Biomedical Engineering: Applications, Basis and Communications

Heterogeneous thyroid nodules have distinct components and vague boundaries in ultrasound (US) images. It is difficult for radiologists and physicians to manually draw the complete shape of a nodule, or distinguish what kind of components a nodule has. Hence, this article presents an automatic process for nodule segmentation and component classification. A decision-tree algorithm is used to segment the possible nodular area. A refinement process is then applied to recover the nodular shape. Finally, a hierarchical method based on support vector machines (SVMs) is used to identify the components in the nodular lesion. Experimental results of the proposed approach were compared with those of other methods.

Hyperthyroidism is a common thyroid disease. Graves' disease is one of the most common etiologies with 70–80% of hyperthyroidism. In clinical diagnosis, physicians generally utilize ultrasound (US) images for inspections. Unfortunately, they are not able to decide the Graves' disease directly on the US images, and have to rely on blood tests to assure it. The blood tests often take weeks to obtain the inspection results. Hence, in this article, we proposed an approach to diagnose Graves' disease on US image directly. The automatic diagnosis meets a short-time inspection, and patients may know their situation quickly. We segment the thyroid regions and utilize a high-performance classifier (support vector machine — SVM) to classify the regions. The Graves' disease is diagnosed using an area measure on the classified regions. Experimental results show effectiveness of the proposed approach.

In this article, we present a method to construct retinal vascular tree with the spatial information and temporal intensity profile of vessels. The significance of the work is to assist in clinical studies of diagnosis of cardiovascular diseases, such as hypertension, which manifest abnormalities in either venous and/or arterial vascular systems. For temporal intensity profile, fluorescein angiogram (FA) is adopted to ensure continuity of vessel segments along time. We separate retinal vascular tree construction into two main steps — vessel extraction and vessel grouping. In vessel extraction, we use an existing algorithm that applies multiscale matched filtering and vessel likelihood measure to extract vessel segments and its spatial information in color retinal image. Vessel segments are grouped using extended Kalman filter to take into consideration continuities in curvature, width, and fluorescein angiogram sequence changes at the bifurcation or crossover point. The algorithm is evaluated on a clinical data set from Taipei Veterans' General Hospital, and the results are compared with the ground truth images provided by a physician. The results show that our method reaches an average correctness rate of 91.48%.

This article presents a new multispectral signature detection approach to magnetic resonance (MR) image classification. It is called the gray system theory (GST) method. In recent years, although GST has been used in various fields, applying the GST to medical images is very rare. The theoretical basis of GST is the gray hazy set; the main element is composed of an uncertain or abstract set and through a series of algorithms, the gray messages or phenomena are transformed into white ones. The concept of the theory is based on the information coverage, which establishes a new regular set based on source data and then the set is mapped to a 2D space. A series of experiments are conducted and compared to the commonly used fuzzy C-means (FCM) method, extension theory, and perceptron for performance evaluation. The results show that the GST method is a promising and effective spectral technique for MR image classification.

Determining the structure of a protein is not an easy task, which usually involved a time-consuming and costly process in the web lab. Using computational methods to predict a protein's tertiary structure from its primary structure (the amino acid sequence) is desirable. Disordered regions are segments of a protein that do not have a fixed conformation, which makes the structure prediction harder. Also, these disordered regions are functionally important for a protein. In this research, we would like to identify such regions with a focus on selecting a proper feature set. Three feature selection methods, namely F-score, information gain (IG), and k-medoids clustering, are used for feature selection. The support vector machine (SVM) is then used for classification. The results show that the classification accuracy can be raised with a smaller feature set. The k-medoids clustering feature selection can reduce the number of features from 440 to 150 and improve the accuracy from 84.66 to 86.81% in five-fold cross validation. It also has a more stable performance than F-score and IG.

