Chunhua Dong

Computer-Aided Diagnosis and Quantification of Cirrhotic Livers Based on Morphological Analysis and Machine Learning

It is widely known that morphological changes of the liver and the spleen occur during the clinical course of chronic liver diseases. In this paper, we proposed a morphological analysis method based on statistical shape models (SSMs) of the liver and spleen for computer-aided diagnosis and quantification of the chronic liver. We constructed not only the liver SSM but also the spleen SSM and a joint SSM of the liver and the spleen for a morphologic analysis of the cirrhotic liver in CT images. The effective modes are selected based on both its accumulation contribution rate and its correlation with doctors opinions (stage labels). We then learn a mapping function between the selected mode and the stage of chronic liver. The mapping function was used for diagnosis and staging of chronic liver diseases.

Simultaneous Segmentation of Multiple Organs Using Random Walks

Random walks-based (RW) segmentation methods have been proven to have a potential application in segmenting the medical image with minimal interactive guidance. However, the approach leads to large-scale graphs due to number of nodes equal to voxel number. Also, segmentation is inaccurate because of the unavailability of appropriate initial seed points. It is a challenge to use the RW-based segmentation algorithm to segment organ regions from 3D medical images interactively. In this paper, a knowledge-based segmentation framework for multiple organs is proposed based on random walks. This method employs the previous segmented slice as prior knowledge (the shape and intensity constraints) for automatic segmentation of other slices, which can reduce the graph scale and significantly speed up the optimization procedure of the graph. To assess the efficiency of our proposed method, experiments were performed on liver tissues, spleen tissues and hepatic cancer and it was extensively evaluated both quantitatively and qualitatively. Comparing our method with conventional RW and state-of-the-art interactive segmentation methods, our results show an improvement in the accuracy for multi-organ segmentation (p < 0.001).

Segmentation of liver and spleen based on computational anatomy models

Accurate segmentation of abdominal organs is a key step in developing a computer-aided diagnosis (CAD) system. Probabilistic atlas based on human anatomical structure, used as a priori information in a Bayes framework, has been widely used for organ segmentation. How to register the probabilistic atlas to the patient volume is the main challenge. Additionally, there is the disadvantage that the conventional probabilistic atlas may cause a bias toward the specific patient study because of the single reference. Taking these into consideration, a template matching framework based on an iterative probabilistic atlas for liver and spleen segmentation is presented in this paper. First, a bounding box based on human anatomical localization, which refers to the statistical geometric location of the organ, is detected for the candidate organ. Then, the probabilistic atlas is used as a template to find the organ in this bounding box by using template matching technology. We applied our method to 60 datasets including normal and pathological cases. For the liver, the Dice/Tanimoto volume overlaps were 0.930/0.870, the root-mean-squared error (RMSE) was 2.906mm. For the spleen, quantification led to 0.922 Dice/0.857 Tanimoto overlaps, 1.992mm RMSE. The algorithm is robust in segmenting normal and abnormal spleens and livers, such as the presence of tumors and large morphological changes. Comparing our method with conventional and recently developed atlas-based methods, our results show an improvement in the segmentation accuracy for multi-organs (p<0.00001).

Zero watermarking for medical images based on DFT and LFSR

Medical images and data may encounter tampering, compromising patients privacy and other safety issues in the process of storage and transmission of information. This paper proposes an algorithm of zero watermarking to address these issues using DFT. The algorithm avoids the sophisticated process of finding the Region of Non-Interest (RONI) of medical images, and uses a part of sign sequence of DFT coefficients as the feature vector of images for enhancing the robustness. Even if the algorithm is public, it can still withstand malicious attacks by using the private key to protect the medical image. Experimental result indicates that the watermarking scheme belongs to the blind watermarking extraction with strong robustness and invisibility to withstand rotation, scaling, translation, cropping and other attacks. Moreover, it can embed much more data compared with the existing watermarking techniques.

Non-rigid image registration with anatomical structure constraint for assessing locoregional therapy of hepatocellular carcinoma

PURPOSE Assessing the treated region with locoregional therapy (LT) provides valuable information for predicting hepatocellular carcinoma (HCC) recurrence. The commonly used of assessment method is inefficient because it only compares two-dimensional CT images manually. In our previous work, we automatically aligned the two CT volumes to evaluate the therapeutic efficiency using registration algorithms. The non-rigid registration is applied to capture local deformation, however, it usually destroys internal structure. Taking these into consideration, this paper proposes a novel non-rigid registration approach for evaluating LT of HCC to maintain the image integrity. METHOD In our registration algorithm, a global affine transformation combined with localized cubic B-spline is used to estimate the significant non-rigid motions of two livers. The proposed method extends a classical non-rigid registration based on mutual information (MI) that uses an anatomical structure term to constrain the local deformation. The energy function can be defined based on the total one associated with the anatomical structure and deformation information. Optimal transformation is obtained by finding the equilibrium state in which the total energy is minimized, indicating that the anatomical landmarks have found their correspondences. Thus, we can use the same transformation to automatically transform the ablative region to the optimal position. RESULTS Registration accuracy is evaluated using the clinical data. Improved results are obtained with respect to all criteria in our proposed method (MI-LC) than those in the MI-based non-rigid registration. The landmark distance error (LDE) of MI-LC is decreased by an average of 3.93mm compared to the case of MI-based registration. Moreover, it could be found regardless of how many landmarks applied in our proposed method, a significant reduction in LDE values using registrations based on MI-LC compared with those based on MI is confirmed. CONCLUSION Our proposed approach can guarantee the continuity, the accuracy and the smoothness of structures by constraining the anatomical features. The results clearly indicate that our method can retain the local deformation of the image. In addition, it assures the anatomical structure stability.

