Red Lesion Detection Using Dynamic Shape Features for Diabetic Retinopathy Screening

ABSTRACT:

The development of an automatic telemedicine system for computer-aided screening and grading of diabetic retinopathy depends on reliable detection of retinal lesions in fundus images. In this paper, a novel method for automatic detection of both microaneurysms and hemorrhages in color fundus images is described and validated. The main contribution is a new set of shape features, called Dynamic Shape Features, that do not require precise segmentation of the regions to be classified. These features represent the evolution of the shape during image flooding and allow to discriminate between lesions and vessel segments. The method is validated per-lesion and per-image using six databases, four of which are publicly available. It proves to be robust with respect to variability in image resolution, quality and acquisition system. On the Retinopathy Online Challenge’s database, the method achieves a FROC score of 0.420 which ranks it fourth. On the Messidor database, when detecting images with diabetic retinopathy, the proposed method achieves an area under the ROC curve of 0.899, comparable to the score of human experts, and it out performs state-of-the-art approaches.

PROJECT OUTPUT VIDEO:

EXISTING SYSTEM:

• A common methodology adopted in the literature for combined MA and HE detection consists in identifying all dark-colored structures in the image, mainly through a thresholding, combined with adapted preprocessing and then in removing the vessels from the resulting set of candidates. Vessel detection is performed using either a multilayer perceptron or multiscale morphological closing.

DISADVANTAGES OF EXISTING SYSTEM:

• The major limitation to this approach is that most of the false positives at the vessel segmentation step are actually lesions. After their removal along with the detected vessels, these lesions are lost and not retrieved in subsequent processing.

PROPOSED SYSTEM:

• The proposed method takes as input a color fundus image together with the binary mask of its region of interest (ROI). The ROI is the circular area surrounded by a black background. It outputs a probability color map for red lesion detection.

• The method comprises six steps.

• First, spatial calibration is applied to support different image resolutions.

• Second, the input image is preprocessed via smoothing and normalization.

• Third, the optic disc (OD) is automatically detected, to discard this area from the lesion detection.

• Fourth, candidate regions corresponding to potential lesions, are identified in the preprocessed image, based on their intensity and contrast.

• Fifth, the DSF together with color features are extracted for each candidate.

• Sixth, candidates are classified according to their probability of being actual red lesions.

ADVANTAGES OF PROPOSED SYSTEM:

• DSFs have proven to be robust features, highly capable of discriminating between lesions and vessel segments.

• The results demonstrate the strong performance of the proposed method in detecting both MAs and HEs in fundus images of different resolution and quality and from different acquisition systems.

MODULES:

1. Image Acquisition

2. Preprocessing

3. Segmentation

4. Feature Extraction

5. Classification

MODULE DESCRIPTION:

1. Image Acquisition:

Images are acquired from Gallery.

2.Preprocessing:

The objective of the preprocessing phase is to apply possible image enhancement techniques to obtain the required visual quality of the ultrasound images.

Image enhancement techniques,

1. Grayscale Image

2. Filtered Image

3. Histogram Equalization Image

1. Grayscale Image:

The ultrasound image is in RGB type which is an additive color of red, green, and blue. The image is converted into gray scale image for further processing.

2. Filtered Image:

A median filter does a very good job at reducing the noise in image.

3. Histogram Equalization Image:

The contrast enhancement of the image can be observed by applying histeq (enhance contrast using histogram equalization).

3. Segmentation:

1. Morphological Operation

2. Thresholding Method

i. Morphological Operation:

Morphology is a technique for extracting the information from an image which is representation and description of region shape. In our paper morphological operations are used in post processing mainly as a filter. Its fundamental operations are Boundary pixels and low frequency pixels are eliminated from image. Then difference image was generated.

Erosion: it shrinks objects in the binary image

Dilation: grows or thickens the objects in binary image

ii. Thresholding Method:

After, Thresholding is employed to segment red lesions in retinal images. Thresholding makes it possible to highlight pixels in an image. Thresholding can be applied to gray scale images or color images. In this discussion gray scale images are used. In Thresholding a pixel intensity value is adjusted, by taking the given value as reference the low intensity pixels will become zero and rest of the pixels will become 1. The result of the Thresholding is a binary image containing black and white pixels.

4. Feature Extraction:

i. Shape Features:

Basically, shape-based image retrieval consists of measuring the similarity between shapes represented by their features. Some simple geometric features can be used to describe shapes. Usually, the simple geometric features can only discriminate shapes with large differences; therefore, they are usually used as filters to eliminate false hits or combined with other shape descriptors to discriminate shapes.

They are not suitable to be stand alone shape descriptors. A shape can be described by different aspects. These shape parameters are Center of gravity, Axis of least inertia, Digital bending energy, Eccentricity, Circularity ratio, Elliptic variance, Rectangularity, Convexity, Solidity, Euler number, Profiles, Hole area ratio, Area, Length, Perimeter.

5. Classification:

The classification process is done over the segmented images. The main novelty here is the adoption of Random Forest. RF classifier is applied over the segmented images and the classification is done.

Random Forest:

One of the foremost common ways or frameworks utilized by knowledge scientists at the ‘rose knowledge science skilled follow cluster’ is Random Forests. The Random Forests formula is one in every of the simplest among classification algorithms -able to classify giant amounts of information with accuracy. Random Forests are associate degree ensemble learning technique (also thought of as a kind of nearest neighbor predictor) for classification associate degreed regression that construct variety of call trees at coaching time and outputting the category that’s the mode of the categories output by individual trees (Random Forests may be a trademark of Leo Bremen and Adele monger for an ensemble of call trees).

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

• System           : Pentium Dual Core.

• Hard Disk      : 120 GB.

• Monitor          : 15’’ LED

• Input Devices : Keyboard, Mouse

• Ram                 : 1GB.

SOFTWARE REQUIREMENTS:

• Operating system : Windows 7.

• Coding Language : MATLAB

• Tool                       : MATLAB R2013A

REFERENCE:

Lama Seoud*, Thomas Hurtut, Jihed Chelbi, Farida Cheriet, and J. M. Pierre Langlois, “Red Lesion Detection Using Dynamic Shape Features for Diabetic Retinopathy Screening”, IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 35, NO. 4, APRIL 2016.