WO2010131944A2

WO2010131944A2 - Apparatus for monitoring and grading diabetic retinopathy

Info

Publication number: WO2010131944A2
Application number: PCT/MY2010/000077
Authority: WO
Inventors: Fadzil B. Mohamad Hani Ahmad
Original assignee: Institute Of Technology Petronas Sdn Bhd
Priority date: 2009-05-13
Filing date: 2010-05-13
Publication date: 2010-11-18
Also published as: DE112010002000T5; MY147093A; WO2010131944A3

Abstract

The present invention relates generally to an apparatus for ascertaining the area of foveal avascular zone (FAZ) (802) of the retina based on the digital map of retinal vasculature for reliable determination of the FAZ area (802) to assist in monitoring and gradin diabetic retinopathy.

Description

APPARATUS FOR MONITORING AND GRADING DIABETIC

RETINOPATHY

1. TECHNICAL FIELD OF INVENTION

The present invention relates generally to an apparatus for monitoring and grading diabetic retinopathy to enable effective identification of those individuals who are at risk of developing diabetic retinopathy and assisting in achieving accurate diagnosis in an efficient manner.

2. BACKGROUND OF THE INVENTION

The finest blood vessels linking arteries to veins, called retinal capillaries (very thin or micro vessels); tend to be damaged due to complications of diabetes mellitus. This progressive damage called diabetic retinopathy occurs due to a combination of micro-vascular leakage and micro-vascular occlusion

Diabetic Retinopathy (DR) is a common complication of diabetes mellitus which is caused by the damage on the retinal vasculature and is the leading cause of blindness in the working-age population. It is a silent disease and may only be realized by the patient when the changes in the retina have progressed to a level where treatment is complicated and nearly impossible. The incidence of DR has increased with increase of life expectancy of diabetics. Roughly, about 50 percent of diabetic patients develop DR after 10 years, 70 percent after 20 years and 90 percent after 30 years of onset of the diabetes. In Malaysia, the diabetic population has increased over four-fold from 300,000 in 1996 to nearly 1.4 million in 2005. In 2007, nearly 15% of the population suffers from diabetes mellitus and about 37% of the diagnosed diabetic population has any form of diabetic retinopathy [National Eye Database 2007].

At the initial stage of the disease, there are often no obvious warning signs and patients suffering from the disease are unaware until it has advanced into more severe levels. Treatment of the disease at this stage may prevent future complications leading to blindness. As the disease progresses, clinically evident retinopathy (pathology) appears namely microanerysms, dot and blot haemorrhages, cotton wool spots, venous caliber changes and retinal capillary non-perfusion. However, changes at hemodynamic (physical aspects of the blood circulation) and cellular levels already take place even when there is no clinically detectable retinopathy. It is believed that visual loss occurring from Diabetic Retinopathy can be prevented by a periodic follow-up that is very important for a timely intervention to reduce the risk of blindness in diabetic patients. Diabetic patients with no Diabetic Retinopathy or mild Diabetic Retinopathy should have their eyes checked by an ophthalmologist at least once a year. Proper screening for retinopathy followed by laser surgery treatment can significantly reduce the incidence of blindness. Screening in Diabetic Retinopathy is a non-diagnostic test or identification of individuals who may be at risk of developing Diabetic Retinopathy performed not only by ophthalmologists but also by highly trained clinicians or medical staff.

Known methods of screening Diabetic Retinopathy is by means of analyzing a patient's fundus photograph (retinal image produced by a fundus camera which is a specialized camera capable of producing fundus images in film or digital form) to identify pathologies for grading purposes. The grades or stages of severity Diabetic Retinopathy are identified by reference to pathologies and symptoms shown on the fundus photographs or images. Therefore, the ophthalmologist will have to carefully study the photographs and diagnose the severity of retinal pathologies. This conventional method [also known as Physician's Global Assessment (PGA)] of making diagnosis requires a high degree of skill and experience and is not only very costly in both time and money but also results in inter and intra variations. The high cost of examination and the shortage of ophthalmologists or highly trained clinicians, particularly in undeveloped or rural areas in developing countries, are prominent factors that hamper patients from obtaining regular examinations. Further mass screening of a large number of diabetic patients annually poses a huge workload for the ophthalmologists or highly trained clinicians as they are required to examine a prohibitively large number of fundus images and the number of images produced without any sign of Diabetic Retinopathy, is extremely large compared to the ones having signs. Potential of inconsistent judgment can occur due to varying degrees of experience of ophthalmologists and highly trained clinicians and so the diagnosis is therefore subjective. Errors due to susceptibility of the observer are inevitable. Such errors will either result in the delay of patients being referred to ophthalmologist for further examination and treatment or false alarm being raised. The degree of severity varies over the retina (fundus) and therefore assessment of the complete fundus is necessary. The optimal time for treatment is before the patient experiences visual symptoms. Therefore, early detection of the disease through regular screening and timely treatment is very crucial to prevent virtual loss and blindness. It is also very important to recognize the stages in which treatment may be beneficial.

