Wei Hu

Microcalcification detection in full-field digital mammograms with PFCM clustering and weighted SVM-based method

Clustered microcalcifications (MCs) in mammograms are an important early sign of breast cancer in women. Their accurate detection is important in computer-aided detection (CADe). In this paper, we integrated the possibilistic fuzzy c-means (PFCM) clustering algorithm and weighted support vector machine (WSVM) for the detection of MC clusters in full-field digital mammograms (FFDM). For each image, suspicious MC regions are extracted with region growing and active contour segmentation. Then geometry and texture features are extracted for each suspicious MC, a mutual information-based supervised criterion is used to select important features, and PFCM is applied to cluster the samples into two clusters. Weights of the samples are calculated based on possibilities and typicality values from the PFCM, and the ground truth labels. A weighted nonlinear SVM is trained. During the test process, when an unknown image is presented, suspicious regions are located with the segmentation step, selected features are extracted, and the suspicious MC regions are classified as containing MC or not by the trained weighted nonlinear SVM. Finally, the MC regions are analyzed with spatial information to locate MC clusters. The proposed method is evaluated using a database of 410 clinical mammograms and compared with a standard unweighted support vector machine (SVM) classifier. The detection performance is evaluated using response receiver operating (ROC) curves and free-response receiver operating characteristic (FROC) curves. The proposed method obtained an area under the ROC curve of 0.8676, while the standard SVM obtained an area of 0.8268 for MC detection. For MC cluster detection, the proposed method obtained a high sensitivity of 92xa0% with a false-positive rate of 2.3 clusters/image, and it is also better than standard SVM with 4.7 false-positive clusters/image at the same sensitivity.

Patch-based denoising method using low-rank technique and targeted database for optical coherence tomography image

Abstract. Image denoising is a crucial step before performing segmentation or feature extraction on an image, which affects the final result in image processing. In recent years, utilizing the self-similarity characteristics of the images, many patch-based image denoising methods have been proposed, but most of them, named the internal denoising methods, utilized the noisy image only where the performances are constrained by the limited information they used. We proposed a patch-based method, which uses a low-rank technique and targeted database, to denoise the optical coherence tomography (OCT) image. When selecting the similar patches for the noisy patch, our method combined internal and external denoising, utilizing the other images relevant to the noisy image, in which our targeted database is made up of these two kinds of images and is an improvement compared with the previous methods. Next, we leverage the low-rank technique to denoise the group matrix consisting of the noisy patch and the corresponding similar patches, for the fact that a clean image can be seen as a low-rank matrix and rank of the noisy image is much larger than the clean image. After the first-step denoising is accomplished, we take advantage of Gabor transform, which considered the layer characteristic of the OCT retinal images, to construct a noisy image before the second step. Experimental results demonstrate that our method compares favorably with the existing state-of-the-art methods.

Implementation of 3-D RDPAD Algorithm on Follicle Images Based CUDA

The 3-D Robust Detail Preserving Anisotropic Diffusion (3-D RDPAD) is an anisotropic diffusion which has many advantages on 3-D ultrasound image processing, especially in 3-D cattle follicle ultrasound Images. The 3-D RDPAD algorithm has good performance both in noise suppression and edge preserving. However, the computation of 3-D RDPAD algorithm requires a huge amount of time to complete because it directly operates on the 3-D image and it has logical judgment in every iteration. In this paper, we proposed a 3-D Parallel RDPAD algorithm using CUDA to solve this problem. By optimizing the logical judgment in diffusion stop condition, the speed of 3-D RDPAD algorithm has a notable improvement. We performed our algorithm on 3-D cattle follicle ultrasound Images and the result of the experiments showed that it revealed a great speedup over the original 3-D RDPAD.

An NTP-Based Test Method for Web Audio Delay

With the application of audio in various fields, the reliability of transmission and delay have been improved to a certain extent. However, there is no quick and effective method to test the delay of audio transmission nowadays. Therefore, it is an urgent matter to propose a reasonable method to test the audio delay. This paper takes web audio as research object, proposing a method of automatically testing web audio delay based on NTP, on the basis of which the delay of web audio is tested from two perspectives namely local delay and network delay. The experimental results show that the auto-delay testing method for web audio proposed in this paper is more flexible and efficient than other testing methods, especially for web audio testing.

An Improved Endpoint Detection Algorithm Based on Improved Spectral Subtraction with Multi-taper Spectrum and Energy-Zero Ratio

Endpoint detection plays a crucial role in speech recognition systems. An effective endpoint detection algorithm can not only reduce the processing time, but also can interfere with the noise of the silent segment. The traditional endpoint detection algorithms are mostly processed in a noise-free environment, so there will be problems such as weak noise immunity. In the problem of low SNR, this paper proposes an improved endpoint detection algorithm based on improved spectral subtraction with multi-taper spectrum and energy-zero ratio. The algorithm uses the improved spectral subtraction method of multi-window spectrum estimation to reduce the speech noise, and then combines the energy-zero ratio with endpoint detection. Experiments show that the proposed algorithm has better robustness under different SNR conditions.

Energy-Aware Fault-Tolerant Scheduling Under Reliability and Time Constraints in Heterogeneous Systems.

As heterogeneous systems have been deployed widely in various fields, the reliability become the major concern. Thereby, fault tolerance receives a great deal of attention in both industry and academia, especially for safety critical systems. Such systems require that tasks need to be carried out correctly in a given deadline even when an error occurs. Therefore, it is imperative to support fault-tolerance capability for systems. Scheduling is an efficient approach to achieving fault tolerance by allocating multiple copies of tasks on processors. Existing fault-tolerant scheduling algorithms realize fault tolerance without energy limit. To address this issue, this paper proposes an energy-aware fault-tolerant scheduling algorithm DRB-FTSA-E. The algorithm adopts the active replication strategy and uses a high utilization of energy consumption to complete a set of tasks with given reliability and time constraints. It finds out all schemes that meet time and system reliability constraints, and chooses the scheme with the maximum utilization of energy consumption as the final scheduling scheme. Experimental simulation results show that the proposed algorithm can effectively achieve the maximum utilization of energy consumption while meeting the reliability and time constraints.

Task scheduling with fault-tolerance in real-time heterogeneous systems

Abstract Nowadays, the performance of heterogeneous systems has been improved dramatically, which also increases the complexity of heterogeneous systems, leading to the growing potential of system failures. Failures can be masked through scheduling approaches. Efficient task scheduling with fault-tolerance can guarantee the execution of tasks and satisfy the real-time nature. In this paper, we address the problem of scheduling tasks on heterogeneous systems with the target to support the maximum number of permanent failures while meeting a given time constraint. The problem is NP-hard and we propose a heuristic algorithm DBSA to solve it. DBSA can dynamically calculate the number of tolerating permanent failures. Firstly, the makespan when systems tolerate a fixed number of failures is calculated. Then, DBSA gets the actual number of tolerating failures by constantly comparing the makespan with a given deadline. Finally, DBSA maps tasks to appropriate processors without violating precedence constraints. Experimental results demonstrate that DBSA can effectively tolerate failures and improves system reliability.