Integration of Digital Mammography into PACS

Shyh-Liang (Andrew) Lou, Ph.D.

Department of Radiology

University of California, San Francisco

Background

Digital mammography has been considered by some as the missing link of PACS since many large scale PAC systems in the world are running without breast images. The principal theoretical advantage of full-field direct digital mammography (FFDDM) comes from decoupling image acquisition, image storage, and image display, so that these three aspects of the imaging chain can be optimized individually, rather than being subject to the technical trade-off that result from using film for all three functions. Thus, the digital image can be captured electronically, stored digitally, and then manipulated, analyzed, and displayed however, whenever, and wherever it is needed. However, there are three major challenges of integrating FFDDM systems into PACS. First, it is the huge file size of the images. Most FFDDM systems generate images with 4K x 5K x 12 bits in size (40 Mbytes), or 160 Mbytes for two craniocaudal (CC) and two mediolateral oblique (MLO) views in a screening examination. For a comparison study, the images from the current and previous examinations are required. In order to integrate FFDDM in PACS, the system has to handle 320 Mbytes for both transmission and storage in a single patient case.          

          Soft-copy display of the FFDDM images is the second challenging task. Monitors currently available can only display 2,048 x 2,560 pixels with 256 gray levels. These monitors are not adequate to present FFDDM images of 4K x 5K pixels and 4,096 gray levels. Third, the workstation has to be as efficient to display digital mammograms as the conventional light box. This paper reports a customized digital mammogram workstation developed in our laboratory with a preprocessing mechanism coupled in the FFDDM to solve the two aforementioned problems of integrating the FFDDM into PACS using current existing technology.

Materials and Methods

The preprocessing mechanism is to segment the breast from the background of air. There are certain advantages by doing so. First, the image file size is reduced by removing the pixels of the background. Second, the laterality and the locations of the breast can be determined which can be used to position and layout the mammogram when it is first displayed on a monitor. Lastly, only breast image pixels are taken into account for determining the parameters of various breast tissues such as dense glandular, fatty tissue, and skin.

Results

We have tested about 300 mammograms generated from a FFDDM system. The digital mammograms include left and right breasts with CC, MLO, and exaggerated MLO views. For this FFDDM, each image file size is about 46 Mbytes. After the image segmentation, the file size is reduced to 9.6 Mbytes (4.8:1 lossless compression ratio) on average.

          We also conducted a study on the performance of mammogram soft-copy display. A mammographer who is experienced with soft-copy reading reviewed 30 cases on the digital mammogram display workstation. Each case contained one CC and one MLO views. In the first test session, the display program provided a window-and-level slide-bar for the user to manually adjust the settings to present the three tissue types. In the second test session, the display program could first automatically present breast features optimally followed by manual minor fine adjustments by the user. The results indicate that in the first test session, it took the mammographer an average of 14.1,13.6, and 12.9 seconds per case to adjust the settings for presenting the three tissue types of glandular, fatty, and skin, respectively. For the second test session, the mammographer spent 2.1,1.4, and 1.0 seconds on fine tuning, respectively.

Conclusions

Based upon our study results, we conclude that the customized digital mammogram workstation with the breast segmentation process significantly increases the feasibility of integrating digital mammography into PACS.

Acknowledgment

This work is supported in part by U. S. Department of Health and Human Services, National Library of Medicine, Army.

Corresponding author:

Shyh-Liang (Andrew) Lou, Ph.D.

Assistant Professor

Dept. of Radiology, UCSF

"Shyh-Liang (Andrew) Lou" <Andrew.Lou(at)Radiology.UCSF.EDU>

Oral presentation at EuroPACS'98, Barcelona, Spain