Yuhua Peng

Texture analysis and classification of ultrasound liver images

Ultrasound as a noninvasive imaging technique is widely used to diagnose liver diseases. Texture analysis and classification of ultrasound liver images have become an important research topic across the world. In this study, GLGCM (Gray Level Gradient Co-Occurrence Matrix) was implemented for texture analysis of ultrasound liver images first, followed by the use of GLCM (Gray Level Co-occurrence Matrix) at the second stage. Twenty two features were obtained using the two methods, and seven most powerful features were selected for classification using BP (Back Propagation) neural network. Fibrosis was divided into five stages (S0-S4) in this study. The classification accuracies of S0-S4 were 100%, 90%, 70%, 90% and 100%, respectively.

New approach to epileptic diagnosis using visibility graph of high-frequency signal.

A new nonlinear approach is presented for high-frequency electrocorticography (ECoG)-based diagnosis of epilepsy. The ECoG data from 3 patients with epilepsy are analyzed in this study. A recently developed algorithm in graph theory, visibility graph (VG), is applied in this research. The approach is based on the key discovery that high-frequency oscillation takes place during epileptic seizure, making it a marker of epilepsy. Therefore, the nonlinear property of the high-frequency signal may be more noticeable. Hence, a complexity measure, called graph index complexity (GIC), is computed using the VG of the patients’ high-frequency ECoG subband. After comparison and statistical analysis, the nonlinear feature is proved to be effective in detection and location of the epilepsy. Two different traditional complexities, sample entropy and Lempel-Ziv, were also calculated to make a comparison and prove that GIC provides better identification.

A Preliminary Study of Fast Virtual Stent-Graft Deployment: Application to Stanford Type B Aortic Dissection

Aortic dissection is the result of blood intruding into the layers of the aortic wall creating a duplicate channel along the aortic course. This considerably changes aortic morphology and thereby alters blood flow, inducing severe pathological conditions. Endovascular stent-graft placement has become an accepted treatment option for complicated Stanford type B aortic dissection. Stent-graft deployment aims to cover the primary entry, preventing most of the inflow to the false lumen, thereby promoting false lumen thrombosis and true lumen expansion. In recent years the application of this treatment has increased continuously. However, a fast and reasonable prediction for the released stent-graft and the resulting aortic remodelling prior to intervention is still lacking. In this paper, we propose a preliminary study on the fast virtual stent-graft deployment algorithm based on contact mechanics, spring analogy and deformable meshes. By virtually releasing a stent-graft in a patient-specific model of an aortic dissection type Stanford B, we simulate the interaction between the expanding stent-graft and the vessel wall (with low computational cost), and estimate the post-interventional configuration of the true lumen. This preliminary study can be finished within minutes and the results present good consistency with the post-interventional computed tomography angiography. It therefore confirms the feasibility and rationality of this algorithm, encouraging further research on this topic, which may provide more accurate results and could assist in medical decision-making.

Contrast response functions with wide-view stimuli in the human visual cortex

In this manuscript, using a novel wide-view visual presentation system that we developed for vision research and functional magnetic resonance imaging (fMRI), we studied contrast response functions in regions of the brain that are central and peripheral to the entire set of visual areas (V1, V2, V3, V3A, MT+), regions that have not been all investigated in previous vision research. Under the stimulus conditions which were 0–20 deg, 20–40 deg, and 40–60 deg eccentricity black-and-white checkerboard patterns, we measured the blood oxygenation level-dependent fMRI contrast response at five contrast levels (6, 12, 24, 48, and 96%) in the visual areas. On the basis of these data, the central and pericentral visual areas had low-contrast gain, whereas the peripheral visual areas had high-contrast gain. In addition, our results showed that the signals fundamentally shift during visual processing through posterior visual cortical areas (V1, V2, and V3) to superior visual cortical areas (V3A and MT+).

Analysis of Frequency Domain of EEG Signals in Clinical Location of Epileptic Focus

As a suddenly abnormal discharge of brain neurons, epilepsy can lead to encephalographic (EEG) abnormalities. Data of 3 epileptic patients recorded by depth intracranial electrodes are analyzed with time–frequency algorithms to observe the abnormal changes in different frequency ranges. Comparing the frequency domain of the normal and epileptic signals that contain several periods of clinical seizures and intervals, a remarkable range of high frequency is identified and corresponds to the seizure. According to the corresponding position of those channels that produced abnormal signals, the epileptic focus is determined theoretically and verified by comparison with the clinical diagnosis and surgical results. Furthermore, the order of the presence of different frequency ranges is listed. The finding proposed in this study could be a fresh method to improve the diagnosis of epilepsy.

An approach to determine myocardial ischemia by hidden Markov models

A hidden Markov model (HMM) of electrocardiogram (ECG) signal is presented for detection of myocardial ischemia. The time domain signals that are recorded by the ECG before and during the episode of local ischemia were pre-processed to produce input sequences, which is needed for the model training. The model is also verified by test data, and the results show that the models have certain function for the detection of myocardial ischemia. The algorithm based on HMM provides a possible approach for the timely, rapid and automatic diagnosis of myocardial ischemia, and also can be used in portable medical diagnostic equipment in the future.

Multi-subject brain decoding with multi-task feature selection

In the neural science society, multi-subject brain decoding is of great interest. However, due to the variability of activation patterns across brains, it is difficult to build an effective decoder using fMRI samples pooled from different subjects. In this paper, a hierarchical model is proposed to extract robust features for decoding. With feature selection for each subject treated as a separate task, a novel multi-task feature selection method is introduced. This method utilizes both complementary information among subjects and local correlation between brain areas within a subject. Finally, using fMRI samples pooled from all subjects, a linear support vector machine (SVM) classifier is trained to predict 2-D stimuli-related images or 3-D stimuli-related images. The experimental results demonstrated the effectiveness of the proposed method.

Simulation based on the heart failure and left ventricular assist devices of the circulatory system

This article established a general model of the human cardiovascular system, also simulated heart failure under different levels of human blood circulatory system, and compared the advantages and disadvantages of the cardiac assist device in a different location (ascending aorta and descending aorta). These experiments have some guiding significances to clinical trials.