Although there currently exists an automated digital system by means of computing techniques to preliminarily detect whether or not any suspicious signs of Diabetic Retinopathy are present in a fundus image to reduce variability and to improve efficiency, as only those individuals who are diagnosed by the automatic digital system (automated preliminary diagnosis) as having early retinal lesions would directly be referred to an ophthalmologist for further evaluation to enable the ophthalmologists to attend to those patients who are actually in need of their expertise there still exist setbacks. This is because typical diagnosis for grading Diabetic Retinopathy progression into its various stages is based on the several pathologies present in the retina individually or in combination. For example, Non-proliferative Diabetic Retinopathy (NPDR) is indicated by the presence of micro aneurysms, retinal haemorrhages, hard exudates and retinal edema, Pre-proliferative Diabetic Retinopathy (PPDR) by lesion features namely cotton wool spots, venous beading and intra-retinal micro-vascular abnormalities (IRMA) and Proliferative Diabetic Retinopathy (PDR) by micro- vascular occlusion (blockage of capillary network) extensively throughout the retina. This means one type of pathology may be used in the diagnosis of each stage of Diabetic Retinopathy or a few types of pathology depending on the images of the pathologies studied. In short, there is no one single type of pathology that can be used in determining all the grades of Diabetic Retinopathy progression. The same goes for diagnosing diabetic maculopathy that is based on the pathologies found in the retina mostly at the macular area. Further, due to the study of various types of pathologies present in the retinal vasculature, the person analyzing the retinal images must be one who is highly trained medical personnel and experienced so as to be able to detect changes, indicating different stages of diabetic retinopathy progression, in each and every type of pathology. The skilled person must therefore be one who has studied in detail the characteristics and changing patterns of each and every type of pathology, who is none other than an ophthalmologist or his trained staff. And because it involves so many types of pathologies the skilled person being human may make errors leading to inaccurate or wrong diagnosis due to the quality of images obtained as a result of inherent images produced by the retina imaging system (302). Although analysis of the various types of pathologies and their changes can be automated, such automation would require use of several different software to cater for the different types of pathology studied, their inconsistent images under variable colour tones (due to differing makes and quality of digital cameras used) and detection of their characteristics and changes that is costly.

It is therefore advantageous if the various stages of diabetic retinopathy progression can be graded using only one single type of structure present in the retinal images such as by means of using the digital map of retinal vasculature to assist in detection of early signs of Diabetic Retinopathy, and the grading of diabetic retinopathy progression by means of determining the degree of enlargement of foveal avascular zone (FAZ) since it has been found there is a correlation between the size of the FAZ and Diabetic Retinopathy progression. In a normal healthy person, there is a region in the macula where there are no vessels in the entire fovea namely FAZ. FAZ is therefore the fovea where there is no blood vessels and is normally located in the very centre of the macula. However, an enlargement of FAZ is usually found in eyes of patients with diabetic retinopathy resulting from a loss of capillaries in the perifoveal capillary network. This is often observed in early Diabetic Retinopathy such as NPDR and also in PDR. The capillaries surrounding an FAZ region may possibly tend to be blocked or damaged as a result of diabetes. It is found that the FAZ dimensions were strongly and positively correlated with the severity of capillary non-perfusion (blockage of capillary) and the presence of proliferative diabetic retinopathy. Early detection of FAZ enlargement at NPDR stage may prevent the progress of the disease to PDR stage and towards visual loss. However, one is not able to currently determine the area of FAZ and its enlargement over time accurately and reliably based on merely studying coloured fundus images and comparing fundus images of a patient taken at different intervals. Fundus image analysis presents several challenges such as image variability (due to pathologies of different patients and due to differing imaging conditions for the same patient), low image contrast of blood vessels against the macular region, improper illumination, glare, fadeout, loss of focus and artefacts arising from reflection, refraction and dispersion. Therefore, an apparatus to enable suitable image enhancing technique is needed to provide a reliable fundus image for analysis of DR based on the

FAZ.

The present invention overcomes the above shortcomings by providing an apparatus for ascertaining the area of FAZ of the retina to effectively analyse and thereafter grade various stages of Diabetic Retinopathy progression so as to enable and assist the physician in screening and monitoring the severity of

DR in an objective manner as early as possible.

3. SUMMARY OF THE INVENTION

Accordingly, it is the primary aim of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases based on the sole criteria of ascertaining the size of the FAZ instead of changes in the various pathologies in the retina.

It is yet another object of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases which enables the determination of FAZ that is bordered by capillary ends in the macular region with good accuracy.

