Search

Respiratory motion results in significant motion blur in thoracic and abdomen PET imaging. The extent of respiratory motion blur is mainly correlated with breathing amplitude, tumor size and location. In this paper we introduce a statistical study to quantitatively show the factors influencing the extent of respiratory motion blur in thoracic PET images. The study is centered on two regression models, one is linked with motion blur induced loss of mean intensity(LMI), tumor motion magnitude and tumor size, and another is to investigate the influence of tumor location, patient gender and patient height on tumor motion magnitude. We use the blur identification and image restoration technique to estimate the tumor motion and compute the LMI. The regression model was validated by simulation and phantom data before extended to 39 cases of clinical lung tumor PET images corrupted with blurring artifact. Results show that the motion magnitude of lung tumor during breathing is 10.9+/-3.7mm in transaxial plane, and it is significantly greater in lower lung lobes than in upper lobes. The LMI is 7.1+/-2.4% in the region of interest (ROI) above 40% of the image's maximum intensity. The least-square estimate of regression equations demonstrates that LMI is proportional to tumor motion magnitude and is inversely proportional to tumor size; the two factors play the same role in determining the extent of respiratory motion blur in thoraco-abdominal PET imaging. The location of tumor was shown as the major factor determining its motion magnitude, while the influencing of patient gender and height on tumor motion was not shown significant.

This paper presents a level set based method for bladder abnormality detection on T1-weighted MR images. First, the bladder wall is segmented by using a coupled level set framework, in which the inner and outer borders of the bladder wall are extracted by two level set functions. Then, the middle layer of the bladder wall is founded and represented by a new level set function. Finally, the new level set function divides the bladder wall into several layers. The inter-layer intensity of all voxels in each layer is sorted in ascending order to generate the inter-layer intensity curve. The results prove the effectiveness of inter-layer intensity curve in indicating the emerging of the bladder abnormalities.

Respiration induces significant movements of thorax and abdomen, which shift target site and thus degrade the accuracy and efficacy of radiotherapy. 4D CT sorting is an indispensable and important step to retrospectively rearrange the reconstructed CT images sampled at scan positions to get a set of volume images corresponding to different respiratory phases. Currently, most current sorting methods depend on external surrogates of respiratory motion, which may not always accurately catch the internal motion especially when irregular breathing occurs and lead to severe image artifacts. The 4D CT images produced by our method present fewer artifacts than that by RPM phase-based sorting, which shows that our method is feasible to sort 4D CT images without using external motion monitoring system.

Four-dimensional computed tomography(4D CT) is significant in radiotherapy treatment planning for thorax and upper abdomen to take their motion induced by respiration into consideration, but its high radiation dose becomes a major concern and impedes its wide application. To solve the problem, we propose an image interpolation approach to get 4D CT simulation images. We simulate 4D CT images at arbitrary intermediate phases by B-Spline deformable model with cosine interpolation of the deformation field, which is obtained by deformable registration of two CT images at end-exhale and end-inhale phases. The mean of absolute differences computed between actual 4D CT images and simulation ones is used to evaluate the accuracy of simulation. Our experiment results show that both linear interpolation and cosine interpolation with proper parameters perform well and the latter performs a little better than the former in general.

Anatomy position annotation of brain CT Axial Slice is an important step in content-based image retrieval. In this paper, we provide an efficient approach to automatically estimate the approximate anatomy position of brain CT axial slices in two steps: First, decide whether the input image is encephalic image or nasal cavity image using vote scheme based on the classification results with features extracted by Gabor filter, sobel operator and gray-level co-occurrence matrix (GLCM) respectively; Second, annotate the approximate anatomy position of encephalic images using nonnegative tensor factorization (NTF). The approach has 99% accuracy in distinguishing between encephalic images and nasal cavity images and over 90% accuracy in automatic position annotation of encephalic images.

Measurement of tissue optical properties with optical coherence tomography (OCT) shows that scattering coefficient has an effect on the OCT signal slope. The slope is calculated by linear fitting with a certain range of OCT signal; hence, the effect of fitting range on the scattering induced change of slope is an important topic. In this paper, we study the scattering induced change of OCT signal slope versus depth in Intralipid suspensions with different concentrations based on Monte Carlo simulation. The results show that the OCT signal slope expresses a contrary change with scattering coefficient below a certain depth in high Intralipid concentrations, so there is an effective fitting depth. The effective fitting depth is approximately 600 μm for 10% Intralipid and increases with decreasing concentrations. The effective fitting depth is due to the face that multiple scattering photons of backscattered light become dominant below the depth, and it is related with the numeric aperture (NA) of the object lens.

