W Mimi Diyana W Zaki

Abnormalities detection in serial computed tomography brain images using multi-level segmentation approach

Segmentation, where pixels are categorized by tissue types, is essential in medical image processing. This paper proposes a multi-level Fuzzy C-Means method to extract an intracranial from its background and skull. Then, a two-level Otsu multi-thresholding method is applied to segment the intracranial structure into cerebrospinal fluid, brain matters and other homogenous regions. Based on symmetrical properties in the intracranial structures, the left-half and right-half segmented intracranial regions are quantitatively compared with respect to the intracranial midline. The segmented regions are found to be very useful in providing information regarding normal and abnormal structures in the intracranial because any asymmetry that is detected would indicate a high probability of abnormalities. Additionally, pixel intensity information such as standard deviation and the maximum value of the pixels of the segmented regions are used to distinguish abnormalities such as bleeding and calcification from normal cases. This experimental work uses a medical image database consisting of 519 normal and 201 abnormal serial computed tomography (CT) brain images from 31 patients. The proposed multi-level segmentation approach proved to effectively isolate important homogenous regions in CT brain images. The extracted features of the regions would provide a strong basis for the application of content-based medical image retrieval (CMBIR).

Diabetic retinopathy assessment: Towards an automated system

Abstract The incidence of diabetes and diabetic retinopathy has been shown to be increasing worldwide. While ophthalmologists struggle to treat this retinopathy, they are also faced with an increment of diabetic referrals for eye screening. Screening and early detection of diabetic retinopathy are crucial to help reduce the incidence of visual morbidity and visual loss. In most countries, diabetic retinopathy assessments are done manually. This is time consuming and is a cause of additional clinical workloads. Clinicians are now aware of the need for an automated system for grading Diabetic Retinopathy (DR) that can help in tracing abnormalities in patients’ retinas based on their fundus images, and assist in grading the retina conditions accordingly. This will lead to more effective assessment methods, as well as providing a second opinion to the ophthalmologist during diagnosis. This paper presents an overview of various methods of automated DR grading assessment systems that can complement manual assessments. Tortuosity of the blood vessels is introduced as one of the significant features that can be quantified and associated with DR stages for the grading assessment. From this review, it can be concluded that the automated system has a huge potential for wider acceptance in real life applications. However, there is still some space for improvement for a more robust system. Nevertheless, the DR automated grading assessment system is foreseen as being widely embraced by researchers and ophthalmologists in the future.

An automated 3D scanning algorithm using depth cameras for door detection

This paper presents an investigation on the characteristics of Microsoft Kinect depth camera for door detection in an indoor environment. Autonomous vehicles usually have to rely on images when navigating indoors due to network limitations of an indoor environment. Locating a door for exit and entryway is one of the problems that need to be tackled when navigating indoors. In this paper, images from a depth camera are captured and used as a tool for detecting doors. The continuously varied ratios and depth differences in the door images have been analysed. An algorithm for door detection was developed using MATLAB. Experiments using different heights and depths of the Kinect sensor have been performed to verify the efficacy of the algorithm for indoor autonomous flying robots like the quadcopter. The algorithm developed is best performed in a clear path of 3.5 meters. The accuracy of the measurement was influenced by the low resolution of the depth images.

Qualitative and quantitative comparisons of haemorrhage intracranial segmentation in CT brain images

This paper presents qualitative and quantitative comparisons of our proposed Multi-level Local Segmentation Approach (MLSA) to segment intracranial structures of the CT brain images for haemorrhage detection. The proposed method is able to overcome the main problem in our database images; the inconsistency of grey level values due to different parameter settings during the scanning process that leads to different objects segmented within the same intensity level, as well as helps to automate the segmentation process. One hundred and fifty haemorrhage CT brain images of thirty one patients from Hospital Serdang and Hospital Putrajaya are used in this work. Performance of the segmentation method is quantitatively and qualitatively compared with available automated methods which are watershed and expectation maximization methods. The results show that the MLSA gives the best segmentation of average Percentage of Correct Classification, PCC = 97.1% with 93% of the haemorrhage cases excellently segmented. Besides, qualitatively, it also portrays good segmentation results. The MLSA proves to be accurate and reliable that would provide a strong basis for the application in content-based medical image retrieval.

Retrieval of intracranial hemorrhages in computed tomography brain images using binary coherent vector

We investigate the use of a new binary coherent vector approach, integrated in a proposed content-based medical retrieval (CBMIR) system, to retrieve computed tomography (CT) brain images. Five types of hemorrhages consisting of 150 plain axial CT brain images are queried from a database of 2500 normal and abnormal CT brain images. Possible combinations of shape features are portrayed as feature vectors and are evaluated based on precision-recall plots. Solidity, form factor, equivalent circular diameter (ECD), and Hu moment are proposed as identifying features of intracranial hemorrhages in CT brain images. In addition to identifying hemorrhages, the proposed approach significantly improves the CBMIR system performance. This retrieval system can be widely useful due to rapid development in computer vision and computer database management, both of which motivated this application of CBMIR.

