Annupan Rodtook

Multi-feature gradient vector flow snakes for adaptive segmentation of the ultrasound images of breast cancer

Segmentation of ultrasound (US) images of breast cancer is one of the most challenging problems of the modern medical image processing. A number of popular codes for US segmentation are based on a generalized gradient vector flow (GGVF) method proposed by Xu and Prince. The GGVF equations include a smoothing term (diffusion) applied to regions of small gradients of the edge map and a stopping term to fix and extend large gradients appearing at the boundary of the object. The paper proposes two new directions. The first component is diffusion as a polynomial function of the intensity of the edge map. The second component is the orientation score of the vector field. The new features are integrated into the GGVF equations in the smoothing and the stopping term. The algorithms, having been tested by a set of ground truth images, show that the proposed techniques lead to a better convergence and better segmentation accuracy with the reference to conventional GGVF snakes. The adaptive multi-feature snake does not require any hand-tuning. However, it is as efficient as the standard GGVF with the parameters selected by the brutal force approach. Finally, proposed approach has been tested against recent modifications of GGVF, i.e. the Poisson gradient vector flow, the mixed noise vector flow and the convolution vector flow. The numerical tests employing 195 synthetic and 48 real ultrasound images show a tangible improvement in the accuracy of segmentation.

Continuous force field analysis for generalized gradient vector flow field

We propose a modification of the generalized gradient vector flow field techniques based on a continuous force field analysis. At every iteration the generalized gradient vector flow method obtains a new, improved vector field. However, the numerical procedure always employs the original image to calculate the gradients used in the source term. The basic idea developed in this paper is to use the resulting vector field to obtain an improved edge map and use it to calculate a new gradient based source term. The improved edge map is evaluated by new continuous force field analysis techniques inspired by a preceding discrete version. The approach leads to a better convergence and better segmentation accuracy as compared to several conventional gradient vector flow type methods.

Migration planning using modified Cuckoo Search Algorithm

Mobile agent is a distributed computing technology which allows the agents to carry their jobs on behalf of users from current processing server to the others. Hence, the migration planning is a significant issue for improving performance of mobile agent. In this paper, we propose an improving method for finding path of mobile agent migration planning based on Cuckoo Search Algorithm. Our proposed method modified the original Cuckoo Search Algorithm in two ways. That is, for the first one: in finding new nest, we propose random replacement instead of Lévy fight algorithm. Then, for the second one: in destroying egg, we propose a context sensitive parameter instead of a constant parameter. The experimental results, comparing with migration using the former proposed method, original Cuckoo Search Algorithm and Ant Colony System, confirm that the proposed migration algorithm improves the path finding within the limited time scale.

Fall detection using directional bounding box

Falls are significant public health problem. In the last few years, several researches based on computer vision system have been developed to detect a person who has fallen to the ground. This paper presents a novel fall detection technique namely the directional bounding box (DBB) to detect a falls event especially a situation of fall direction paralleling the line of cameras sight. The DBB is constructed with perspective side view transformation of depth information. Moreover, a new aspect ratio namely the center of gravity point (COG) is proposed to monitor human movement. The proposed technique was evaluated with the video data set gathering from a RGB-D sensor. The experimental result of the proposed technique was better both accuracy and response times than previous works.

A filter bank method to construct rotationally invariant moments for pattern recognition

We propose multiresolution filter bank techniques to construct rotationally invariant moments. The multiresolution pyramid motivates a simple but efficient feature selection procedure based on a combination of a pruning algorithm, a new version of the Apriori mining techniques and the partially supervised fuzzy C-mean clustering. The recognition accuracy of the proposed techniques has been tested with the reference to conventional methods. The numerical experiments, with more than 50,000 images taken from standard image datasets, demonstrate an accuracy increase ranging from 5% to 27% depending on the noise level.

Phase portrait analysis for automatic initialization of multiple snakes for segmentation of the ultrasound images of breast cancer

Segmentation of ultrasound (US) images of breast cancer is one of the most challenging problems of modern medical image processing. A number of popular codes for US segmentation are based on the active contours (snakes) and on a variety of modifications of gradient vector flow. The snakes have been used to locate objects in various applications of medical images. However, the main difficulty in applying the method is initialization. Therefore, we suggest a new method for automatic initialization of active contours based on phase portrait analysis (PPA) of the underlying vector field and a sequential initialization of trial multiple snakes. The PPA makes it possible to exclude the noise and artifacts and properly initialize the multiple snakes. In turn, the trial snakes allow us to differentiate between the seeds initialized inside and outside the desired object. While preceding methods require the manual selection of at least one seed point inside the object or rely on the particular distribution of the gray levels, the proposed method is fully automatic and robust to the noise, as can be seen from the tests with synthetic and real images.