It is yet another object of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases which is able to provide a comprehensive grading for severity of Diabetic Retinopathy and Diabetic

Maculopathy.

It is yet another object of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases wherein the initial task of analyzing the retinal images is automated thereby dispensing with highly trained medical personnel during initial screening of patients thereby saving cost.

It is yet another object of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases which is capable of reducing the huge workload of professional ophthalmologists thus freeing them from performing repetitive tasks in initial screening of Diabetic

Retinopathy and permitting them to concentrate on patients that require their expertise.

It is yet another object of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases which is efficient and reliable thereby allowing detection of Diabetic Retinopathy progression at an early stage thus enabling prevention of progress of the disease by directly referring the patient to a professional. It is yet another object of the present invention to provide an apparatus for analyzing the retina for ocular manifested diseases capable of allowing images of tiny vessels to be detected and extracted for further study.

It is yet another object of the present invention to provide an apparatus for automatically analyzing the retina for ocular manifested diseases which is safe as it does not require any invasive method or contrasting agent to be injected into the patient for a better image production.

Other and further objects of the invention will become apparent with an understanding of the following detailed description of the invention or upon employment of the invention in practice.

According to a preferred embodiment of the present invention there is provided,

An apparatus for analyzing the retina for ocular manifested diseases comprising,

at least a retina imaging system (302), comprising at least one retina imaging device and at least one image capturing device;

at least one any acceptable processing means (304),

at least one user interface system (306),

characterized in that said any acceptable processing means (304) is specifically provided with a predetermined algorithm using any acceptable programming software to cater for low and varying contrast image to enhance the retinal vessels (104) from its background prior to extraction of the enhanced retinal vessels for ascertainment of the area of the foveal avascular zone (802);

further characterized in that

said any acceptable processing means (304) is further provided with a predetermined algorithm to ascertain the relative area of the foveal avascular zone (802) to diagnose or grade the disease by means of computerized techniques.

4. BRIEF DESCRIPTION OF THE DRAWINGS

Other aspects of the present invention and their advantages will be discerned after studying the Detailed Description in conjunction with the accompanying drawings in which:

FIG. 1 is a digital fundus image showing several structures present in the retina.

FIGS. 2A and 2B show two tables namely the International Clinical Diabetic Retinopathy Disease Severity Scale and the International Clinical Diabetic Macula Edema Disease Severity Scale.

FIG. 3 shows a highest level block diagram of the apparatus for monitoring and grading diabetic retinopathy. FIG. 4 shows a block diagram of the flow for one of the algorithms in the said processing system using CLAHE to perform image enhancement.

FIG. 5A shows a block diagram of the flow in one of the stages stated in FIG 4 called image enhancing using CLAHE.

FIGS. 5B and 5C show the flow similar to FIG 5A but represented in another form.

FIGS. 6A and 6B show two fundus images one prior to being fed into an image enhancing stage and the other subsequent to it.

FIG. 7 shows a block diagram of the flow in one of the steps staged in FIG 4 called extracting blood vessel using morphological technique such as bottom hat to extract the enhanced retinal vasculature.

FIG. 8 shows an illustration of the FAZ area which is plotted and delineated using the semi-automated means.

FIG. 9 shows a block diagram of the flow for another algorithm that can be implemented in the said processing system.

FIGS. 1OA and 1OB show block diagram of the flow in one of the steps stated in FIG 9 called enhancing image and extracting blood vessel using PCA and FastICA. FIGS. HA and HK show an example of the layout of the user interface system as a platform for users to key in appropriate information into the processing system to be analyzed.

5. DETAILED DESCRIPTION OF THE DRAWINGS

In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be understood by those or ordinary skill in the art that the invention may be practised without these specific details. In other instances, well known methods, procedures and/ or components have not been described in detail so as not to obscure the invention.

The invention will be more clearly understood from the following description of the components of the apparatus thereof, given by way of example only with reference to the accompanying drawings. In the descriptions that follow, like numerals represent like elements in all figures. For example, where the numeral (2) is used to refer to a particular element in one figure, the numeral (2) appearing in any other figure refers to the same element.

Referring to FIG. 1, there is shown a digital fundus image showing several structures present in the retina (100). Many retinal diseases and systematic diseases will lead to evident fundus changes that can be observed by fundus photographs. Fundus images are presented as arrays of pixels having discrete intensity values. Analyzing and interpreting fundus images have become a necessary and important diagnostic procedure in ophthalmology. The vasculature in the retinal layer, referred to hereafter as the retinal vasculature can be viewed non-invasively using the fundus camera. The bright round structure in FIG. 1 is known as the optic disc (102). Retinal vessels (104) comprising arteries and veins are seen as dark red curvature lined structure. The darker region at the right side of the image is known as the macular (106). Although the macular (106) has been delineated as shown in FIG. 1, it is to be understood that the said delineation is not the exact size but is delineated for easy reference. The centre of the macular (106) with no vessels is the most accurate vision zone and is known as the fovea (108).