The Medical Image Watermarking Algorithm with Encryption by DCT and Logistic

When medical images transmitted and stored in hospitals, it require strict security, confidentiality and integrity. However, the transmission of wireless and wired networks has made the medical information vulnerable to attacks like tampering, hacking etc. And the ROI of medical image is unable to tolerate significant changes. In order to dealing these problems, we have proposed an algorithm that introducing the digital watermarking technology to increase the security of medical images. The scheme uses a part of sign sequence of DCT coefficients as the feature vector of images. It can avoid the sophisticated process of finding the Region of Interest (ROI) of medical images. At the same time, the watermarking image is encrypted by Logistic Map to enhance its confidentiality. The experimental results show that the scheme has strong robustness against common attacks and geometric attacks. Moreover, compared with the existing medical watermarking techniques, it can embed much more data, less complexity and make embed multi watermarks realized.

A DWT-DCT Based Robust Multiple Watermarks for Medical Image

In the watermarking algorithm of the medical image, the principal problems faced are how to determine the region of interest (ROI) and improve the hiding capacity. This paper proposes a robust multiple watermarks algorithm for the medical image based on DWT and DCT, which can effectively solve the referred problems. The algorithm combines the visual feature vector of images, encryption technology with the third party authentication, and avoids the tedious process for selecting ROI. This algorithm can enhance medical image security, confidentiality and integrity in the application for the clinical. The results of experiment indicate that the watermark scheme has strong robustness, and can embed much more data compared with the existing watermarking techniques.

The medical images watermarking using DWT and Arnold

Medical image data require strict security, confidentiality and integrity when transmitted between hospitals. The involvement of wireless transmission has made the medical data vulnerable to attacks like tampering, hacking etc. To dealing the challenges we have proposed a algorithm that introducing the digital watermarking technology to increase the security of medical images when they transmitted through a wireless network. The scheme uses a part of sign sequence of DWT-DCT coefficients as the feature vector of images for enhancing the robustness against common attacks and geometric attacks. The watermarking image is scrambled by Arnold transform to enhance its privacy. The experimental results show that the scheme has benefits at visual invisibility and robustness. Moreover, it can embed much more data, less complexity and more practicability than the existing watermarking techniques, because it avoids the sophisticated process of finding the Region of Interest (ROI) of medical images.

Nonrigid registration for evaluating locoregional therapy of hepatocellular carcinoma

The assessment of the treated margin with locoregional therapy (LT), for hepatocellular carcinoma (HCC), is the common method for predicting HCC recurrence in most hospital. However, tumors sometimes cannot be removed clearly with LT in limited conditions. The therapeutic efficiency of HCC is often evaluated by comparing 2D fusion images of computed tomography (CT) or magnetic resonance imaging (MRI) between the preoperation and the postoperation. However, judgment about whether the tumors exist in the treated margin after LT by using 2D slices sometimes is difficult. It is desirable to develop a suitable image registration algorithm to automatically align the two volumes in order to transform the treated margin of the postoperative volume to the tumor of the preoperative volume to assess the therapeutic efficiency after treatment of HCC. With taking these into consideration, this paper proposed an automatic 3D fusion imaging approach for medical image by using the nonrigid registration method that aligning an ablative margin - that is the treated margin after LT, onto the locations of HCC. In our registration algorithm, a rigid global transformation combined with localized B-spline is used to estimate the significant nonrigid motions of the liver between before and after LT. Our proposed approach can ensure the feasibility, the accuracy and the efficacy to assess the treated margin for HCC. Furthermore, this method can be adapted to register multi-modality medical images. We demonstrate the effectiveness of our proposed method by comparing the difference criterions of fusion evaluation on medical images. The results clearly indicate that our method extremely useful in the evaluation of the treated margin, in addition, it remain the motion and local deformation of the volume.

Hybrid method combining superpixel, random walk and active contour model for fast and accurate liver segmentation

Organ segmentation is an important pre-processing step in surgery planning and computer-aided diagnosis. In this paper, we propose a fast and accurate liver segmentation framework. Our proposed method combines a knowledge-based slice-by-slice Random Walk (RW) segmentation algorithm (proposed in our previous work) with a superpixel algorithm called the Contrast-enhanced Compact Watershed (CCWS) method to reduce computing time and memory costs. Compared to the commonly used Simple Linear Iterative Clustering (SLIC), we demonstrate that our CCWS is more appropriate for liver segmentation. To improve the methods accuracy, we use a modified narrow band active contour model as a refinement after the initial segmentation. The experiments showed that the superpixel-based slice-by-slice RW could segment the entire liver with improved speed, and the modified active contour model is more precise than the original Chan-Vese Model. As a result, the proposed framework is able to quickly and accurately segment the entire liver.