Imaging and Distributed Computing Group
Information and Computing Sciences Division
Lawrence Berkeley Laboratory
Berkeley, CA 94720

Publication number: LBL-35331




The goal of the Whole Frog Project is to be able to represent the anatomy of a frog in a computer in 3D space in such a way that a high school student will find it useful in learning physiology, hopefully developing a sense of interest in using computers at the same time.


The original plan is to make use of Magnetic Resonance Imaging (MRI) to produce a sequence of 2D gray scale image slices of the whole frog. Photographs are also to be taken of real slices of the frog, to be used as an aid in the visual segmentation of the animal. Thus the position of the frog needs to be maintained as closely as possible during the MRI and actual slicing. To accomplish this task, the frog's toes are glued to a platform, and its upper body is lifted and rested on Styrofoam during the MRI scan. After the the scan, the frog is moved to a freezer with its position unaltered for later slicing and photographing. The result of the 50 slices of MRI data is not too encouraging. The bones are very well represented, but the abdominal region is distorted. It is almost impossible to distinguish the main organs. However, the photograph data greatly exceeds our expectations. Not only does it have better resolution than the MRI data, its colors help distinguish different organs. Consequently, the photograph data is used instead of the MRI data.


After the frog has been frozen to -30 degree C, it is placed on a metal plate. Then a sheet of aluminum foil is placed under the metal plate, and folded up on all sides around the frog. Wax is poured on the metal plate, around the frog in order to stabilize it during the vivisection procedure. As soon as the wax solidifies, the plate is mounted in a cyrotome, a machine designed to make micro meter thin slices in organic material . A sharp blade is mounted at one end of the cyrotome. The cutting procedure involves the movement of the plate toward the blade for a slice and then retraction (under the blade for another slice). As the plate retracts to the extreme end, its height is increased by a precise level of set measurement. Then, it moves toward the blade for the next slice. In this experiment , the frog is raised by 50 micron for each slice. A camera and lighting apparatus are placed over the cryotome, and are focused on the frog. The cryotome's blade is then adjusted to cut at the proper angle. With the frog, the blade, leverage, the lighting and camera set correctly, the cryotome is turned on. Inside the cryotome, the metal plate slides past a stationary blade, removing a 50 micron slice of the frog with every pass. A picture is taken of the sliced surface of the frog every five slices. By the end of this process, 136 pictures are taken.

The next step in this process involves making the photograph data assessable to computer manipulation. First, a digitizing camera is used to capture the pictures for later use in the computer. After all the photographs are converted to images in the computer, it is necessary to use some image processing techniques to align and enhance the images.


In order to display an individual part of the frog in 3D space, there is a need to segment or isolate that part. The segmentation is done on the images of the 2D slices and is called a mask, which is made up of white and black pixels. The segmentation is done using a program called Segmentation Analyzer(Segal) which uses the technique of thresholding, region growing and hand drawing to establish the boundary of the region of interest within each 2D slice. With thresholding, an area of interest , make up of a range of gray values, is given. The result of thresholding is to have the pixels that fall in a particular range to be set to white while the rest are set to black. This produces a mask of black and white value. The growing method also involves a given range of gray values. In addition one or more points(pixel) are given. Any pixels in the neighborhood of the given pixels that fall in the range of the gray values will get set to white and the rest set to black. Most of the time, the growing or the thresholding method do not segment or isolate the desired object perfectly. So the hand drawing is always used to touch up the final mask.

Segal has one additional feature which helps the process of segmentation. This feature allows a few masks to be loaded at once and enables different masks to be exclusive. When a mask becomes exclusive, its region is completely ignored by any of the segmentation methods. In other words, that region can not be drawn on. This method avoids overlapping masks when segmenting objects are adjacent to each other, like bone and muscle. If one is to start segmenting the more obvious objects then the hardest objects will not be too difficult to segment later with the help of exclusive masks.

In this project, 13 masks are generated. They are used to define the skeleton, eyes, brain, kidney, heart, lung, nerve, blood vessel, stomach, large intestine, liver, small intestine, and muscles.


Volume data is formed when all the slices of data images are stacked up. Each point in that volume space is called a voxel. A program named Sunvoxel is used to convert the 3D segmented image data of the frog into a projected view that can be displayed on the workstation screen. This program is capable of direct rendering of 3D data in a semi- transparent surface. It can handle multiple substances (e.g heart, lung, etc) with transparency and color value assigned to them. Each substance is classified by its range of voxel values. With the aid of the masks, the pixels of each substance of the frog are scaled to a certain range so that they can be treated as a distinct substance. With the classification, color, and opacity set, the frog can be rotated to any direction for viewing. By changing the opacity value of the substances, one can view objects that are blocked by other objects.


