Arslan Shaukat

Automated Detection and Grading of Diabetic Maculopathy in Digital Retinal Images

Diabetic maculopathy is one of the retinal abnormalities in which a diabetic patient suffers from severe vision loss due to the affected macula. It affects the central vision of the person and causes blindness in severe cases. In this article, we propose an automated medical system for the grading of diabetic maculopathy that will assist the ophthalmologists in early detection of the disease. The proposed system extracts the macula from digital retinal image using the vascular structure and optic disc location. It creates a binary map for possible exudate regions using filter banks and formulates a detailed feature vector for all regions. The system uses a Gaussian Mixture Model-based classifier to the retinal image in different stages of maculopathy by using the macula coordinates and exudate feature set. The evaluation of proposed system is performed by using publicly available standard retinal image databases. The results of our system have been compared with other methods in the literature in terms of sensitivity, specificity, positive predictive value and accuracy. Our system gives higher values as compared to others on the same databases which makes it suitable for an automated medical system for grading of diabetic maculopathy.

Hybrid Features and Mediods Classification based Robust Segmentation of Blood Vessels

Retinal blood vessels are the source to provide oxygen and nutrition to retina and any change in the normal structure may lead to different retinal abnormalities. Automated detection of vascular structure is very important while designing a computer aided diagnostic system for retinal diseases. Most popular methods for vessel segmentation are based on matched filters and Gabor wavelets which give good response against blood vessels. One major drawback in these techniques is that they also give strong response for lesion (exudates, hemorrhages) boundaries which give rise to false vessels. These false vessels may lead to incorrect detection of vascular changes. In this paper, we propose a new hybrid feature set along with new classification technique for accurate detection of blood vessels. The main motivation is to lower the false positives especially from retinal images with severe disease level. A novel region based hybrid feature set is presented for proper discrimination between true and false vessels. A new modified m-mediods based classification is also presented which uses most discriminating features to categorize vessel regions into true and false vessels. The evaluation of proposed system is done thoroughly on publicly available databases along with a locally gathered database with images of advanced level of retinal diseases. The results demonstrate the validity of the proposed system as compared to existing state of the art techniques.

A computer aided system for grading of maculopathy

In medical imaging, digital images are analyzed to develop computer aided diagnostic (CAD) systems using state of the art image processing and pattern recognition techniques. Diabetic maculopathy is one of the retinal abnormalities in which diabetic patient suffers from severe vision loss due to affected macula. In this paper, we propose an automated system for the grading of diabetic maculopathy to assist the ophthalmologists in early detection of the disease. We present a three stage system consisting of macula detection, exudate extraction and grading of maculopathy. First stage uses optic disc and blood vessels to extract macula from retinal image. Exudate extraction stage extracts all possible exudates from retina using filter bank and support vector machines. Finally, the system grades the input image in different stages of maculopathy by using the macular coordinates and exudate feature set. The evaluation of proposed system is performed by using publicly available standard retinal image databases.

Automated Plant Disease Analysis (APDA): Performance Comparison of Machine Learning Techniques

Plant disease analysis is one of the critical tasks in the field of agriculture. Automatic identification and classification of plant diseases can be supportive to agriculture yield maximization. In this paper we compare performance of several Machine Learning techniques for identifying and classifying plant disease patterns from leaf images. A three-phase framework has been implemented for this purpose. First, image segmentation is performed to identify the diseased regions. Then, features are extracted from segmented regions using standard feature extraction techniques. These features are then used for classification into disease type. Experimental results indicate that our proposed technique is significantly better than other techniques used for Plant Disease Identification and Support Vector Machines outperforms other techniques for classification of diseases.

Person identification using vascular and non-vascular retinal features

Novel methods for personal identification using retinal images.Vascular based method involves the use of vessel properties of retinal images with improved vessel segmentation algorithm by catering pathological lesions.Non-vascular based method uses novel structural features structure to perform person identification. Display Omitted Retina recognition is the most stable and reliable biometric system due to its stability, uniqueness and non-replicable nature of vascular pattern. On the other hand, the complexity of vascular pattern in diseased retina makes the extraction of blood vessels very hard, which majorally effects the recognition rate. The main aim of this paper is to design a robust retinal recognition system with reduced computational complexity and to explore novel retinal features. This paper presents two different approaches for retinal recognition; one is vascular-based feature extraction with an improved vessel segmentation algorithm and second is non-vascular based feature extraction. Vascular-based method uses vessel properties of retinal images and aims to improve the efficiency of retinal recognition system. Whereas, non-vascular based method intends to analyze non-vessel properties of retinal images in order to reduce time complexity. The proposed system is assessed on two local and three public databases.