Typically the grading for Diabetic Retinopathy and Diabetic Maculopathy are performed based on the pathologies present in the retinal vasculature as shown in FIG. 2-A and 2-B which respectively illustrate two tables namely the International Clinical Diabetic Retinopathy Disease Severity Scale and the International Clinical Diabetic Macula Edema [Diabetic Macula Edema is a specific type of Diabetic Maculopathy] Disease Severity Scale

As is known, the dimensions and the changes of the vascular tree induced by the progression of Diabetic Retinopathy can become valuable diagnostic indices of the disease and therefore, the morphological changes in the retinal vasculature are of great interest to ophthalmologists. However, image analysis of the retinal vasculature is a complicated task particularly because of the variability of the fundus images in terms of the colour or gray levels, the diverse morphology of the retinal anatomical pathological structures and the existence of particular features in different patients that may lead to an erroneous interpretation. There are several challenges of retinal vessels extraction that may be outlined as follows:-

(a) A wide range of vessel widths - from a pixel to 12 pixels wide (b) Low contrast of vessels to its surrounding areas. Narrow vessels often have the lowest contrast.

(c) A variety of non-vessel structures appearing in the surroundings, including the border of the camera's aperture, the optic disc, and pathologies (lesion features). The lesion features may appear as a series of bright spots, sometimes with narrow darker gaps in between which are a challenge for automatic vessel extraction.

(d) Some wider vessels have a bright strip along the centre called the "central reflex", causing a complicated intensity cross-section. This may be hard to be distinguished locally from two side-by-side vessels.

Therefore, to enable an accurate study of the retinal vasculature for determination and analysing the area of FAZ (802), to reliably and accurately diagnose stages of ocular manifested diseases, digital images need to first undergo an image processing technique automatically by intelligent computerized analysis systems whereby a technique is employed to manipulate digital images to preferably perform image enhancement, image noise removal, image segmentation or extraction and image statistical extraction to facilitate determination of the area of FAZ (802) based on the extracted vessel.

Referring to FIG. 3, there is shown a highest level block diagram of the apparatus for monitoring and grading diabetic retinopathy. It contains three major components: retina imaging system (302), processing system (304) and user interface (306). The said retina imaging system (302), preferably a fundus camera, comprises of two components: retina imaging device (308) and image capturing device (310). The retina imaging device (308) enables the image of the retina to be shown when the patient's eyes are examined. The image capturing device (310) is attached to the said retina imaging device (308) in order to capture the image of retina that is shown. The said image capturing device (310) can be of a digital camera. The functionality of the retina imaging system (302) is to capture images of retina. The said retina imaging system (302) is a specialized low powered microscope with an attached camera designed to photograph the interior surface of the eye (fundus / retina). The said retina imaging system (302) provides an upright, magnified view of the fundus / retina. A typical fundus camera views 20 to 50 degree of the retinal area. The fundus camera of the present invention is developed using 45 to 50 degrees of camera view, however it is also possible to operate the retina imaging system (302) using 20 to 30 degrees of camera view.

Once a button on the fundus camera is being pressed, the image capturing device (310) captures the image of retina that is shown on the retina imaging device (308). After capturing the image of the retina, said image is automatically transferred to said processing system (304).

Once said processing system (304) receives images from said image capturing device (310), it performs processing and analysis to the said image. The user interface (306) functions as a platform for the user to provide input to said processing system (304) through any acceptable input means such as keyboard and mouse or obtain results of analysis from said processing system (304) through any acceptable displaying means such as a monitor.

Two types of algorithm can be implemented in said processing system

(304). Referring to FIG 4, there is shown a block diagram of the flow for one of the algorithms in the said processing system (304). The software program that is being used by the said processing system (304) can be Visual C++ or

Matlab. After the image of retina is being captured by the retina imaging system (302) and transferred to the said processing system (304) through the image capturing device (310), said processing system (304) performs the following steps: i. performing pre-processing to said image (402); ii. enhancing image using CLAHE (404); iii. extracting blood vessel using morphological technique such as bottom hat (406); iv. detecting blood vessel end-point (408); v. performing analysis (410).