As breast cancer is a substantial threat to the lives of women, it has become a major health issue in the world over the past 50 years, and its incidence has increased in the recent years. Early diagnosis and suitable treatment is relatively important. In the process of breast screening, tissue biopsy is an important operation in determining the presence of breast cancer. It not only provides an accurate diagnosis of the disease but also determines the prognosis for breast cancer. The main goal of this study is to develop a breast cancer diagnosis system based on histopathology and a sequence of image-processing technologies to analyze H&E stained images of breast tissues. The proposed system can automatically detect the mitosis of nuclei and analyze the size and the shape of nuclei to evaluate the duct structure of the breast tissue. Moreover, it provides physicians quantitative prognosis and classification of tissue malignancy, which will improve the diagnostic accuracy and efficiency of the cancer.

This study proposes a diagnosis system for liver masses based on the improved radial basis function (RBF) neural networks. In this article, RBF networks are improved by sigmoid function and the growing and pruning algorithm. The proposed improved RBF networks adopt the sigmoid function as their kernel due to its increased flexibility over the Gaussian kernel. Furthermore, the growing and pruning algorithm is used to adjust the network size dynamically according to the neuron's significance. This investigation formulates discriminating among cysts, hepatoma, cavernous hemangioma, and normal tissue as a supervised learning problem. The current work calculates several texture and gray-level features derived from regions of interest as input in the proposed classifier. Receiver operating characteristic (ROC) curves evaluate the diagnosis performance, and demonstrate the proposed method's good performance.

In this article, we propose an automatic lung segmentation method. We design an ROI-adjusted initial boundary derivation method to estimate a proper initial lung boundary for ASM deformation by deriving the translation and the scaling parameters from the lung ROI. An adaptive ASM, using k-means clustering and silhouette-based cluster validation technique, is proposed to adapt to the lung shape change so that the lung shape variation among people can be overwhelmed. The experimental results indicate that the performance of the proposed lung segmentation method is superior to the traditional ASM approach.

Objective: To design a novel percutaneous stented valve and model its implantation in the aorta.

Background: The dimensions of stented aortic valve components govern its ability to prevent backflow of blood into the left ventricle. Whilst the theoretical parameters for the best stent performance have already been established, an effective valve model and its suitability along with the stent are lacking.

Methods: This article discusses the design of a stented valve suitable for percutaneous aortic valve replacement. Steps involved in 3D CAD-based geometric modeling of the stented aortic valve and its implantation in the aorta are presented. Conceptual designing of individual components was used to build the total geometric model.

Results: A novel geometric model of percutaneous stented aortic valve was generated. The improved design enhances its performance during and after implantation.

Conclusion: The blunt hooks in the stent model prevent its migration in either direction by getting embedded in the aortic endothelium. This novel stent aortic valve may be of great interest to designers of future bioprosthetic heart valve models, as well as to surgeons involved in minimally invasive valve surgeries.

Magnetic resonance (MR) imaging plays an important role in the preoperative evaluation of the living-related liver donors to reduce postoperative morbidity and complications. Though invasive, intraoperative cholangiography is still the golden standard in demonstrating anatomical variation of bile ducts. We design this study to compare the validity and reliability between these two imaging modalities and retrospectively searched the database of living-related liver transplantation (LRLT) center during the period from April 2005 to September 2008. The clinical information and images of total 58 donors (female : male = 35 : 23) were reviewed. Five types of branching patterns of the bile ducts were encountered. The rate of disagreement in the classification of branching pattern between MR and intraoperative cholangiography was 5% (3/58). Intraoperative cholangiography could detect significantly more division degree of right hepatic ducts (mean degree: 7) than MR cholangiography (mean degree: 4) (using paired t-test, p < 0.05). No significant difference existed in the detection of division degree of left hepatic ducts. However, MR cholangiography was significantly better than intraoperative cholangiography in demonstrating major branches of the left hepatic ducts including tributaries of segment 2 and 3 (58/58 versus 35/58) (using McNemar's chi-square test, p < 0.05). We conclude that MR cholangiography is nearly as good as intraoperative cholangiography in identifying the branching pattern of bile ducts, and demonstrates the major left hepatic ducts better.