Nonnegative Matrix Factorization (NMF) is a powerful tool for gene expression data analysis as it reduces thousands of genes to a few compact metagenes, especially in clustering gene expression samples for cancer class discovery. Enhancing sparseness of the factorisation can find only a few dominantly coexpressed metagenes and improve the clustering effectiveness. Sparse p-norm (p > 1) Nonnegative Matrix Factorization (sp-NMF) is a more sparse representation method using high order norm to normalise the decomposed components. In this paper, we investigate the benefit of high order normalisation for clustering cancer-related gene expression samples. Experimental results demonstrate that sp-NMF leads to robust and effective clustering in both automatically determining the cluster number, and achieving high accuracy.

Optical coherence tomography (OCT) presents its capability of noninvasive glucose measurement in tissue-simulating phantoms and biological tissues. However, speckle noise substantially deteriorates the accuracy of the measurements with this technique. The speckle noise can be suppressed by averaging of in-depth scans. We studied the suppression of speckle noise for accurate measurement of backscattering signal and the maximum precision of glucose measurement in Intralipid suspensions with the OCT technique. Our results show that the precision of glucose measurement can achieve plusmn4.4 mM for 10% Intralipid and plusmn2.2 mM for 3% Intralipid after the averaging of 50,000 scans.

Retrieving images is a useful tool to help radiologist to check medical image and diagnosis. In this paper, we propose a retrieval system of computer aided brain magnetic resonance imaging diagnosis , which integrated archiving and communication system, feature extraction and storage, feature match, and user interfaces. The system can be access in three ways, (i) using our system including PACS and image retrieval system, (ii)embedded the software into current PACS, and (iii) internet connection to submit an image as a query, and receive in return a set of images from the the web system. The system utilizes sophisticated image processing algorithms, combined with robust database technology and state-of-the-art web-enabled hardware, and packages these capabilities into a tightly-integrated diagnosis environment for the radiologist. This paper outlines the architecture, algorithms, and the test/verification procedures that our system has implemented.

Disease categorization based on medical image is a very challenge problem, because disease anatomical distribution on image is complex. In this paper, we describe an approach of region feature extraction, which is to partition the image into different regions and extract the local features in each region, to classify different diseases. Our experiments are based on a database consisting of 50 samples with two different disease types(stroke and tumor). The experimental results demonstrate our method is effective in classification precision comparing with Gabor feature extraction.

In this paper, we propose an effective pre-processing method of CT brain images to provide consistent feature for content-based medical image retrieval. The key steps in pre-processing include cutting out background region, using ellipse to correct lean imaging angle, and normalization. We take vast CT brain images as experimental data to evaluate the method. Experimental results show the effect of pre-processing.

Content-based image retrieval for medical image is a main way for computer-aided diagnosis. To storage medical image into the huge databases is its premise. In this paper, we provides efficient approach to develop the archives of large brain CT medical data, including quality control, ensuring consistency/integrity of information, expert verification of institutional findings, image correction, the validation for registration, and region of interest focusing on attention mechanism. The database is highly promising and shows the profit when it is used to implement brain CT image assistant diagnosis systems.

With contrast to conventional MRI, diffusion tensor image (DTI) can demonstrate the anatomic structure and pathological process of white matter tracts and it is widely used by clinician. However, at present, DTI and MRI images are diagnosed respectively and clinician gets the image information objectively, and this leads to different diagnosis conclusions. In this article, we develop an integrated approach that aims to maximize the amount of image information for clinician by combining anatomical and functional MR images with nerve fiber bundles derived from diffusion tensor imaging to assist treatment solution for brain tumor. Firstly, segmentation on MRI T2 is operated to decide solid tumor border or edema area et al; and then DTI and MRI images are fused together, which can supplied not only individual information but also fusion information. This new visualization facility which is intuitive for clinician to understand and use can adequately mine this source of image information. Finally, white matter fibers affected by tumor were analyzed quantitatively. The combined visualization and computational methods have the potential to assist in preoperative surgery and radiotherapy planning as well as postoperative treatment evaluation.

There exist two classical linear methods for feature extraction, i.e. principal component analysis (PCA) and Fisher discriminant analysis (FDA). PCA best represents the data while FDA best separates the data in the least squares sense with different scatter measures from samples. This paper discusses a regularized scatter measure (RSM) as a linear combination of within-class and between-class scatters for feature extraction. The tradeoff between for representation and for discrimination is controlled via some suitable regularization parameters and the corresponding eigenvalue problem is resolved without singularity. Experiments on two different size data sets demonstrate the effectiveness of the method. In addition, we can see that the counterpart of PCA, i.e. minor component analysis (MCA), is to optimize one special case of RSM. And this provides another easy way for understanding why MCA outperforms PCA for feature extraction in one-class classification problem.

