Computer Aided Detection of Eye Diseases - Glaucoma & Diabetic Retinopathy and Watermarking of Medical Images for Tele-Ophthalmology

Glaucoma is an optic neuropathy which can cause damage to the optic nerve and result in permanent vision loss.

Under normal scenarios, the circulation cycle of aqueous humor is performed well. However, when some aberrations occur, this fluid is not able to flow out. This causes a blockage in the channel and causes a pressure inside the eye. This intra-ocular pressure creates an inward force on the retina.

Usually, an optic disc is a flat structure, which can become non-flat due to increased IOP. The pressure causes the disc to bend and results in cupping of the disc. This is clinically termed as Optic Cup.

Why Image Processing?

Currently, a clinical diagnosis for Glaucoma detection is performed under the supervision of a trained ophthalmologist. There is no such tool designed yet which can give a decision in absence of medical expert. So, an automated diagnostic tool can be used for screening purpose at primary care centres. Such a tool can act as an assisting tool and screen the affected patients so that proper medical care can be provided to them.

The medical signs such as cupping of optic disc, haemorrhages, blood vessel abnormality, peri-papillary atrophy and RNFL defects are some clinical observations used by medical experts across the globe. These clinical observations impart visual changes to a fundus image and can easily be differentiated from an image which has no signs.

Visual changes like change in intensity and colour of pixels, tortuosity of blood vessels can be observed which can be easily detected using IMAGE PROCESSING techniques. Also, the diagnosis is done by considering the clinical input from medical experts, so there are less chances of wrong diagnosis.

Challenges

Glaucoma is a disease majorly related to the optic nerve head region. Structural changes may occur in the optic nerve head region which may characterize the presence of Glaucoma. Since, the disease is related to optic nerve head region, so this region should be segmented accurately using image processing techniques. However, the presence of medical signs like bright lesions (exudates and cotton soft wools) and some artefacts which gets introduced in the images during acquisition process, affect the accuracy of optic disc segmentation. The following can be considered some important challenges which are faced during ONH segmentation:

• Exudates
• Reflections (artefacts introduced due to illumination)
• Correct ONH localization
• Choroid Vessels
• Blood Vessels overlaying on the ONH region

The use of 3 different clinical parameters make SOP-G robust in giving a decision for a fundus image.

Post diagnosis, a report is generated with a decision regarding the class of the tested fundus image. With the use of the 3 clinical parameters, the fundus images are classified as follows:

NOTE: The diagnosis report should be clinically correlated with an Ophthalmologist.

Diabetic retinopathy also known as diabetic eye disease, is a medical condition in which damage occurs to the retina due to diabetes. It can eventually lead to vision loss and blindness.

• Progressive dysfunction of the retinal blood vessels caused by chronic hyperglycemia
• DR can be a complication of diabetes type 1 or diabetes type 2
• Initially, DR is asymptomatic, if not treated though it can cause low vision and blindness.
• Diabetic retinopathy is responsible for 1.8 million of the 37 million cases of blindness throughout the world
• People with diabetes are 25 times more likely to become blind than the general population.

• All forms of diabetic eye disease have the potential to cause severe vision loss and blindness.
• Diabetic retinopathy involves changes to retinal blood vessels that can cause them to bleed or leak fluid, distorting vision.

Fundus image with DR abnormalities

• Segmentation of Hard Exudates (HE) is a difficult task as the image characteristics of HE is similar to Optic Disc and also because of uneven illumination.
• In segmentation of Red Lesions, (microaneurysms and hemorrhages) blood vessels are hindrance as they have same texture as red lesions.
• Localization of macula is also a difficult job as the area of macula is not distinct or clear structure like blood vessel or optic disc.

Several image processing techniques including Image Enhancement, Segmentation, Morphology, Geometrical based feature extraction and Classification has been developed for the early detection of DR on the basis of features such as blood vessels, exudates hemorrhages and microaneurysms.

• A GUI based prototype has been developed for computer aided diagnosis of Retinal disease Diabetic Retinopathy generates a diagnosis report with a single click.
• This GUI based prototype for DR diagnosis is user friendly and can be operated by a normal computer operator.
• This prototype will be designed to be compatible with the existing computer/Laptops and medical eye image capturing cameras.
• A GUI based prototype has been developed in which a patient identity is inserted in a perceptually transparent manner including patient information in the medical image without losing the medical information.

Signs of Diabetic Retinopathy:

In this system the grading of NPDR is done on the basis of no. of two clinical signs of NPDR, bright and dark lesion i.e. hard exudates, red spots, cotton wools, hemorrhages. The following rule is used to grade the input image into different categories

Here,
• MA - Microaneurysms,
• HE - Hard Exudates,
• HM - Hemorrhages

NOTE: The diagnosis report should be clinically correlated with an Ophthalmologist.