Muhammad Usman Akram

Core point detection using improved segmentation and orientation

Core point detection is very important in fingerprint classification and matching process. Usually fingerprint images have noisy background and the local orientation field also changes very rapidly in the singular point area. It is difficult to locate the singular point precisely. In this paper, we present a new algorithm for optimal core point detection using improved segmentation and orientation. In our technique detects core point accurately by extracting best region of interest(ROI) from image and using fine orientation field estimation. We present a modified technique for extracting ROI and fine orientation field. The distinct feature of our technique is that it gives high detection percentage of core point even in case of low quality fingerprint images. The proposed algorithm is applied on FVC2004 database. Results of experiments demonstrate improved performance for detecting core point.

Automated segmentation of subretinal layers for the detection of macular edema

Macular edema (ME) is considered as one of the major indications of proliferative diabetic retinopathy and it is commonly caused due to diabetes. ME causes retinal swelling due to the accumulation of protein deposits within subretinal layers. Optical coherence tomography (OCT) imaging provides an early detection of ME by showing the cross-sectional view of macular pathology. Many researchers have worked on automated identification of macular edema from fundus images, but this paper proposes a fully automated method for extracting and analyzing subretinal layers from OCT images using coherent tensors. These subretinal layers are then used to predict ME from candidate images using a support vector machine (SVM) classifier. A total of 71 OCT images of 64 patients are collected locally in which 15 persons have ME and 49 persons are healthy. Our proposed system has an overall accuracy of 97.78% in correctly classifying ME patients and healthy persons. We have also tested our proposed implementation on spectral domain OCT (SD-OCT) images of the Duke dataset consisting of 109 images from 10 patients and it correctly classified all healthy and ME images in the dataset.

Glaucoma detection using novel optic disc localization, hybrid feature set and classification techniques

Glaucoma is a chronic and irreversible neuro-degenerative disease in which the neuro-retinal nerve that connects the eye to the brain (optic nerve) is progressively damaged and patients suffer from vision loss and blindness. The timely detection and treatment of glaucoma is very crucial to save patient’s vision. Computer aided diagnostic systems are used for automated detection of glaucoma that calculate cup to disc ratio from colored retinal images. In this article, we present a novel method for early and accurate detection of glaucoma. The proposed system consists of preprocessing, optic disc segmentation, extraction of features from optic disc region of interest and classification for detection of glaucoma. The main novelty of the proposed method lies in the formation of a feature vector which consists of spatial and spectral features along with cup to disc ratio, rim to disc ratio and modeling of a novel mediods based classier for accurate detection of glaucoma. The performance of the proposed system is tested using publicly available fundus image databases along with one locally gathered database. Experimental results using a variety of publicly available and local databases demonstrate the superiority of the proposed approach as compared to the competitors.

An automated system for the grading of diabetic maculopathy in fundus images

Computer aided diagnosis systems are very popular now days as they assist doctors in early detection of the disease. Diabetic maculopathy is one such disease which affects the retina of the diabetic patients. It affects the central vision of the person and causes blindness in severe cases. In this paper, an automated system for the grading of diabetic maculopathy has been developed, that will assist the ophthalmologists in early detection of the disease. Here, we propose a novel computerized method for the grading of diabetic maculopathy in fundus images. Our proposed system comprises of preprocessing of retinal image followed by macula and exudate regions detection. This is followed by feature extractor module for the formulation of feature set. SVM classifier is then used to grade the diabetic maculopathy. The publicly available fundus image database MESSIDOR has been used for the validation of our algorithm. The results of our proposed system have been compared with other methods in the literature in terms of sensitivity and specificity. Our system gives higher values of sensitivity and specificity as compared to others on the same database.

Decision Support System for Detection of Papilledema through Fundus Retinal Images

A condition in which the optic nerve inside the eye is swelled due to increased intracranial pressure is known as papilledema. The abnormalities due to papilledema such as opacification of Retinal Nerve Fiber Layer (RNFL), dilated optic disc capillaries, blurred disc margins, absence of venous pulsations, elevation of optic disc, obscuration of optic disc vessels, dilation of optic disc veins, optic disc splinter hemorrhages, cotton wool spots and hard exudates may result in complete vision loss. The ophthalmologists detect papilledema by means of an ophthalmoscope, Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and ultrasound. Rapid development of computer aided diagnostic systems has revolutionized the world. There is a need to develop such type of system that automatically detects the papilledema. In this paper, an automated system is presented that detects and grades the papilledema through analysis of fundus retinal images. The proposed system extracts 23 features from which six textural features are extracted from Gray-Level Co-occurrence Matrix (GLCM), eight features from optic disc margin obscuration, three color based features and seven vascular features are extracted. A feature vector consisting of these features is used for classification of normal and papilledema images using Support Vector Machine (SVM) with Radial Basis Function (RBF) kernel. The variations in retinal blood vessels, color properties, texture deviation of optic disc and its peripapillary region, and fluctuation of obscured disc margin are effectively identified and used by the proposed system for the detection and grading of papilledema. A dataset of 160 fundus retinal images is used which is taken from publicly available STARE database and local dataset collected from Armed Forces Institute of Ophthalmology (AFIO) Pakistan. The proposed system shows an average accuracy of 92.86% for classification of papilledema and normal images. It also shows an average accuracy of 97.85% for classification of already classified papilledema images into mild and severe papilledema. The proposed system is a novel step towards automated detection and grading of papilledema. The results showed that the technique is reliable and can be used as clinical decision support system.

