960 pixel resolution that is about 4MB per image.These images acquired by 3 ophthalmologic departments using a color video 3CCD camera on a Topcon TRC NW6 non-mydriatic retinograph with a 45 0 Field of View (FOV).The images captured using 8 bit per color on plane at the pixel ranges are 1440*960, 2240*1488, 2304*1536.1200 images are divided into two sub images i.e. 800 images acquired with pupil dilation (one drop of Tropicamide at 0.5%) and 400 images are without dilation.

(i) Training Set:
This dataset is used for testing and improving the available algorithms as well as for validating the methods used to evaluate the algorithms.This database contains 200 images.
For each image, it indicated at least: • The number of and/or the surface of the micro aneurysms and Non-Microaneurysms.

(ii) Evaluation Set:
This dataset contains thousand images since its purpose is evaluating the proposed algorithms.
Usually, Diabetic Retinopathy stages are classified as mild, moderate, severe and PDR.But in our proposed process we have to categorize diseases like Microaneurysms and Non-Microaneurysms.Our dataset is processed and improve classification accuracy.

Performance Metrics
Performance of the test classifier can be measured in the form of sensitivity (or) Recall, specificity and Accuracy (Correct classification and Detection rate).True positive (TP), True Negative (TN), False Positive (FP) and False Negative (FN) are the test outcomes.In general True positive is correctly identified, False Positive is incorrectly identified, True negative is correctly rejected and finally false negative is incorrectly rejected.Performance metrics such as sensitivity, specificity and accuracy are in follows:

(i) Sensitivity:
Sensitivity can be measured by the proportion of positives, disease affected peoples can be correctly identified.This can be computed as follows: (ii) Specificity: Specificity can be measured by the proportion of negatives, peoples does not could affected are correctly identified.This can be computed as follows: Figure 1 il Hemorrhag Microaneu Hemorrhag Hemorrhag geometrica Hard exud Yellow spo Many tec supervised (Janakiram to achieve approxima robustness approach p Figure

Figure
Figure 7. F

Figure
Figure 8. F

Figure
Figure 10.N Figur

Table 3
of 99% and accuracy is 99%.In future, we will decide to separate the diseases with some different properties and use different database for processing.