Research On Application Of Multi-scale Analysis In Medical Imaging Diagnostic System

With the advances in digital medical imaging devices, diagnostic imaging approach has been a fundamental change for physician. Currently, Medical Imaging Diagnosis System(MIDS) based on "soft reading " approach is gradually popular in the clinical application. In Medical Imaging Diagnosis System, the process of "soft reading " for medical images is as follows: Firstly the collected medical images will be transmitted and saved in the Imaging Server, and then the images are queried and read by the Imaging Workstation, through the observation of the images displayed on the screen, the doctor gives the clinical diagnosis. Therefore, the efficient storage, a high-speed transmission and a high-definition screen display are the key to improve the efficiency and accuracy for imaging diagnostic.The thesis focus on the status that exorbitant dose CT examination in recent years and thousands of two-dimensional image sequences taken in one CT examination, which posed a serious challenge to MIDS in the effective data storage, high-speed transmission and a three-dimensional display quality, conduct the research on the application of medical diagnostic imaging systems on the basis of multi-scale geometric analysis.And the main contributions of this thesis are as follows:1. We investigated and studied the application of medical image compression and interpolation based on the multi-scale geometric analysis, then gave a detailed anatomy of the multi-scale geometric analysis of two representative tools: wavelet and shearlet. Then the thesis discussed the transformation characteristics of these two tools as well as the distribution regularity of transform coefficients. 2. Aiming to the status that JPEG 2000 has been supported by the international DICOM standard, the two image compression algorithms of JPEG2000 are mainly studied: the SPIHT coding algorithm based on the tree structure and EBCOT coding algorithm based on the block structure. Through the simulation experiment, we verify and compare the effectiveness and superiority of these two compression methods.3. In order to optimize the efficiency of the huge storage and transmission for the medical CT images, in accordance the compression superiority of the EBCOT coding, EBCOT coding is selected as the basis for optimization, then put forward a new adaptive truncation coding compression algorithm based on the wavelet transform is researched. First the CT images are de-noised by the Bayesian thresholding, Secondly, the processed wavelet coefficients are block coded by truncating adaptively the bit stream. Where, the channel numbers are decided according to the relative importance of wavelet sub-bands. At last, the reconstructed image is obtained by decoding and inverse wavelet transform. The experimental results show that the proposed algorithm not only can reduce the computational complexity, but also can improve the quality of the reconstructed image.4. In order to improve the diagnostic accuracy of diagnostic imaging systems, in the case of adding a specific dose of Poisson noise, a new interpolation algorithm is proposed based on Non Subsampled Shearlet Transform for low-dose CT images. Compared with the inter-layer interpolation algorithm based on wavelet transform, the experimental results illustrate that, in a certain dose of noise, the algorithm can get a better interpolation reconstruction effect.