Wavelet-based vector quantization for high-fidelity compression and fast transmission of medical images

Abstract

Compression of medical images has always been viewed with skepticism, since the Ioss of information involved is thought to affect diagnostic information. However, recent research indicates that some waveletbased compression techniques may not effectively reduce the image quality, even when subjected to compression ratios up to 30:1. The performance of a recently designed wavelet-based adaptive vector quantization is compared with a well-known waveletbased scalar quantization technique to demonstrate the superiority of the former technique at compression ratios higher than 30:1. The use of higher compression with high fidelity of the reconstructed images allows fast transmission of images over the Internet for prompt inspection by radiologists at remote Iocations in an emergency situation, while higher quality images follow in a progressive manner if desired. Such fast and progressive transmission can also be used for downloading large data sets such as the Visible Human at a quality desired by the users for research or education. This new adaptive vector quantization uses a neural networks-based clustering technique for efficient quantization of the wavelet-decomposed subimages, yielding minimal distortion in the reconstructed images undergoing high compression. Results of compression up to 100:1 are shown for 24-bit color and 8-bit monochrome medical images. Copyright 9 1998 by W.B. Saunders Company C OMPRESSION OF MEDICAL IMAGES forfast transmission is not accepted universally because of possibility of losing diagnostic information. Recently, however, it has been reported I that up to 30:1 compression is acceptable using waveletbased compression techniques, rather than JPEG compression, by which more than 10:1 compression is not considered acceptable by most radiologists.Wavelet coding 2,3 is a class of subband coding that allows selective coding of subbands significant to visual perception. The encoding process actually determines the distortion introduced by quantization. It is well known from rate distortion theory that rector quanfization yields opfimal distortion. However, implementafion of vector quantization specifically the encoding process is quite involved. On the other hand, the decoding process in vector quantization is relafively simple and can be achieved from look-up tables. Generally, a simple statisfical clustering technique is used for generating the code vectors in vector quanfization. 4,5 We have developed novel clustefing techniques 6,7 for on-line generation of code vectors 8,9 and applied them to quantizafion of wavelet-transformed subimages. The selection of the number of clusters generated is decided by the quantization level to be applied, ie, whether coarse or fine quantization is needed. The code vectors representing the clusters are generated here by an adapfive vector quantization technique, namely AFLC-VQ (adaptive fuzzy leader clustefing-vector quantization), lo AFLC-VQ is based on an integration of self-organizing neural networks with fuzzy membership values of data samples for...

Preliminary clinical evaluation of a multi-site telemammography system in a screening mammography environment

“…In remote or underserved clinics in may be necessary to evaluate personnel qualifications in regards to deciding if teleradiology is appropriate and the necessary radiographic procedures. [1][2][3] Many teleradiology systems employ image processing techniques to manage the digital image data in terms of data acquistion, 4-9 transmission time (e.g., compression, 4,10,11,12 cropping, 13 image selection 14 ), and image display. 7,8,10,11,13,14,15 The effects of data management techniques on diagnostic image quality are application specific.…”

Line vector quantization using noninteger subsampled wavelet pyramids and its application in medical imaging

“…The explicit operation of convolving the mid and high frequency subbands with 2-channel bank filters can be avoided. If we denote by 3 T the decomposition matrix of a 3-channel PR bank whose upper rows consist of the shifted low-pass filter, and by 2 T a matrix whose upper third rows have ones on the main diagonal, and the lower two-thirds rows are the reconstruction matrix of a 2-channel bank, then The question may arise: why use spectral weighting to produce homogeneity in the combined band instead of designing the mid and high-pass filters with heavily overlapping (extremely aliased) spectra? The answer lies in the fact that filters in a PR bank "rely" on each other for aliasing cancellation.…”

“…Compressions of up to 18:1 can be achieved if the image is segmented prior to compression, with only the region of interest compressed losslessly 2 ). Lossy compression algorithms, (those algorithms that permit information to be lost gradually), have been reported to produce visually acceptable reconstructed images with compression ratios up to 100:1 3 . While lossy compression is not allowed for diagnostic images in current use, it is allowed for archiving images that have not been in use for a significant period of time.…”

Teleradiology and screening mammography: a telemammography system evaluation and comparison to clinical results

“…There are several image processing techniques commonly used in teleradiology to facilitate handling large amounts of data, which include image compression, image cropping, image selection, and display format. [17][18][19][20][21][22][23][24] In addition, the qualifications of personnel necessary for the successful implementation may need to be evaluated. [25][26][27] We have design and tested a telemammography system capable of handling the large data requirements of mammography that is operated by mammography technologists at the remote sites and experienced radiologists who review transmitted examinations at the central site.…”