Image Compression in Digitized Mammograms

Mónica Penedo¹, María J. Lado¹, Pablo G. Tahoces², Miguel Souto¹, Juan J. Vidal¹

^1. Department of Radiology of the University of Santiago de Compostela, SPAIN
(Complejo Hospitalario Universitario de Santiago)
². Department of Electronics and Computer Science of the University of Santiago de Compostela, SPAIN

Introduction

Digital technology is likely to replace conventional screen-film dedicated mammography. However, in digital mammography, detectability of radiological features, such as microcalcifications, is hampered by the quality of the digitized images. Therefore, archiving and retaining the amount of data generated in mammography will need an efficient data compression scheme that could reduce the digital data without degradation of the image quality.

Wavelet transforms are particularly well-suited for image compression [1]. An alternative scheme for encoding wavelet coefficients, called embedded zerotree wavelet algorithm (EZW), was described by Shapiro [2]. Some of the ideas underlying EZW have been substantially modified and enhanced by Said and Pearlman [3]. Their new and different implementation is called Set Partitioning in Hierarchical Trees (SPIHT).

In this work we have evaluated the image quality of mammographic images compressed with a wavelet-based method using an SPIHT coding algorithm by calculating the root mean square (RMS) error.

Material and Methods

Digital Images

Three conventional mammograms containing clusters of subtle microcalcifications were digitized resulting in images of 4096x4974 pixels quantized on 12 bits. Because of computational requirement for processing the entire breast image, we manually extracted an ROI of 1024x1024 pixels from each image. This three ROIs were compressed at ratios of 40:1 and 80:1 and then decompressed by using a wavelet-based method. For all images and at all compression ratios the RMS error was calculated.

Wavelet-based compression method

The compression method used computes a discrete wavelet transform followed by a coding of the wavelet coefficients with an SPIHT algorithm. This algorithm is based on: (1) exploitation of the hierarchical structure of the wavelet transform, (2) partial ordering of the transformed coefficients by magnitude, and (3) ordered bit plane transmission of refinement bits for the coefficient values. The discrete wavelet transform uses the Daubechies filters with a basis of length four to obtain a seven-level decomposition of the original image.

Results

A wavelet-based compression method using an SPIHT coding algorithm was applied to three 1024x1024-pixel ROIs at 12 bits. The RMS error between the original and the decompressed images at each compression ratio was calculated. The numerical results are presented in Table I.

Table I. RMS error for all compressed-decompressed images at two different ratios

Discussion

SPIHT has been applied to several chest radiographs by Savcenko et al [4]. They used a wavelet transform with the 9-tap/7-tap biorthogonal filters to obtain a five-level decomposition of the original images. They found no substantial difference in the diagnostic accuracy between images compressed at 40:1 and 80:1 and uncompressed images. Manduca [1] evaluated a conventional wavelet compression method using the same filters. An 8 bit digital mammogram compressed at a 25:1 ratio with a RMS error of 2.01 was obtained with this method.

The results obtained with this wavelet compression method using an SPIHT coding algorithm indicate that mammographic images could be compressed at ratios between 40:1 and 80:1 without significant image degradation. This suggests that digital mammograms could be compressed at an 80:1 ratio without degradation in the visibility of microcalcifications. Future work will address, using a larger set of images, the compression of the whole mammogram and focus on the diagnostic accuracy of the compressed images.

References

1.- Manduca A., “Compressing Images with Wavelet/Subband Coding”. IEEE Engineering in Medicine and Biology 14(5):639-646, 1995.

2.- Shapiro J.M., “Embedded image coding using zerotrees of wavelet coefficients”. IEEE Transactions on Signal Processing 41(12):3445-3462, 1993.

3.- Said A., Pearlman W.A., “A new fast and efficient image codec based on set partitioning in hierarchical trees”. IEEE Trans on Circuits and Systems for Video Technology 6:243-250, 1996.

4.- Savcenko V., Erickson B.J., Palisson P.M., Persons K.R., Manduca A., Hartman T.E., Harms G.F., Brown L.R., “Detection of Subtle Abnormalities on Chest Radiographs after Irreversible Compression”. Radiology 203 (3):609-616, 1998.

Corresponding author:

          Mónica Penedo

          Departamento de Radiología, Facultad de Medicina

          C/ San Francisco, 1

          15704 Santiago de Compostela, Spain

          Phone: 34 - 81- 570982

          Fax: 34 - 81- 547031

          e-mail: mrmocs(at)uscmail.usc.es

---

Oral presentation at EuroPACS'98, Barcelona, Spain

---