As shown in FIG. 3 after the said retina imaging system (302) captures the fundus image, said image is transferred to the processing system (304) wherein the said input image undergoes a preliminary processing technique (402) applied to enhance the vessels (104) and reduce background noise including border of the aperture of the retina imaging system (302). As the green channel shows the best contrast of vessels (104) to their surrounding, the green band of the input image is amongst the first to be extracted in the image pre-processing stage (402). In this image pre-processing stage (402), a combination of geometric operation and median filtering is performed on the green band image. Geometric operation such as scaling up the image to preferably two times larger than its original size is applied prior to median filtering to preserve one-pixel width vessels. This is done so that one-pixel width vessel will not be eliminated during median filtering that will be performed to smooth out the image. Median filtering is chosen due to its smoothing while preserving edges effect. After median filtering, the image will be scaled down to its original size resulting in a pre-processed image. Referring now to FIG. 5A, there is shown a block diagram of the flow in one of the stages stated in FIG 4 called image enhancing using CLAHE. Meanwhile, FIG. 5B and FIG. 5C are the flow similar to FIG. 5A but represented in another form. Due to poor illumination, lack of dynamic range in the imaging sensor or even wrong settings of lens during image acquisition, low-contrast images may still result. Thus the pre-processed image goes through an image enhancing stage (404) where it undergoes an enhancing technique suitable for low and varying contrast images; where the lowest and highest occupied bins are close to the minimum and the maximum of the full range of pixel values that the image type concerned allows. The said image enhancing stage (404) employs a technique which is different from the global contrast stretching technique normally used in enhancing vessels located in both dark and bright regions where the said contrast stretching only attempts to increase the dynamic range of the gray levels in the image being processed and works by stretching the range of intensity values it contains to span a desired range of values (that is the full range of pixel values that the image type concerned allows). Preferably in the said image enhancing stage (404) a window (tiles) based enhancement technique called contrast limited adaptive histogram equalization (hereinafter referred to as "CLAHE") is employed. Preferably the said stage (404) utilising

CLAHE comprises various sub-stages: partitioning sub-stage (502), performing histogram equalization sub-stage (504) and aligning using bilinear interpolation sub-stage (506). For partitioning sub-stage (502), an input image is partitioned into small regions or blocks as indicated by the reference letters (A), (B), (C) and (D) as shown in FIG 5B for histogram equalization (504) to be applied to each one resulting in small histogram equalized regions or blocks as indicated by the reference letters (A'), (B'), (C) and (D') of FIG 5B. The performing histogram equalization sub-stage (504) comprises of a clipping sub-stage (508) to clip the height of histogram in each contextual region to avoid amplifying any noise that might be present in the image thus limiting the contrast especially in uniform areas. When the histogram is clipped as shown by shaded box as indicated by FIG. 5C, the number of excess pixels is counted and then they are redistributed across the whole histogram (504). The redistribution of the pixels is to spread out the frequency of the intensity value. This will eventually increase the contrast between the blood vessels and the other areas. The aligning using bilinear interpolation sub-stage (506) is to calculate the new gray level allocation of pixels within the small block by using bilinear interpolation to reduce or eliminate artificially induced boundaries.

Although FIG. 5B and 5C show the image input is partitioned into four small blocks (A), (B), (C) and (D) it is to be understood that the said image may be partitioned into more than four. Unlike typical contrast stretching, CLAHE operates on small data regions (tiles), rather than the entire image. Each tile's contrast is enhanced, so that the histogram of the output region approximately matches the specified histogram. This evens out the distribution of used grey values and thus makes hidden features of the image more visible. After undergoing image enhancing in the image enhancing using CLAHE stage (404), the vessels in both bright region (optic disc) and dark regions are evenly enhanced and can be distinguished as shown in FIGS.

6A and 6B which illustrates two fundus images one prior to being fed into an image enhancing stage (404) and the other subsequent to it. It will be seen that the one undergoing image enhancement through the said image enhancing using CLAHE stage (404) (FIG. 6B) has a well-contrasted image.

Referring now to FIG 7, there is shown a block diagram of the flow in one of the steps stated in FIG 4 called extracting blood vessel using morphological technique such as bottom hat to extract the enhanced retinal vasculature. The image from the image enhancing using CLAHE stage (404) goes through a morphological process, preferably bottom-hat (406) and a specified structuring element in order to perform higher contrast to the blood vessels. The said morphological process (406) comprises several sub-stages: filtering using bottom hat, filtering using averaging filter and enhancing contrast. The

Bottom-hat technique is used to isolate dark objects on light surroundings that are convex preferably using structuring elements of the linear type as the vessels are mainly linear in form. The size of the structuring element is preferably chosen to ensure that only the primary vessels including their branching and intersection points are extracted and the structuring elements are posed in different orientations using a rotating angle from 0 to 180 degrees. A Bottom-hat technique using the said linear structuring element allows vessels even in low local contrast regions to be extracted regardless of their sizes and directions. Background noise removal preferably using the averaging filter technique is then carried out to reduce background noise and non-vessel features at the background that are being enhanced as well while undergoing the Bottom-hat technique. Further enhancement of extracted vessels may then be carried out using normal contrast stretching technique in this blood vessel extraction using morphological process such as bottom hat stage (406).

