Search

Diffusion tensor imaging is an essential technique to study tissue brain microstructure and the distribution of white matter fiber tracts. However, affected by the diffusion-weighted signal attenuation and long echo time, diffusion tensor images suffer from serious low signal-to-noise ratio problem. Therefore, efficient denoising techniques are crucial for enhancing image quality. This paper starts with the principle of diffusion tensor imaging and the types of noise. Then it discusses the classical diffusion tensor image denoising algorithms, including algorithms based on traditional image processing and deep learning. Special emphasis is given to the status and shortcomings of diffusion tensor image denoising research. The denoising evaluation criteria and commonly used public datasets are also introduced, followed by experiments and quantitative analysis on the diffusion tensor image denoising methods mentioned in this paper. Finally, it concludes with a summary and an outlook for the field’s future research directions.

As global aging worsens and deep learning advances, the imaging classification of Alzheimer’s disease (AD) based on deep learning has become a hot topic of research. This paper reviewed the common deep learning models, evaluation criteria and public datasets in AD imaging classification tasks, discussed the application of different image modalities in AD imaging classification. The content was focused on the improvement of deep learning models applied to AD imaging classification. The studies of model interpretability were also introduced. Finally, the paper summarized and compared the classification models mentioned, identified the brain regions related to AD image classification, and outlined the future research directions in this field.

The classification of the breast imaging-reporting and data system (BI-RADS) based on magnetic resonance imaging (MRI) refers to the classification of the degree of lesions according to the image signs of lesions, which is usually subjective. Moreover, the benign and malignant lesions of BI-RADS 3-5 are overlapping, which is prone to unnecessary invasive treatment due to high diagnostic categories in clinical diagnosis. To address these problems, this research applied radiomics for feature extraction and fusion of T1-weighted (T1W) and dynamic contrast-enhanced (DEC) MRI. The least absolute shrinkage and selection operator (LASSO) algorithm was used to screen out the optimal feature collection of each type of MR image. Support vector machine (SVM), random forest (RF), K-nearest neighbour (KNN) and logistic regression (LR) algorithms were applied for BI-RADS 3-5 classification, based on which the benign and malignant lesions were further classified. The results showed that the classification accuracy of breast BI-RADS 3-5 by four radiomics models based on feature fusion was 81.25%, 87.50%, 78.38%, and 81.25%, respectively. Their accuracy in distinguishing the benign and malignant breast lesions was 90.91%, 93.55%, 92.73%, and 94.55%, respectively. This indicates that the combination of radiomics and machine learning correlation algorithm has a good effect on breast MRI BI-RADS classification and benign and malignant differentiation, and feature fusion can further improve the accuracy of classification prediction.

Constrained spherical deconvolution can quantify white matter fiber orientation distribution information from diffusion magnetic resonance imaging data. But this method is only applicable to single shell diffusion magnetic resonance imaging data and will provide wrong fiber orientation information in white matter tissue which contains isotropic diffusion signals. To solve these problems, this paper proposes a constrained spherical deconvolution method based on multi-model response function. Multi-shell data can improve the stability of fiber orientation, and multi-model response function can attenuate isotropic diffusion signals in white matter, providing more accurate fiber orientation information. Synthetic data and real brain data from public database were used to verify the effectiveness of this algorithm. The results demonstrate that the proposed algorithm can attenuate isotropic diffusion signals in white matter and overcome the influence of partial volume effect on fiber direction estimation, thus estimate fiber direction more accurately. The reconstructed fiber direction distribution is stable, the false peaks are less, and the recognition ability of cross fiber is stronger, which lays a foundation for the further research of fiber bundle tracking technology.

Oral teeth image segmentation plays an important role in teeth orthodontic surgery and implant surgery. As the tooth roots are often surrounded by the alveolar, the molar's structure is complex and the inner pulp chamber usually exists in tooth, it is easy to over-segment or lead to inner edges in teeth segmentation process. In order to further improve the segmentation accuracy, a segmentation algorithm based on local Gaussian distribution fitting and edge detection is proposed to solve the above problems. This algorithm combines the local pixels' variance and mean values, which improves the algorithm's robustness by incorporating the gradient information. In the experiment, the root is segmented precisely in cone beam computed tomography (CBCT) teeth images. Segmentation results by the proposed algorithm are then compared with the classical algorithms' results. The comparison results show that the proposed method can distinguish the root and alveolar around the root. In addition, the split molars can be segmented accurately and there are no inner contours around the pulp chamber.

With the exacerbation of aging population in China, the number of patients with Alzheimer's disease (AD) is increasing rapidly. AD is a chronic but irreversible neurodegenerative disease, which cannot be cured radically at present. In recent years, in order to intervene in the course of AD in advance, many researchers have explored how to detect AD as early as possible, which may be helpful for effective treatment of AD. Imaging genomics is a kind of diagnosis method developed in recent years, which combines the medical imaging and high-throughput genetic omics together. It studies changes in cognitive function in patients with AD by extracting effective information from high-throughput medical imaging data and genomic data, providing effective guidance for early detection and treatment of AD patients. In this paper, the association analysis of magnetic resonance image (MRI) with genetic variation are summarized, as well as the research progress on AD with this method. According to complexity, the objects in the association analysis are classified as candidate brain phenotype, candidate genetic variation, genome-wide genetic variation and whole brain voxel. Then we briefly describe the specific methods corresponding to phenotypic of the brain and genetic variation respectively. Finally, some unsolved problems such as phenotype selection and limited polymorphism of candidate genes are put forward.

