Search

In this paper, we present a multiple ocular diseases detection scheme based on joint sparse multi-task learning. Glaucoma, Pathological Myopia (PM), and Age-related Macular Degeneration (AMD) are three major causes of vision impairment and blindness worldwide. The proposed joint sparse multitask learning framework aims to reconstruct a test fundus image with multiple features from as few training subjects as possible. The linear version of this problem could be casted into a multi-task joint covariate selection model, which can be very efficiently optimized via kernelizable accelerated proximal gradient method. Extensive experiments are conducted in order to validate the proposed framework on the SiMES dataset. From the Area Under Curve (AUC) results in multiple ocular diseases classification, our method is shown to outperform the state-of-the-art algorithms.

In recent years, there has been increasing interest in the use of automatic computer-based systems for the detection of eye diseases such as glaucoma, age-related macular degeneration and diabetic retinopathy. However, in practice, retinal image quality is a big concern as automatic systems without consideration of degraded image quality will likely generate unreliable results. In this paper, an automatic retinal image quality assessment system (ARIES) is introduced to assess both image quality of the whole image and focal regions of interest. ARIES achieves 99.54% accuracy in distinguishing fundus images from other types of images through a retinal image identification step in a dataset of 35342 images. The system employs high level image quality measures (HIQM) to perform image quality assessment, and achieves areas under curve (AUCs) of 0.958 and 0.987 for whole image and optic disk region respectively in a testing dataset of 370 images. ARIES acts as a form of automatic quality control which ensures good quality images are used for processing, and can also be used to alert operators of poor quality images at the time of acquisition.

Optic disc segmentation from retinal fundus image is a fundamental but important step in many applications such as automated glaucoma diagnosis. Very often, one method might work well on many images but fail on some other images and it is difficult to have a single method or model to cover all scenarios. Therefore, it is important to combine results from several methods to minimize the risk of failure. For this purpose, this paper computes confidence scores for three methods and combine their results for an optimal one. The experimental results show that the combined result from three methods is better than the results by any individual method. It reduces the mean overlapping error by 7.4% relatively compared with best individual method. Simultaneously, the number of failed cases with large overlapping errors is also greatly reduced. This is important to enhance the clinical deployment of the automated disc segmentation.

We introduce the experiences of the Singapore ocular imaging team, iMED, in integrating image processing and computer-aided diagnosis research with clinical practice and knowledge, towards the development of ocular image processing technologies for clinical usage with potential impact. In this paper, we outline key areas of research with their corresponding image modalities, as well as providing a systematic introduction of the datasets used for validation.

Optic disc segmentation in retinal fundus image is important in ocular image analysis and computer aided diagnosis. Because of the presence of peripapillary atrophy which affects the deformation, it is important to have a good initialization in deformable model based optic disc segmentation. In this paper, a superpixel classification based method is proposed for the initialization. It uses histogram of superpixels from the contrast enhanced image as features. In the training, bootstrapping is adopted to handle the unbalanced cluster issue due to the presence of peripapillary atrophy. A self-assessment reliability score is computed to evaluate the quality of the initialization and the segmentation. The proposed method has been tested in a database of 650 images with optic disc boundaries marked by trained professionals manually. The experimental results show an mean overlapping error of 10.0% and standard deviation of 7.5% in the best scenario. The results also show an increase in overlapping error as the reliability score reduces, which justifies the effectiveness of the self-assessment. The method can be used for optic disc boundary initialization and segmentation in computer aided diagnosis system and the self-assessment can be used as an indicator of cases with large errors and thus enhance the usage of the automatic segmentation.

The optic cup segmentation is critical for automated cup-to-disk ratio measurement, and hence computer-aided diagnosis of glaucoma. In this paper, we propose a novel sector-based method for optic cup segmentation. The method comprises two parts: intensity-based cup segmentation with shape constraints and blood vessel-based refinement. The initial estimation of the cup is obtained by applying a statistical deformable model on the vessel free image. At the same time, blood vessels within the optic disk are extracted, after which vessel bendings and vessel boundaries in the nasal side are located. Subsequently, these key points in the blood vessels are used to fine tune the cup. The algorithm is evaluated on 650 fundus images from the ORIGA(-light) database. Experimental results show that the Dice coefficient for the optic cup segmentation can be as high as 0.83, which outperforms other existing methods. The results demonstrate good potential for the proposed method to be used in automated optic cup segmentation and glaucoma diagnosis.