It is observed that morphological transformation has certain limitations such as the inability to highlight bifurcation and intersection points effectively where the profiles are larger than the Structuring Element. It is observed that the intersection point mostly at primary vessels can only be weakly or partially extracted.

After extraction of the said vessels (406), images of the extracted vessels are proceeded to a blood vessel end-point detection stage (408) to determine the location of the extremities (or end-points) of the blood vessels. The said detection is done either by semi-automated or fully automated means to determine end-points based on selected extremities (or end-points) of the extracted capillaries to facilitate calculation, plotting and delineating of the area of FAZ (802) at the next stage. In semi-automated means human intervention is employed to select the extremities or terminal points of the capillary of the extracted vessel manually assisted by automatic zoom function. In fully automated means selection of the extremities or terminal points of the capillary are carried out using computerized techniques without any human intervention except for defining the centre of macula. With the detection of end-points, the FAZ area (802) can be plotted by connecting all the end-points of the blood vessels to delineate the FAZ area (802) for analysis thereafter. An illustration of the FAZ area (802) which is plotted and delineated using the semi-automated means is shown in FIG. 8.

Referring to FIG 9, there is shown a block diagram of the flow for another algorithm that can be implemented in the said processing system (304). The software program that is being used by the said processing system (304) can be Visual C++ or Matlab. After the image of retina is being captured by the retina imaging system (302) and transferred to the said processing system (304) through the image capturing device (310), said processing system (304) performs the following steps: i. performing pre-processing to said image (402); ii. enhancing image and extracting blood vessel using principal component analysis, (hereinafter referred to as "PCA") and fast independent component analysis (hereinafter referred to as

"FastICA") (902); iii. detecting blood vessel end-point (408); iv. performing analysis (410). Referring now to FIG 1OA and 1OB, there is shown block diagram of the flow in one of the steps stated in FIG 9 called enhancing image and extracting blood vessel using PCA and FastICA. FIG 1OA refers to the sub-stages to perform PCA while FIG 1OB refers to the sub-stages to perform FastICA. Alternative to the image enhancing using CLAHE and blood vessel extraction using morphological process such as bottom hat as in FIG 4, algorithm known as PCA and FastICA may be employed after the image pre-processing stage (402) in place of said image enhancing using CLAHE and blood vessel extraction using morphological process such as bottom hat. The enhancing images using PCA and FastICA involves the determination of retinal pigments namely haemoglobin, melanin and macular pigment from the fundus images based on the distribution of the retinal pigments. PCA is used to determine the principal value or the maximum energy of the image being analyzed. The sub-stages in performing PCA are: calculating covariance matrix of the retinal image in red, green and blue, calculating Eigen values and eigenvectors, arranging eigenvectors and multiplying arranged eigenvectors with data sets to create a PCA subspace.

Principal Components Analysis is a multivariate data analysis method. The eigenvectors are computed from the covariance matrix of the image. Arranging the eigenvectors based on its sorted eigenvalues will make the principal component of resulted image contained most of total variance of the data sets. Finally, the linear transformation between image and its eigenvectors will make the output image orthogonal and uncorrelated.

As for the sub-stages of FastICA, they are: reconfiguring matrix W and measuring for non gaussianity until maximum value is reached, and multiplying matrix W with said PCA subspace. Matrix W is estimated so that the non-gaussianity of the resulted image is optimum. The estimation is performed by optimizing a certain variable in the non-gaussianity measurement function. Finally, the transformation between PCA output and matrix W will make the component of the resulted image independent from each others.

The retinal pigments from macular images do not create artefacts in its enhanced state and has been found to perform better than CLAHE. The haemoglobin and macular pigment are used to reveal retinal blood vessels and the macular region for studies to be conducted and can be used for image enhancement in particular to visualise very fine vessels that are poorly contrasted without the need of injecting any contrasting agent. As vessels are the predominant and most reliable structures in the fundus images, reliable vessel extraction is a requirement for subsequent retinal image analysis. Vessel enhancement as achieved by CLAHE or PCA-FastICA described above increases the contrast of vessels to the background and this makes the task to extract vessel from the background less difficult.