Photographing Lighting

The first problem involves the inability to keep the lighting constant when taking the original set of photographs at the cryotome. The need to produce pictures that are consistently lit comes from the fact that, once inside the computer, different pixel values will be assigned to each tiny piece of the picture. Since most organs contain generally one color throughout, the entire organ is expected to have fairly consistent pixel values. Consistency of pixel values aid in mask making. When taking the pictures, the external light from a window near the cytotome was not taken into consideration. The sectioning and photography process took about eight hours, and consequently, as the day progressed, the pictures became darker as the day did. The artificial light that was provided by a lamp set up did little to stop the contrast between the daylight and night pictures.


The next problem came about because the zoom factor was changed at two points while taking the original pictures at the cryotome. Consistency of magnification throughout all the pictures is important because it alleviates the need to adjust the digitized images to the correct size in the computer. Another cause for readjustment of the digitized pictures came from our inability to maintain the correct position of the frog on the sliding track of the cryotome with each photograph. Each slight deviation from the correct position made it necessary to center each frame of the digitized pictures in the computer.

Digitizing Lighting

We also had a problem with keeping the lighting constant throughout the digitizing process. The lighting for all the 136 pictures had to be the same in order to obtain consistent pixel values. This was very difficult using the high intensity reflective lights necessary for such a task. This problem caused some of the frames used in the computer to be darker than others.


After digitizing all the photos, the process of visual segmentation of the frog had to begin. Since the frog could only be seen in frozen cross section, identifying all the organs posed a challenge. Organs could only be seen as micron thin slices were cut away, causing them to look somewhat distorted. The visual segmentation process was also complicated by the fact that the representation of the organs had to be made with hand drawings using a computer mouse pointing device. The accuracy of the organs was thus limited to the experimentors drawing ability.

Inconsistent Data

The photograph data was not perfect. First the thickness of each slice was not constantly 250 micron throughout the 136 slices. Photos were taken every 5 slices of 50 micron thick. But the crytome machine jammed at least 6 times. Each time it was jammed, the plate was lowered and the slicing motion was tested on a slice of the frog to find out what had gone wrong. With each of the test slices, the thickness was not guaranteed to be 50 microns, in fact it was most likely less than 50 microns. To make matters worse, the the machine sometimes jammed on a test slice. There might also have been human error involved in keeping count on the number of slices for the photographing. Secondly, the plane of slicing was changed by the loosening of the blade. One factor causing the machine to jam was due to the loosening of the blade. When the blade was tightened its angle might have changed. Thus, the plane of slicing was affected. All these problems were evident when a mapping of the photograph data to the MRI data was conducted. Since the MRI data has 50 slices and the photograph data has 136 slice, every 2 to 3 slices of data in the photograph data should be taken out to match each MRI data. But at some levels, as many as 10 slices were necessary to be taken out.

Higher Resolution Desire

Finally, not enough slices were made to make a good representation of all the major systems. Veins in the circulatory system are often smaller than 50 microns. Thus the circulatory system was seen in only broken pieces in the reconstruction from 250 micron photographs.


Dr. Paul Licht, a specialist in amphibian anatomy, was consulted on how another experiment of the same sort might be completed with more success. First, he advised us to perform a regular segmentation of a second frog, using it as a reference for the visual segmentation. Next, he suggested taking more pictures in order to capture the entire circulatory system. The circulatory system could also be highlighted by injecting florescent dye into the frog's blood stream. He also made note of the fact that our frog was an immature male, and he suggested that we get a more mature specimen for a more complete representation of the reproductive system.

Most of the problems encountered can be avoided in the next experiment. Misalignment can be solved by setting up markers around the frog so that every photograph captures the markers. The zoom factor of the camera lens should be kept constant throughout the photographing. Since the camera is hanging, it might be advisable to lock the zoom lens to prevent it from sliding down.

Minimal inconsistent data can be achieved by avoiding jamming the machine. The blade should be tightened frequently to avoid loosening. Also foreign objects other than the frog and wax should be avoided. The Styrofoam used to hold the frog up was believed to have caused the loosening of the blade as well. It will be wise to make note of the current frame number when the machine jams. Also try to avoid removing the metal plate as it will change the slicing angle. Lastly, avoid leaning on the machine to prevent moving the machine.

During the digitizing procedure, make sure that all four light bulbs are changed when one or more bulbs go out. It will be a good idea to have a photographer set up the camera and lighting before the slicing procedure.

A better data set should be obtained with the proper light setting. This should also help segmentation. As Dr. Licht suggested a real dissection of a frog would help the segmentation. The computer mouse is not a good device to trace objects during segmentation. An alternative device like a pen and digitizing tablet might do a better job.

Berkeley Lab | DST | Notice to Users | Whole Frog Project | Virtual Frog

Page last modified: 03/25/19
Contacts: Bill Johnston, David Robertson