Objective: Nonnegative matrix factorization (NMF) has been proven to be a powerful clustering method. Recently Cichocki and coauthors have proposed a family of new algorithms based on the @a-divergence for NMF. However, it is an open problem to choose an optimal @a. Methods and materials: In this paper, we tested such NMF variant with different @a values on clustering cancer gene expression data for optimal @a selection experimentally with 11 datasets. Results and conclusion: Our experimental results show that @a=1 and 2 are two special optimal cases for real applications.

In microarray data analysis, each gene expression sample has thousands of genes and reducing such high dimensionality is useful for both visualization and further clustering of samples. Traditional principal component analysis (PCA) is a commonly used method which has problems. Nonnegative Matrix Factorization (NMF) is a new dimension reduction method. In this paper we compare NMF and PCA for dimension reduction. The reduced data is used for visualization, and clustering analysis via k-means on 11 real gene expression datasets. Before the clustering analysis, we apply NMF and PCA for reduction in visualization. The results on one leukemia dataset show that NMF can discover natural clusters and clearly detect one mislabeled sample while PCA cannot. For clustering analysis via k-means, NMF most typically outperforms PCA. Our results demonstrate the superiority of NMF over PCA in reducing microarray data.

Glioma is one of malign tumors due to the special construction of the glia cell and its character of infiltration. The treatment, such as surgical resection and radiotherapy, needs the precise tumor boundary. To identify noninvasively the margin of the tumor, using metabolic information by proton magnetic resonance spectroscopic imaging (1H-MRSI) has been approved to be a powerful tool. In this paper we adopt 1H-MRSI data for feature extraction and employ support vector machine(SVM) to classify every voxel in the region of interest (ROI) into either glioma or normal tissue, and then to infer the margin of glioma. Experimental results on 1H-MRSI glioma data demonstrate that proposed method is effective and show a better performance compared with recent popular method.

Brain computed tomography (CT) image based computeraided diagnosis (CAD) system is helpful for clinical diagnosis and treatment. However it is challenging to extract significant features for analysis because CT images come from different people and CT operator. In this study, we apply nonnegative matrix factorization to extract both appearance and histogram based semantic features of images for clustering analysis as test. Our experimental results on normal and tumor CT images demonstrate that NMF can discover local features for both visual content and histogram based semantics, and the clustering results show that the semantic image features are superior to low level visual features.

This paper considers gene expression data classification by discriminative mixture models in which sparseness of training data features controls the learning rate. Our goal is to improve the sparseness of training features reduced by nonnegative matrix factorization (NMF). We use the generalized Lp-norm NMF for reducing the high dimensional gene expression data. Experimental results on four real gene expression datasets show that, the classification accuracy can be significantly improved by using the generalized method, and especially that it is first to adopt L2-norm NMF for dimension reduction.

Purpose: High radiation dose in CT scans increases a lifetime risk of cancer and is always a major concern. Recently, iterative reconstruction algorithm with Total Variation (TV) regularization has been developed to reconstructCTimages from highly undersampled data in order to reduce the imaging dose. Nonetheless, CTimagesreconstructed in this approach are sometimes over‐smoothed and edge information is lost. In this work, we developed an iterative CTreconstruction algorithm with edge preserving TV regularization to accurately reconstructCTimages from highly undersampled data. Methods and Material: The CTimage is reconstructed by minimizing TV norm under a constraint posed by few X‐ray projections. To avoid over‐smoothed edge, an edge‐preserving penalty was proposed and added to the TV norm. We tested our reconstruction algorithm on 4 cases, namely, Shepp‐Logan head phantom, NCAT phantom at thorax region, a real patient head case and a real patient lung case. The reconstruction accuracy was evaluated using relative error and edge correlation coefficient. Conventional FDK filtered backprojection algorithm and conventional TV regularization method without edge preserving term were also studied for comparison purpose. Results: It is found that 30 equi‐angle projections are enough for reconstructing the CTimages in phantom experiments and 40 projections for real patient experiments. Our approach outperforms the FDK algorithm and the conventional TV algorithm in terms of the relative error and the edge correlation coefficient in all the four test cases. In particular, the reconstructed images using our approach contain less noise and sharp edges. Conclusion: The edge‐preserving TV performs better in CTimages reconstruction. About 30∼40 X‐ray projections are enough to accurately reconstruct the CTimages, which implies a potential radiation dose reduction by about 20 times.