Subsequently in the analysis stage (410), as per in FIGS 4 and 9, the FAZ area (802) is calculated and DR is analysed by using a DR analysis algorithm which can be developed using Visual C++, MATLAB or any appropriate programming tools for subsequent grading of DR and MR. Visual C++ or MATLAB provides programming logic that can realize the DR algorithm. The

DR algorithm is divided into modules. The modules are then attained using Visual C++ or MATLAB. The analysis stage (410) is operated by an automated computerized technique to calculate the FAZ area (802) and subsequently grade DR and DM based on the relative area of the FAZ (802). From the FAZ (802) area distribution, it is found that there is overlapping of

FAZ (802) areas between two stages of the disease. The resultant overlapping area ranges are highlighted to indicate progression of the disease to another stage. The area of FAZ (802) that overlaps the later stage is used as the upper bound of the current stage while the minimum FAZ area (802) of the later stage is used as the lower bound of the later stage. The FAZ area (802) ranges that show progressions of the disease stage by stage are analyzed. The lower bounds may indicate the maximum FAZ area (802) for the previous stage and thus will be the indicator of progression to the next stage while the upper bounds may indicate the maximum FAZ area (802) for progression to the next stage. Area within the progression range shows that the disease is in high risk of progressing to the next stage. Using the above apparatus, the severity of Diabetic Retinopathy and Diabetic Maculopathy can be easily graded. Using the above apparatus preferably five ranges of FAZ area (802) for Diabetic Retinopathy grading are obtained namely normal, mild non-proliferative diabetic retinopathy, moderate non-proliferative diabetic retinopathy, severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy.

The grading of Diabetic Retinopathy and its progressing stages based on the size of FAZ (802) can be graded in the following ranges:

(a) Normal range

(b) Progression from normal to NPDR range

(c) NPDR range

(d) Progression from NPDR to severe NPDR/ PDR range

(e) Severe NPDR / PDR range

While the grading of Diabetic Retinopathy by the said apparatus has been described and illustrated in some detail as being in five ranges as stated above, ranges below or above five may also be applicable as what is advantageous in accordance with the present invention is the fact that the present invention allows accurate and reliable ascertainment of the FAZ area (802) to aid grading of Diabetic Retinopathy and its progressing stages.

The said apparatus enabling determination of the FAZ area (802) can also be used to grade Diabetic Maculopathy (DM). For severity grading of

Diabetic Maculopathy, the maximum value for FAZ area (802) for each stage is investigated and used to mark the border of progression of the disease.

Using the above apparatus, preferably three ranges of FAZ area (802) for

Diabetic Maculopathy grading and its progressing stages are obtained namely Normal, Diabetic Maculopathy (Observable) and Severe Diabetic

Maculopathy (Referable).

While the grading of Diabetic Maculopathy has been described and illustrated in some detail as being in three ranges as stated above, ranges below or above three may also be applicable as what is advantageous in accordance with the present invention is the fact that the present invention allows accurate and reliable ascertainment of the area of FAZ (802) to aid grading of Diabetic Maculopathy and its progressing stages.

The said invention has disclosed an apparatus for determination of the FAZ area (802) for grading the severity of Diabetic Retinopathy or Diabetic Maculopathy based on the digital map of retinal vasculature obtained from extraction of retinal vasculature to enable the vessel ends and pathologies surrounding FAZ area (802) to be derived for reliable determination of the FAZ area (802).

Although the algorithms utilized in the image enhancing technique has been described and illustrated in some detail as being CLAHE and FastICA, other image enhancing algorithms may also be applicable as what is advantageous in accordance with the present invention is the fact that the image enhancing technique specifically caters for low and varying contrast image to enhance the retinal vessels from its background.

Although the said apparatus been described and illustrated as for use in diagnosing and grading Diabetic Retinopathy and Diabetic Maculopathy, the extracted vessels can also be used to assist in diagnosis of hypertension based on the turtuosity of the retinal vessels.

FIGS HA - HK show an example of the layout of the user interface (306) system as a platform for users to key in appropriate information into the processing system (304) to be analyzed, for the user to control the operations of the processing system (304) and for the user to observe the outcome of the analysis. FIG HA shows the opening view of the user interface. FIG HB shows the area in the user interface (306) whereby information of the person being examined such as personal particulars can be entered to the processing system (304). FIG HC shows an image of the retina being displayed at the user interface after being captured by the retina imaging system (302). FIGS HD-HG show a sample of the user interface if the user chooses to perform the image enhancement using CLAHE. FIG HD shows a button in the user interface enabling the user to enhance the said retina image using CLAHE.

FIG HE shows a button in the user interface enabling the user to automatically analyze and perform grading towards the DR of the patient through automatic detection and determination of blood vessel end points as shown in FIG HF. FIG HG shows the outcome of the analysis, whereby the range of the DR is shown on the user interface. FIGS HH-HK show a sample of the user interface if the user chooses to perform the image enhancement and blood vessel extraction using PCA and FastICA. FIG HH shows a panel in the user interface allowing the user to navigate the appropriate location and size of the FastICA region to be analyzed. FIG 111 shows the button at the user interface allowing the user to enhance the selected image from FIG HH using PCA and FastICA. FIG HJ shows the user interface whereby the user can manually place the end points of the blood vessel for the analysis. The end point of the blood vessels can also be determined automatically. FIG HK shows the outcome of the analysis, whereby the range of the DR is shown on the user interface after analysis is being done. Although the user interface is being explained as per in FIGS HA - HK, the layout of the user interface, amount of information given by the user to the user interface and the amount of information provided to the user by the user interface can vary depending on the individual needs, as long as the said information is sufficient for the user to know the grading of DR and DM of the patient using the said apparatus.