We present a regional propagation approach based on retinal structure priors to localize the optic cup in 2D fundus images, which is the primary image component clinically used for identifying glaucoma. This method provides three major contributions. First, it proposes processing of the fundus images at the superpixel level, which leads to more descriptive and effective features than those employed by pixel based techniques, without additional computational cost. Second, the proposed approach does not need manually labeled training samples, but uses the structural priors on relative cup and disc positions. Third, a refinement scheme that utilizes local context information is adopted to further improve the accuracy. Tested on the ORIGA-light clinical dataset, which comprises of 325 images from a population-based study, the proposed method achieves a 34.9% non-overlap ratio with manually-labeled ground-truth and a 0.104 absolute cup-to-disc ratio (CDR) error. This level of accuracy is much higher than the state-of-the-art pixel based techniques, with a comparable or even less computational cost.

The macula is the part of the eye responsible for central high acuity vision. Detection of the macula is an important task in retinal image processing as a landmark for subsequent disease assessment, such as for age-related macula degeneration. In this paper, we have presented an approach to automatically determine the macula centre in retinal fundus images. First contextual information on the image is combined with a statistical model to obtain an approximate macula region of interest localization. Subsequently, we propose the use of a seeded mode tracking technique to locate the macula centre. The proposed approach is tested on a large dataset composed of 482 normal images and 162 glaucoma images from the ORIGA database and an additional 96 AMD images. The results show a ROI detection of 97.5%, and 90.5% correct detection of the macula within 1/3DD from a manual reference, which outperforms other current methods. The results are promising for the use of the proposed approach to locate the macula for the detection of macula diseases from retinal images.

Optic disc segmentation from retinal fundus image is a fundamental but important step for automatic glaucoma diagnosis. In this paper, an optic disc segmentation method is proposed based on peripapillary atrophy elimination. The elimination is done through edge filtering, constraint elliptical Hough transform and peripapillary atrophy detection. With the elimination, edges that are likely from non-disc structures especially peripapillary atrophy are excluded to make the segmentation more accurate. The proposed method has been tested in a database of 650 images with disc boundaries marked by trained professionals manually. The experimental results by the proposed method show average m(1), m(2) and m(VD) of 10.0%, 7.4% and 4.9% respectively. It can be used to compute cup to disc ratio as well as other features for application in automatic glaucoma diagnosis systems.

Min-Redundancy Max-Relevance (mRMR) is a feature selection methodology based on information theory. We explore the mRMR principle for automatic glaucoma diagnosis. Optimal candidate feature sets are acquired from a composition of clinical screening data and retinal fundus image data. An mRMR optimized classifier is further trained using the candidate feature sets to find the optimized classifier. We tested the proposed methodology on eye records of 650 subjects collected from Singapore Eye Research Institute. The experimental results demonstrate that the new classifier is much compact by using less than ¼ of the initial feature set. The ranked feature set also enables the clinicians to better access the diagnostic process of the algorithm. The work is a further step towards the advancement of the automatic glaucoma diagnosis.

The optic nerve head (optic disc) plays an important role in the diagnosis of retinal diseases. Automatic localization and segmentation of the optic disc is critical towards a good computer-aided diagnosis (CAD) system. In this paper, we propose a method that combines edge detection, the Circular Hough Transform and a statistical deformable model to detect the optic disc from retinal fundus images. The algorithm was evaluated against a data set of 325 digital color fundus images, which includes both normal images and images with various pathologies. The result shows that the average error in area overlap is 11.3% and the average absolute area error is 10.8%, which outperforms existing methods. The result indicates a high correlation with ground truth segmentation and thus demonstrates a good potential for this system to be integrated with other retinal CAD systems.