No abstract

<span>Segmentation of blood vessels (BVs) from retinal image is one of the important steps in developing a computer-assisted retinal diagnosis system and has been widely researched especially for implementing automatic BV segmentation methods. This paper proposes an improvement to an existing retinal BV (RBV) segmentation method by combining the trainable B-COSFIRE filter with adaptive thresholding methods. The proposed method can automatically configure its selectivity given a prototype pattern to be detected. Its segmentation performance is comparable to many published methods with the advantage of robustness against noise on retinal background. Instead of using grid search to find the optimal threshold value for a whole dataset, adaptive thresholding (AT) is used to determine the threshold for each retinal image. Two AT methods investigated in this study were ISODATA and Otsu’s method. The proposed method was validated using 40 images from two benchmark datasets for retinal BV segmentation validation, namely DRIVE and STARE. The validation results indicated that the segmentation performance of the proposed unsupervised method is comparable to the original B-COSFIRE method and other published methods, without requiring the availability of ground truth data for new dataset. The Sensitivity and Specificity values achieved for DRIVE and STARE are 0.7818, 0.9688, 0.7957 and 0.9648, respectively.</span>

Ocular imaging has developed rapidly and plays a critical role in clinical care and ocular disease management. Development of image processing technologies pertinent to ocular diseases has paved the way for automated diagnostic systems including detection techniques using deep learning (DL) approaches. The prevalence of an abnormal tissue layer in the conjunctiva, known as pterygium eye disease, is increasing due to lack of awareness. Despite the non-cancerous/benign nature of pterygium, a clinical diagnosis from an ophthalmologist is still required to prevent the pterygium tissues from extending into the pupil, which would result in blurred vision. However, current diagnostic methods are mostly dependent on human expertise. Automated detection can potentially serve as an assistive method to reduce diagnosis time by applying a DL approach. Considering the lack of comprehensive research work on pterygium detection using DL, we propose a new architecture consisting of an improved CNN-based trained network named VggNet16-wbn that is derived from VggNet16, a pre-trained CNN algorithm. This paper presents an overview of the DL as a core approach to the transfer learning (TL) concept, as well as current efforts towards automated ocular detection approaches. A new architecture of a CNN-based trained network was proposed based on a network assessment from six CNN pre-trained networks to detect pterygium. This work consists of two main modules, namely, data acquisition and DCNN classification. The proposed trained network, VggNet16-wbn, shows the best performance with 99.22% accuracy, 98.45% sensitivity, and a perfect score on specificity and area under the curve metrics. This work has high potential for creating a pterygium screening system that can be used as a baseline for fully automated detection using a DL approach.

No abstract

For timely diagnosis of retinal disease, routine retinal monitoring of people with high risk should be put in place. To assist the ophthalmologists in performing retinal analysis efficiently and accurately, numerous studies have been conducted to propose an automated retinal diagnosis system. One of the crucial steps for such a system is accurate detection of retinal blood vessels from retinal image. In this paper, we investigated the use of automatic binarization methods on pre-processed fundus image to detect retinal blood vessels. Three methods for binarization were investigated in this study, namely Otsu's method, ISODATA and K-means clustering method. The resulting binarized output indicated good detection of large vessels but most of the smaller vessels were left undetected. To address this issue, Gabor wavelet filter was used to enhance the small blood vessel structures before binarization of the filter output. Combining the binary images from both binarization with and without Gabor filter resulted in significant improvement of the overall detection rate of the retinal blood vessels. The proposed method proved to be comparable to other unsupervised techniques in the literature when validated using the publicly available fundus image database, DRIVE.

Recent technological advancements have changed significantly the way news is produced, consumed, and disseminated. Frequent and on-spot news reporting has been enabled, which smartphones can access anywhere and anytime. News categorization or classification can significantly help in its proper and timely dissemination. This study evaluates and compares news category predictors' performance based on four supervised machine learning models. We choose a standard dataset of British Broadcasting Corporation (BBC) news consisting of five categories: business, sports, technology, politics, and entertainment. Four multi-class news category predictors have been developed and trained on the same dataset: Naïve Bayes, Random Forest, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). Each category predictor's performance was evaluated by analyzing the confusion matrix and quantifying the test dataset's precision, recall, and overall accuracy. In the end, the performance of all category predictors was studied and compared. The results show that all category predictors have achieved satisfactory accuracy grades. However, the SVM model performed better than the four supervised learning models, categorizing news articles with 98.3% accuracy. In contrast, the lowest accuracy was obtained by the KNN model. However, the KNN model's performance can be enhanced by investigating the optimal number of neighbors (K) value.

Retinal image analysis has been widely used for early detection and diagnosis of multiple systemic diseases. Accurate vessel extraction in retinal image is a crucial step towards a fully automated diagnosis system. This work affords an efficient unsupervised method for extracting blood vessels from retinal images by combining existing Gabor Wavelet (GW) method with automatic thresholding. Green channel image is extracted from color retinal image and used to produce Gabor feature image using GW. Both green channel image and Gabor feature image undergo vessel-enhancement step in order to highlight blood vessels. Next, the two vessel-enhanced images are transformed to binary images using automatic thresholding before combined to produce the final vessel output. Combining the images results in significant improvement of blood vessel extraction performance compared to using individual image. Effectiveness of the proposed method was proven via comparative analysis with existing methods validated using publicly available database, DRIVE.

Body Sensor etworks (BS s) are increasingly being used in pervasive sensing environments including healthcare, sports, wellbeing, and gaming. Activity segmentation using BS is challenging and the use of manual annotation is subjective and error prone. In this paper, we investigate a semisupervised activity segmentation method using a Multiple Eigenspace (MES) technique based on Principal Components Analysis (PCA). Results show that the method can reliably perform activity segmentation and the classification results based on HMMs demonstrate the practical value of the proposed technique.