It will be understood by those skilled in the art that changes and modifications may be made to the invention without departing from the spirit and scope of the invention. Therefore it is intended that the foregoing description merely for illustrative purposes and not intended to limit the spirit and scope of the invention in any way but only by the spirit and scope of the appended claim.

Claims

WHAT IS CLAIMED IS:

1. An apparatus for analyzing the retina for ocular manifested diseases comprising,

at least one any acceptable processing means (304),

at least one user interface system (306),

characterized in that

said any acceptable processing means (304) is specifically provided with a predetermined algorithm using any acceptable programming software to cater for low and varying contrast image to enhance the retinal vessels (104) from its background prior to extraction of the enhanced retinal vessels for ascertainment of the area of the foveal avascular zone (802);

further characterized in that

2. An apparatus for analyzing the retina for ocular manifested diseases as in Claim 1 wherein the said apparatus is used to grade severity of diabetic retinopathy.

3. An apparatus for analyzing the retina for ocular manifested diseases as in Claim 1 wherein the said apparatus is used to grade severity of diabetic maculopathy.

4. An apparatus for analyzing the retina for ocular manifested diseases as in Claim 1 or 2 wherein the said apparatus allows grading of diabetic retinopathy and its progressing stages based on the relative area of the foveal avascular zone (22) into the following ranges:-

Normal

Mild non-proliferative diabetic retinopathy Moderate non-proliferative diabetic retinopathy Severe non-proliferative diabetic retinopathy Proliferative diabetic retinopathy

5. An apparatus for analyzing the retina for ocular manifested diseases as in Claim 1 or 3 wherein the said apparatus allows grading of diabetic maculopathy and its progressing stages based on the relative area of the foveal avascular zone (802) into the following ranges: -

Normal

DM (Observable) Severe DM (Referable)

6. An apparatus for analyzing the retina for ocular manifested diseases as in Claim 1 wherein ascertainment of the foveal avascular zone (802) is semi- automated or fully automated.

7. An apparatus for analyzing the retina for ocular manifested diseases as in Claim 1 wherein said predetermined algorithm comprises the following steps:

i. performing pre-processing on retinal image captured from said retina imaging system (402);

ii. enhancing said image (404, 902);

iii. extracting blood vessels from said image (406, 902);

iv. detecting end point of blood vessels to ascertain area of foveal avascular zone (FAZ) (408);

v. analyzing FAZ area to determine severity of diabetic retinopathy (410).

8. An apparatus for analyzing the retina for ocular manifested diseases as in any one of the preceding claims wherein the said image enhancing stage (404) employs the algorithm known as Contrast Limited Adaptive Histogram Equalization (CLAHE).

9. An apparatus for analyzing the retina for ocular manifested diseases as in any one of the preceding claims wherein the said image enhancing stage and blood vessels extraction from said image (902) employs the algorithm known as principal component analysis (PCA) and Fast Independent Component Analysis (FastICA).

10. An apparatus for analyzing the retina for ocular manifested diseases as in any one of the preceding claims wherein the said blood vessel extraction stage (406) employs a technique to isolate dark objects on light surroundings that are convex known as bottom-hat technique in extracting the enhanced retinal vessels.

11. An apparatus for analyzing the retina for ocular manifested diseases as in any one of the preceding claims wherein the structuring element used during blood vessel extraction stage (406) is a linear type.

12. An apparatus for analyzing the retina for ocular manifested diseases as claimed in any one of the preceding claims which is capable of diagnosing hypertension.

13. An apparatus for analyzing the retina for ocular manifested diseases as claimed in any one of the preceding claims wherein said any acceptable programming software is preferably Visual C++, Matlab or any equivalent programming software.

14. An apparatus for analyzing the retina for ocular manifested diseases as claimed in any one of the preceding claims wherein said user interface system (306) comprises at least one displaying means and at least one input means to allow the user to provide information to said any acceptable processing means (304) and portray results of said disease grading.

15. An apparatus for analyzing the retina for ocular manifested diseases as claimed in any one of the preceding claims wherein said retina imaging device is a fundus camera.

16. An apparatus for analyzing the retina for ocular manifested diseases as claimed in any one of the preceding claims wherein said image capturing device is a digital camera attached to said fundus camera to capture images of retina portrayed by said fundus camera.