An adaptive nonlinear enhancement method using sigmoid function for iris segmentation in pterygium cases

Pterygium is an eye related disease affected by the fibrovascular tissue that encroaches into the corneal region. Recently, image processing techniques have been explored in the development of pterygium detection system. An iris segmentation module is needed to develop an automatic pterygium detection system of the anterior segment photographed images (ASPI). Qualitatively, the invasion of the pterygium tissues on the iris will result in the imperfect circular iris feature. Thus, an adaptive nonlinear enhancement method using sigmoid function have been proposed in this work to enhance the ASPI. The cutoff and gain factor of the sigmoid function are adaptively calculated based on the tested images. Fifty eight ASPI of various sizes contributed by RAFAEL have been tested using the proposed enhancement method. The proposed method proves to give better visual results, later contributes to more accurate segmented iris regions with accuracy and specificity values of 0.9353 and 0.8818, respectively.

Iris segmentation method of pterygium anterior segment photographed image

Pterygium is an eye disease that commonly affects people living in areas near the equator such as Malaysia, Indonesia etc. and who are expose to excessive wind, sunlight, or sand. It is a form of tissue overgrowth found in the eye. Recently, anterior segment photographed images (ASPIs) have been used for early detection of the disease by incorporating digital image processing (DIP) algorithms techniques which has triggered our interest to investigate such possibilities. As such, this paper reports the early results of iris segmentation of ASPIs that can be used later for pterygium detection. The work involves using the normalized HSV colour space of the iris ASPIs. By using the subtraction method, the iris threshold value was calculated to segment the iris. It is found out that the proposed algorithm can correctly segment the iris with pterygium cases.

Fast, accurate and memory efficient pupil localization based on pixels properties method

One of the most crucial processes in the iris recognition system is iris segmentation. In iris segmentation, the iris region is extracted from the eyelids, eyelashes, eyebrows, pupil and sclera to identify the uniqueness of its patterns. The segmentation process is crucial since the poor and false segmentation can affect the accuracy of the iris recognition system. Since the system will deal with a lot of irises, it must be fast enough to execute the iris segmentation. The speed also crucial in order to implement it in real-time. Pupil localization is a part of iris segmentation process which is to locate pupil boundaries. The pupil boundaries can provide the information of pupil center and radius as a based to locate the iris boundaries. This research is important since poor pupil localization due to reflections and eyelashes interferences can reduce the segmentation accuracy of the iris recognition system. Moreover, the modified segmentation methods based on Hough transform and Integro-differential operator are still complex and require a lot of time. Furthermore, there are many pupil and iris localization methods that assume the shape of pupil as a circle which is not accurate. In this paper, the new pupil localization method which is based on pixels properties is proposed to obtain fast, accurate and memory efficient of pupil localization in the iris recognition system. The proposed method is compared with other methods based on localization accuracy, time to execute and memory usage. According to the results, the proposed method recorded high localization accuracy, low execution time and low memory usage than the other state-of-the-art methods. This proves that the proposed pupil localization method is appropriate and suitable to be implemented in real-time systems.

Effects of different classifiers in detecting infectious regions in chest radiographs

This paper presents the effects of different types of classifiers when analysing the normal and infectious regions in chest radiographs. Three types of classifiers are experimented on: Rule-based, Bayesian and k-nearest neighbours. The evaluation is based on a few criteria, namely, the classification accuracy, misclassification (error), speed, Kappa statistic, ROC area, and other performance measures specifically the true and false positive rates, and precision and recall. The dataset consists of image features from a total of 102 chest radiographs. The normal and infectious lung regions are extracted and divided into non-overlapping sub-blocks prior to the image feature computation. The quantitative results are presented and discussed for consideration in further analysis of infectious lungs.

Content-Based Medical Image Retrieval System for Infections and Fluids in Chest Radiographs

This paper presents a retrieval system based on the image’s content for the application in medical domain. This system is aimed to assist the radiologists in healthcare by providing pertinent supporting evidence from previous cases. It is also useful for the junior radiologists and medical students as teaching aid and training mechanism. The system is tested to retrieve the infections and fluid cases in chest radiographs. We explored several feature extraction techniques to see their effectiveness in describing the low-level property of the radiographs in our dataset. These features are Gabor transform, Discrete Wavelet Frame and Grey Level Histogram. The retrieval of these cases was also experimented with a number of distance metrics to observe their performances. Promising measures based on recognition rate are reported.