Scale and Rotation Invariant Texture Classification Using Covariate Shift Methodology

In this letter, we propose to tackle rotation and scale variance in texture classification at the machine learning level. This is achieved by using image descriptors that interpret these variations as shifts in the feature vector. We model these variations as a covariate shift in the data. This shift is then reduced by minimising the Kullback-Leibler divergence between the true and estimated distributions using importance weights (IW). These IWs are used in support vector machines (SVMs) to formulate the IW-SVMs. The experimental results show that IW-SVMs exhibit good invariance characteristics and outperform other state-of-the-art classification methods. The proposed methodology gives a generic solution that can be applied to any texture descriptor that models the transformations as a shift in the feature vector.

A Gaussian mixture model based system for detection of macula in fundus images

Digital fundus imaging is used to diagnose various eye diseases like diabetic retinopathy, diabetic maculopathy and age related macular degeneration. Macula is the main central part of retina which is responsible for sharp vision and any changes in macula cause severe effects on vision. In this paper, we propose a novel method for automated detection of macula from digital fundus images. The proposed system performs preprocessing, optic disc detection and blood vessel segmentation prior to macula detection. In macula detection, it formulates a feature vector and uses Gaussian Mixture Model for detection of macular region. We evaluate the proposed technique using publicly available fundus image database MESSIDOR. The results show the validity of proposed system and are found to be competitive with previous results in the literature.

Emotional state recognition from speech via soft-competition on different acoustic representations

This paper presents our investigations on automatic emotional state recognition from speech signals using ensemble based methods based on different acoustic representations/feature measures. In our work, we employ various types of acoustic feature measures where none of the feature measures is optimal for emotional state classification. It is observed that different feature measures may be complementary and used simultaneously to yield a robust classification performance. Therefore, we employ a probabilistic method of combining classifiers based on different feature measures. The combination method that uses different feature measures simultaneously yields high recognition rates on various emotional speech corpora for both full feature set and language-independent feature subset. The ensemble method also outperforms a composite-feature representation and two other methods reported in literature. In addition, the classification accuracies achieved by our combination method are competitive with those mentioned in literature for different emotional speech corpora.

ECG Based Biometric Identification for Population with Normal and Cardiac Anomalies Using Hybrid HRV and DWT Features

Electrocardiograms (ECG) emerged as a novel biometric identification system in the past decade which yields high level of uniqueness and permanence. Moreover ECG provides inherent characteristic of liveness of a person, so it can furnish a superior solution as compared to other biometric techniques. This research provides with the complete systematic approach for ECG based person identification in various cardiac conditions and consists of ECG preprocessing, feature extraction, feature reduction and classifier performance. Segmentation of ECG involve R-peak detection, however system is independent of fiducial detection and does not require any extensive computational complexity. Feature extraction involve fusion of discrete wavelet transform (DWT) of cardiac cycle and heart rate variability (HRV) based features. Feature reduction is performed with best first search and classification is performed by using Random Forests. System is tested on three publicly available databases like MIT-BIH/Arrhythmia (MITDB), MIT-BIH/Normal Sinus Rhythm (NSRDB) and ECGID database (ECG-IDDB) including all subjects. HRV effects are removed from MITDB to confront with cardiac disorders that cause problems in identification and accuracy of 95.85% was achieved with false acceptance rate (FAR) of 4.15% and false rejection rate (FRR) of 0.1%. System is also tested on normal population based databases and accuracy of 100% is achieved using NSRDB database and 83.88% for a challenging ECG-ID database.

Daily sound recognition for elderly people using ensemble methods

This paper presents our investigations on automatic daily sound recognition using ensemble methods. Two benchmark datasets RWCP-DB and Sound Dataset are utilized for this purpose. A set of acoustic features for daily sound recognition is identified and used. First, sound classification is carried out using individual classifiers on both datasets. As the classification accuracy comes out lower with base classifiers as compared to the results reported in literature, ensemble methods are then employed for classification task. The ensemble methods prove to be effective and robust in recognizing daily sounds as they yield high recognition rates. The classification accuracies achieved by our proposed setup of ensemble methods are higher than those mentioned in literature for the two daily sound datasets.