Effective medical image enhancement method can not only highlight the interested target and region, but also suppress the background and noise, thus improving the quality of the image and reducing the noise while keeping the original geometric structure, which contributes to easier diagnosis in disease based on the image enhanced. This article carries out research on strengthening methods of subtle structure in medical image nowadays, including images sharpening enhancement, rough sets and fuzzy sets, multi-scale geometrical analysis and differential operator. Finally, some commonly used quantitative evaluation criteria of image detail enhancement are given, and further research directions of fine structure enhancement of medical images are discussed.

Present study used diffusion tensor image and tractography to construct brain white matter networks of 15 cerebral palsy infants and 30 healthy infants that matched for age and gender. After white matter network analysis, we found that both cerebral palsy and healthy infants had a small-world topology in white matter network, but cerebral palsy infants exhibited abnormal topological organization: increased shortest path length but decreased normalize clustering coefficient, global efficiency and local efficiency. Furthermore, we also found that white matter network hub regions were located in the left cuneus, precuneus, and left posterior cingulate gyrus. However, some abnormal nodes existed in the frontal, temporal, occipital and parietal lobes of cerebral palsy infants. These results indicated that the white matter networks for cerebral palsy infants were disrupted, which was consistent with previous studies about the abnormal brain white matter areas. This work could help us further study the pathogenesis of cerebral palsy infants.

Mild cognitive impairment(MCI) is a clinical transition state between age-related cognitive decline and dementia. Researchers can use neuroimaging and neurophysiological techniques to obtain structural and functional information about the human brain. Using this information researchers can construct the brain network based on complex network theory. The literature on graph theory shows that the large-scale brain network of MCI patient exhibits small-world property, which ranges intermediately between Alzheimer’s disease and that in the normal control group. But brain connectivity of MCI patients presents topologically structural disorder. The disorder is significantly correlated to the cognitive functions. This article reviews the recent findings on brain connectivity of MCI patients from the perspective of multimodal data. Specifically, the article focuses on the graph theory evidences of the whole brain structural and functional and the joint covariance network disorders. At last, the article shows the limitations and future research directions in this field.

In this paper, an improved empirical mode decomposition (EMD) algorithm for phonocardiogram (PCG) signal de-noising is proposed. Based on PCG signal processing theory, the S1/S2 components can be extracted by combining the improved EMD-Wavelet algorithm and Shannon energy envelope algorithm. Firstly, by applying EMD-Wavelet algorithm for pre-processing, the PCG signal was well filtered. Then, the filtered PCG signal was saved and applied in the following processing steps. Secondly, time domain features, frequency domain features and energy envelope of the each intrinsic mode function's (IMF) were computed. Based on the time frequency domain features of PCG's IMF components which were extracted from the EMD algorithm and energy envelope of the PCG, the S1/S2 components were pinpointed accurately. Meanwhile, a detecting fixed method, which was based on the time domain processing, was proposed to amend the detection results. Finally, to test the performance of the algorithm proposed in this paper, a series of experiments was contrived. The experiments with thirty samples were tested for validating the effectiveness of the new method. Results of test experiments revealed that the accuracy for recognizing S1/S2 components was as high as 99.75%. Comparing the results of the method proposed in this paper with those of traditional algorithm, the detection accuracy was increased by 5.56%. The detection results showed that the algorithm described in this paper was effective and accurate. The work described in this paper will be utilized in the further studying on identity recognition.

Computer-aided detection (CAD) of pulmonary nodule technology can effectively assist the radiologist to enhance lung nodule detection efficiency and accuracy rate, so it can lay the foundation for the early diagnosis of lung cancer. In order to provide reference for the scholars and to develop the CAD technology, we in this paper review the technology research and development of CAD of the pulmonary nodules which is based on CT image in recent years both home and abroad. At the same time, we also analyse the advantages and shortcomings of different methods. Then we present the improvement direction for reference. According to the literature in recent years, there still has been large development space in CAD technology for pulmonary nodules. The establishment and improvement of the CAD system in each step would be of great scientific value.

This paper presents a probability segmentation algorithm for lung nodules based on three-dimensional features. Firstly, we computed intensity and texture features in region of interest (ROI) pixel by pixel to get their feature vector, and then classified all the pixels based on their feature vector. At last, we carried region growing on the classified result, and got the final segmentation result. Using the public Lung Imaging Database Consortium (LIDC) lung nodule datasets, we verified the performance of proposed method by comparing the probability map within LIDC datasets, which was drawn by four radiology doctors separately. The experimental results showed that the segmentation algorithm using three-dimensional intensity and texture features would be effective.

In medical imaging field, doctors often complete the intra-subject registration of multi modality images by choosing a pair of anatomic landmarks. It is hard to choose the same landmark accurately in different imaging modality. For multi-modality image registration, mutual information measure is widely used because it suffers little from the difference among different modalities, but it has disadvantages of large amount of calculation and slow converges. In this paper, considering the convenience in practice, we firstly register different modality images by choosing a pair of corresponding landmarks. In order to decrease the error during choosing the landmarks, we further do the registration using mutual information method. The advantage of this solution is that the landmark based registration presents a good starting point for further mutual information (MI) registration, while the MI method decreases the difficulty of choosing landmarks. Experimental results showed that this registration solution was fast, accurate, and would have a good clinical potential application in the future.