Applied Mechanics and Materials Vols. 195-196

Abstract: The visual quality of medical images is an important aspect in PACS implementation. In this study, on the basis of wavelet analysis, a denoising and enhancement algorithm for medical image is proposed. The algorithm mainly includes six steps. At first, an effcient method is investigated for Poisson Noise remove. Second, diagnosis features of the denoised image are enhanced by compressing the dynamic range. Third, we extract the high frequency component of the original image by the designed lowpass filter. Fourth, the extracted high frequency component are segment into diagnosis feature component in the high signal range, the diagnosis feature component in the low signal range, and the noise component. Five, we reconstruct an image using image fusion. Finally, we make DICOM calibration for used display and decide parameters of the image fusion, resulting in the diagnosis image. Experimental results show that this new scheme offers effective noise removal in medical images and enhancing sharpness. More importantly, this scheme can improve the diagnostic value of the display image on the commercial display successfully.

Abstract: Most computer assisted language learning (CALL) systems use acoustic models trained by MLE (Maximum Likelihood Estimation) for pronunciation proficiency evaluation. However, MLE ignores information of other phones during training stage and cant distinguish confusing phones well. This paper introduced discriminative measures of minimum phone/word error to refine acoustic models to deal with the problem. This paper analyzed discriminative trained acoustic models on Putonghua proficiency test in detail and found that: 1) They are much more distinguishable than MLE ones; 2) Even though the training and test are mismatch, they still perform significantly better than MLE-trained models under the same phone boundaries. The final system performance has approximately 4.5% relative improvement.

Abstract: Watermarking, as a new technology, has been applied widely in medical images field for its advantages. This paper emphasizes on these works as follows: firstly analyzes the secure requirements of medical images and watermarking functions from the practical applications view, and then reviews some typical algorithms of the existing watermarking and presents a universal evaluating scheme for medical digital images watermarking, finally, some further works are proposed. The purpose for this paper is to study the present watermarking methods and promote the new schemes emergence effectively.

Abstract: Outward directed surface extraction from imaging modalities is the first task in the design of implants. In this paper a method based on level set method is proposed to extract the directed surface from CT images. The process is composed of two steps. In the first step, Level Set method with a new speed function is employed to evolve zero level set to its destination and used to cut the desired bone part from the input CT images. In the second step, a simple method is used to extract the directed surface, usually the outward surface, from the separated bone part by removing the interior surface. The experimental results show the proposed method works well.

Abstract: Airborne pollen is the main cause of pollen allergy, so statistic of the amount and distribution of pollen in the air is important. This paper presents a method to identify airborne pollen grains in optical microscopy images. After pollen region is segmented by thresholding, global shape descriptor and Fourier descriptor are used to extract shape features, gray level co-occurrence matrix is employed to extract texture features, and finally pollen grains are classified by a k-nearest neighborhood classifier. In the experiment with 55 cases of Poaceae, 44 cases of Moraceae and 48 cases of Pinaceae, a classification rate of 82.31% can be obtained with an accuracy of 85.45%, 61.36%, and 97.91% for each family.

Abstract: Recent development of the lung nodule computer-aided diagnosis (CAD) in helical computed tomography (CT) images has shown great potential in the diagnosis and treatment of the lung cancer. One key technology of the CAD system is the classification of the nodules and non-nodules. In this paper, we try to solve the problem using the ensemble relevance vector machine (ERVM). The contribution of our work includes: 1) relevance vector machine (RVM) is used as the classifier of the CAD system. It has been proven that RVM is comparable to SVM in the generalization capability with a much sparser solution; 2) the ensemble skill is used to process the imbalanced candidates, to acquire a high detection rate with relative low false positives. The proposed method is evaluated using 20 helical CT scans, provided by Guangzhou military hospital. Compared with SVM, RVM is slightly better in accuracy. However, it is much sparser, showing great potential in predicting the huge volume CT scans. 84% of the 25 true nodules are identified by ERVM and only 9.8% of the non-nodules are misclassified. A 40% gain in sensitivity is acquired with the ensemble skill. The results show a fast and satisfactory classification rate.

Abstract: Electrocardiogram (ECG) is a convenient, economic, and non-invasive detecting tool in myocardial ischemia (MI). And its clinical appearance is mainly exhibited by ST-segment deviation. In the paper, the concepts of Correlation Coefficient Entropy (CCE) and Inverse Correlation Coefficient Entropy (ICCE) were proposed and used to compare the differences in morphology variability between ST segments induced by Heart Rate (HR) and by MI. After the Long-Term ST database (LTST) verification, the obvious results obtained with both methods. Whats more, It showed that CCE was better than ICCE comparatively.

Abstract: The traditional wavelet-based image fusion method has some problems, in order to further satisfy the requirements of image fusion on the right direction, this paper studies the principle and performance contourlet transformation based on the proposed sensitivity analysis. In this method, contourlet multi-resolution, locality and direction effectively capture the source image in detail, texture, direction of information, enhance the visibility of the fused image and obtain the higher quality medical images.

Abstract: Given the time-domain measurement of picoseconds pulses and frequency-domain measurement of high-frequency modulation have limitations to realize, this paper introduces the experiment research which use square-wave modulation laser source as incident light source of the diffusion optical tomography in homogeneous medium. Based on the finite-element method, the forward model can be established. The analytical solution of the homogeneous medium can obtained from the derivation of the diffusion equation, after that, we build the optical experiment platform, derive the experimental solution from the platform. Finally, the consistent result is obtained by the comparison of the simulation, the analytical solution and the experimental solution. The result clearly demonstrated the accuracy and effectiveness of the proposed method which use square-wave modulation laser source as incident light source to measure the light intensity.

Abstract: Four-dimensional computed tomography (4D CT) which clearly includes the temporal changes in anatomy during the diagnosis, planning, and delivery of radiotherapy has great promise. Deformable image registration has the potential to reduce the geometrical uncertainty of the target, and makes it possible to signally improve the treatment accuracy by optimizing treatment in response to anatomical uncertainty. In this paper, we used Scale Invariant Feature Transform (SIFT) algorithm to extract landmark points, and we proposed a registration method based on B-Spline model, then used a limited memory quasi-Newton method to optimize the system, also calls the limited memory BFGS (L-BFGS) method. The deformable registration model B-Spline model can derive the images at all intermediate phases from sets of 3D images acquired at a few known phase points. Because 4D CT can track the location of region of interest (ROI) and tumors over several respiratory cycles, so 4D CT can make the apparent size of the tumor which is caused by breathing motion more accurate. The method is evaluated on 10 4D-CT data sets of patients in a breathing cycle.