Automatic exudates detection in retinal images using efficient integrated approaches

Diabetic Retinopathy with exudates causes a major problem in human visualization and becomes a cause of blindness to diabetic patients. In addition, the numbers of diabetic retinopathy patients are increasing while the numbers of doctors are not easily increased in the same proportion. This circumstance causes a heavy work load for doctors. In the past, the medical image processing research has shown that simply getting a second opinion can significantly help physicians diagnosis. This research proposes a method to detect exudates from diabetic retinopathy images. The early exudates detection of diabetic retinopathy patients will reduce seriousness in diabetic retinopathy. The proposed method for detecting exudates consists of 5 major steps as follows: 1) To improve the quality of images by using the contrast limited adaptive histogram equalization (CLAHE) 2) To apply the object attribute thresholding algorithm (OAT) for non-retinal object removal, 3) To implement Frangis algorithm based on Hessian filtering for blood vessel detection 4) To detect the retinal optic disc by applying the combination between multi-resolution analysis and Hough transform and 5) To classify exudates in the remaining region with algorithms of hierarchical fuzzy-c-mean clustering. The performance of the proposed method is evaluated on DIARETDB, which is the retinal image database of the Lappeenranta University of Technology, where the performance is good enough for exudates detection.

Selection of multiresolution rotationally invariant moments for image recognition

We propose multiresolution filter bank techniques to construct rotationally invariant moments. The multiresolution pyramid motivates a simple but efficient feature selection procedure based on a combination of a pruning algorithm, a new version of the Apriori mining techniques and partially supervised fuzzy C-mean clustering. The recognition accuracy of the proposed techniques has been tested with the reference to conventional methods. The numerical experiments, with more than 50,000 images, demonstrate an accuracy increase ranging from 5% to 27% depending on the noise level.

Semi-automated detection of breast mass spiculation using active contour

The use of computer research for breast cancer diagnosis in digital mammograms has been studied by some researchers for years. The researches based on medical image processing were developed and published continuously. Theirs objective are to create a diagnostic tool that can increase the accuracy of risk analysis for breast cancer. At the early stage, cancers may be identified as spiculated masses revealing architectural distortion. This research proposes a semi-automated method to detect architectural distortion characterized by thin lines radiating from its margins. It will help physicians as second or minor opinion before biopsy operation. The proposed method involves following major steps in sequence. A combination of the object attributes thresholding, hill-climbing and region growing algorithm is applied to digital mammogram for background and breast pectoral muscle removal. The second is a region of interest (ROI) selection based on image splitting and breast ratio estimation. In the third step, the shade corrections of ROI are considered by using the contrast-limited adaptive histogram equalization. Next, we apply the modified hierarchical clustering to detect and enhance the possible cluster of spiculated masses. The other clusters will be a significant reduction. The final step is established to segment spiculated shape by employing the parametric active contour method. The numerical experiments of the proposed method are performed by testing on the digital database for screening mammography (DDSM) made up by the University of South Florida.

Intrusion feature selection using Modified Heuristic Greedy Algorithm of Itemset

This paper proposes the Modified Heuristic Greedy Algorithm of Itemset (MHGIS) as a feature selection method for Network Intrusion Data. The proposed method can be use as an alternative method to gain the proper attributes for the proposed domain data: Network Intrusion Data. MHGIS is modified from original Heuristic Greedy Algorithm of Itemset (HGIS) to increase efficiency for finding proper feature. In our work, we compare our result with the common method of feature selection is which the Chi-Square (Chi2) feature selection. There are 4 main steps in our experiment: Firstly, we start with data pre-processing to discard unnecessary attributes. Secondly, MHGIS feature selection and Chi2 feature selection have been employed on the pre-processed data, to reduce the number of attributes. Thirdly, we measure the recognition performance by using supervised learning algorithms which are C4.5, BPNN, RBF and SVM. Lastly, we evaluate the results received from MHGIS and Chi2. From the KDDCup99 dataset, we got 13,499 randomly sampling patterns with 34 data dimensions. With the use of MHGIS and Chi2 algorithms, we obtain 14 and 26 features respectively. The result shows that, the classification accuracies measure by C4.5 over the MHGIS selection algorithm produces better accuracies as compare to the Chi2 feature selection and HGIS feature selection over all types of classification methods.