Retinal images: Noise segmentation

In automated diagnosis of diabetic retinopathy, retinal images are used. The retinal images of poor quality need to be enhanced before the extraction of features and abnormalities. Segmentation of retinal images is essential for this purpose. The segmentation is employed to smooth and strengthen images by separating the noisy area from the overall image thus resulting in retinal image enhancement and less processing time. In this paper, we present a novel automated approach for segmentation of colored retinal images, which involves two steps. In the first step, we create binary noise segmentation mask to segment the retinal image. Second step creates final segmentation mask by applying morphological techniques. We used standard retinal image databases Diaretdb0 and Diaretdb1 to test the validation of our segmentation technique. Experimental results indicate our approach is effective and can get higher segmentation accuracy.

A novel multiprocessor architecture for k-means clustering algorithm based on network-on-chip

The k-means clustering is one of the widely used algorithms in Data Mining and Machine Learning domains due to the simplicity, efficiency and scalability involved. The algorithm allocates N data-points or samples to k-clusters employing the minimum distances from respective cluster centroids. Distance calculation is intrinsically a computationally intensive task which is usually accelerated by using specific hardware platforms like Field Programmable Gate Arrays (FPGAs) and Graphic Processing Unit (GPUs) etc. Hardware implementations absolve k-means from these exhaustive computations by using the inherent parallelism of the k-means clustering technique. In this paper we propose a Multi-Processor based sequentially unfolded architecture for k-means clustering using N-tiles in a collaborative working environment. The tiles work independently in parallel and largely interchange data at the end of iteration. In proposed framework the exchange of data between titles is carried out using a Network-on-Chip (NoC) inter-connect to elevate communication bottleneck caused by concurrent working of tiles. The modularity of the proposed model permits scalability with respect to the number of working tiles. The performance evaluation of proposed architecture is done using Speed, Area and average Throughput.

Classifying Normal Sinus Rhythm and Cardiac Arrhythmias in ECG Signals Using Statistical Features in Temporal Domain

Any morphological abnormality or atypical group of conditions in a cardiac rhythm indicates a typical class of arrhythmias. ECG plays a vital role in diagnosis of many such cardiac disorders. Some arrhythmias including ventricular fibrillation and premature ventricular contraction can be fatal if not dealt on time. Clinical analysis of ECGs by physicians may result into an inaccurate as well as a time-consuming analysis of a critically serious arrhythmia patient mostly involving measuring the ECG statistics from calipers. Considering the importance of morphological shapes and statistics of ECG signals in arrhythmia diagnosis, these features are input to a dedicated system which act as key markers to categorize various arrhythmias automatically. This paper highlights the development of an algorithm for classifying 15 different cardiac arrhythmias using a novel statistical feature set of ECG signals in time domain. Rhythm annotations from the bench mark MIT-BIH Cardiac Arrhythmia database have been used to organize the data as rhythms. The proposed method has been simulated and tested in MATLAB and results have been discussed in detail.

Improved automated detection of glaucoma from fundus image using hybrid structural and textural features

Glaucoma is a group of eye disorders that damage the optic nerve. Considering a single eye condition for the diagnosis of glaucoma has failed to detect all glaucoma cases accurately. A reliable computer-aided diagnosis system is proposed based on a novel combination of hybrid structural and textural features. The system improves the decision-making process after analysing a variety of glaucoma conditions. It consists of two main modules hybrid structural feature-set (HSF) and hybrid texture feature-set (HTF). HSF module can classify a sample using support vector machine (SVM) from different structural glaucoma condition and the HTF module analyses the sample founded on various texture and intensity-based features and again using SVM makes a decision. In the case of any conflict in the results of both modules, a suspected class is introduced. A novel algorithm to compute the super-pixels has also been proposed to detect the damaged cup. This feature alone outperformed the current state-of-the-art methods with 94% sensitivity. Cup-to-disc ratio calculation method for cup and disc segmentation, involving two different channels has been introduced increasing the overall accuracy. The proposed system has given exceptional results with 100% accuracy for glaucoma referral.

Parallel architecture for implementation of frequent itemset mining using FP-growth

Frequent itemset mining is a fundamental step in analysis of big data where correlation among the raw data in deemed necessary. In modern era the amount of data available for processing has grown exponentially, making it a stepper task for mining algorithms to provide solution in a timely manner. The software implementations are normally not efficient in handling such datasets thus focus on parallel architecture seems imminent. In this paper we propose a Multi-Processor based sequentially unfolded architecture for implementation of FP-Growth algorithm. The proposed framework exploits the inherent parallelism available in the FP-Growth algorithm such that N-processing entities (PEs) can work in a collaborative environment. The processing entities work in an independent manner in parallel and largely interchange data at the close of each iteration. The overall architecture is modular which permits scalability of the design with regards to the number of parallel processing entities. The performance of the framework is evaluated using benchmark datasets and their results show a linear increase in the speedup of our proposed framework with increase in PEs.