World Wide Web Viewing and Distribution of Visible Human Images

Kevin W. McEnery, M.D.

Lucky K. Kelley, B.A.

Elizabeth A. Kelly, MLS

S. Bruce Dowton, M.D. (Syd.)

Washington University School of Medicine
St. Louis, MO

Purpose:

Develop a method for efficient distribution of the Visible Human image database at a medical university campus. Reduction in the bandwidth requirements for distribution was accomplished through a combination of image cropping, background color reduction, and reduction to an optimized 256 (8 bit) color look-up table. Image sets were specifically cropped to the head & neck, thorax, abdomen/pelvis, spine, and lower extremity. A CGI application provides a universal interface with any currently available WWW browser including Netscape, Mosaic, AOL browser, and MS Navigator.

METHODS:

Image FTP and post-processing

The raw visible human data set consists of over 1878 images 4 Mb compressed images and uncompressed 7 Mb of data per image. For the medical educator, while these images are of high detail, the 7 Mb images make them impractical to efficiently browse and manipulate.

As supplement medical education, we determined that every 5^th image (5mm increment) was sufficient for illustration of cross-sectional anatomy. A total of 375 images were retrieved via the Internet from the National Library of Medicine FTP site. Uncompressed this data set would have required 2.6 Gb of image data storage. However, we wanted to maintain the entire project on an available computer with a 1 Mb hard drive.

We chose to reduce total image storage requirements through a combination of cropping, reducing the background to a single color, and reducing the color depth to 16 bit and 8 bits for Internet distribution. Other types of compression are available including JPEG and GIF. GIF compression was ultimately used for distribution because it could be viewed directly with any available WWW browser.

The raw image size of an individual image was 2048 x 1206. It was determined that cropping the raw images, a standard image size of 1771 x 1021 at 16 bits/pixel. We chose this size as the smallest size which would completely accommodate the largest thorax image with both arms included.

To allow viewing of the images in the same orientation as is standard for medical imaging, the images were flipped in both the horizontal and vertical plane.

An important aspect of the color reduction process was the elimination of the various blue colored hues of the surrounding blue ice enclosure and replacement with a solid black background. This was essential in the final reduction to 8 bit color. As the background was represented by a single color (black), the 256 color look-up table could represent the natural brown hues of the cross sectional images. Without this color reduction, the color table had large predominance of the blue hues found in the surrounding ice resulting in visible deterioration of image quality.

A Power Macintosh 7200 with Ethernet connection accomplished the downloading and subsequent image processing. Fetch was the FTP application. The same Macintosh computer functions as the image server for the project.

This entire process of downloading, decompression, image cropping, background color reduction, color table reduction/optimization, flipping the images from top to bottom and left to right, was achieved entirely through the scripting of these various program applications. These included Fetch (for FTP), PhotoShop, and Debabelizer (image processing). The entire 374 image data set was downloaded over the course of a three day week-end without an operator in attendance. The final full-resolution, 16 bit image set is approximately 390 Mb. Several anatomy focused image sets were abstracted from this reduced size master image set. (Table 1).

CGI browser application

A WWW server was established on a Macintosh 7200. Raw images were further reduced in size and then a composite image was created for each image with a line drawing indicating the approximate level of the represented anatomic section. (Illustration 1). A CGI application written in Applescript creates all HTML pages based upon user interactions.

In interacting with the program, the user selects the anatomic area and then the exact level to be viewed by clicking on the drawing icon at the level to be viewed. The prior and next image in the series can be selected. If a user wanted a full resolution image, a single click on the anatomic image would retrieve the full resolution image. This could be saved by the user for incorporation in to other projects.

The network bandwidth saved by this approach is substantial. The user can view the image series and then determine the exact areas to be studied. The navigation icon images are about 70 Kb each. It is assumed that while there would be extensive interaction with the icon images, only those high resolution images that are of specific interest to the user will be downloaded.

Conclusion:

The Internet provides a method for efficient network browsing and display of the visible human images. Through the use of an interactive browser, faculty and students are able to navigate to a specific area of interest and then download a high resolution image which can be incorporated into interactive and printed teaching materials. The utilization of varying resolution of images provides an efficient retrieval method. We anticipate widespread utilization of this resource.

Table 1. Relative images sizes

Number Pixel Size Image Size

Entire image set Single Total< /B>

Compressed 1878 2048 x 1206 4 Mb 7512 Mb

Non-compressed 1878 2048 x 1206 7 Mb 13176 Mb

Selected images

Compressed 375 2048x1206 4 Mb 1500 Mb

Non-compressed 24bit 375 2048 x 1206 7 Mb 2625 Mb

Cropped 16 bit color 375 1021 x 771 96 Kb - 2 Mb 390 Mb

Number Pixel size Image size Kb (8 bit GIF)

Index images 390 x 260 55-100

Head & Neck 75 649 x 729 100-357

Thorax 46 990 x 838 638-867

Abdomen 101 1090 x 1021 800-950

Spine 151 470 x 455 179-255

Lower Extremity 210 660x833 - 300x350 383-459

Table 1. Relative images sizes
	Number	Pixel Size	Image Size
Entire image set			Single	Total< /B>
Compressed	1878	2048 x 1206	4 Mb	7512 Mb
Non-compressed	1878	2048 x 1206	7 Mb	13176 Mb

Selected images
Compressed	375	2048x1206	4 Mb	1500 Mb
Non-compressed 24bit	375	2048 x 1206	7 Mb	2625 Mb
Cropped 16 bit color	375	1021 x 771	96 Kb - 2 Mb	390 Mb


	Number	Pixel size	Image size Kb (8 bit GIF)
Index images		390 x 260	55-100
Head & Neck	75	649 x 729	100-357
Thorax	46	990 x 838	638-867
Abdomen	101	1090 x 1021	800-950
Spine	151	470 x 455	179-255
Lower Extremity	210	660x833 - 300x350	383-459

illus1.gif - 49.4 k

Illustration 1. Interface appearance of the navigation, interface. Image size of 60 Kb makes navigation through the images practical even with a modem. Clicking on the image downloads a "full resolution image". Left icon indicates approximeate level of current image.