TECHNICAL NOTE

Widefield Microscopy Images of Tissue Sections by Computer Imaging Techniques

Correspondence to: Isamu Ikeda, Dept. Dermatology, Mitsui Omuta Hospital, Tenryo 1-100, Omuta City, Fukuoka 836, Japan.

	Summary

Top Summary Introduction Materials and Methods Results Discussion Literature Cited

A fine photomicrograph covering one whole specimen is very useful in the study of skin histopathology. However, because it is almost impossible to take such a picture with a conventional photomicroscope, we attempted to make one with the aid of a computer. The large field was divided into small fields, which were individually recorded through a photomicroscope. The images were then digitized and processed with a computer to reconstruct the largefield image. A fine seamless image was reconstructed with this method. We can thus extend the field of a photomicroscope with the aid of computer imaging techniques, without impairing the quality. (J Histochem Cytochem 45:461-466, 1997)

Key Words: photomicrography, computer imaging

	Introduction

Top Summary Introduction Materials and Methods Results Discussion Literature Cited

In the study of skin histopathology, a scanning view of an entire specimen is very important in understanding of the structure of the affected area in contrast to the surrounding normal area. A fine photomicrograph covering the entire specimen would be very useful for this purpose, but it is very difficult to take such a photograph with a conventional photomicroscopy system. This is why the field of view of the photomicroscope is limited to a square covering approximately one third of an inch, even at the lowest magnification. This is large enough for normal use, but for recording a larger specimen this limitation precludes visualization of the entire object.

Two approaches have been used to obtain largefield photomicrographs. One is a macrophoto technique. A photograph of the entire specimen is taken directly using a camera equipped with a macro lens. This method is simple and easy, but it has a serious weak point. Because the macro lens' depth of focus is very broad in comparison to the photomicroscope, almost the entire thickness of the specimen is recorded on the film, making it difficult to distinguish the fine structure of the specimen. This method is not suitable for many purposes.

Another way to attempt a largefield image is a patchwork of printed micrographs. The large field is first divided into small parts, which are individually recorded on film. Then the micrographs are printed on paper and manually reconstructed to make the large images. In theory, the quality of the product is equal to that of a normal photomicrograph, but in practice it does not provide adequate picture quality. There are bizarre patterns of shading, irregularities of darkness, and obvious seam lines in the reconstructed picture (Figure 1).

View larger version (96K): [in this window] [in a new window]   Figure 1. A largefield photomicrograph constructed by the conventional patchwork method. Irregular shading and seamlines are obvious. Bar = 1 cm.

These problems stem from the fact that a conventional photographic system cannot provide a perfect micrograph. There are always some minor errors, such as peripheral darkness or distortion in the printed images. These are not obvious when the images are viewed individually but become apparent when they are combined.

The problems associated with reconstructing fine micrographs are as follows:

1. Irregular shading in each small micrograph. This appears repeatedly in the reconstructed large micrograph and makes a "dirty" pattern of shading. The causes of this phenomenon are determined by investigating the photomicrographic system (Figure 2). First, imperfect illumination on the object is one of the factors. The brightness at the stage of the microscope should be even at all points. In practice, however, the shape of the light source and the specifications of the lens set influence the distribution of brightness. Second, the power of the lens set in the microscope and/or the enlarger is occasionally not great enough to deliver enough light to the peripheral area of the field. These factors cause irregular shading in the printed images.

   
   

   View larger version (28K): [in this window] [in a new window]   Figure 2. Principles of the photomicrographic system. There are some lens sets that may cause irregular darkness and/or distortion in the image.

2. Distortion of each small micrograph. The small reconstructed images do not fit perfectly with each other. This is because they are somewhat distorted through the lens sets of the microscope and/or the enlarger. The imperfect flatness of the films may be one of the causes. The distortion is usually symmetrical, and is more severe in the peripheral area where the micrographs are joined.

3. Irregularity of darkness among the small micrographs. The darkness of the images is not even, making the seam lines obvious. This irregularity is largely caused by the specifications of the AE (auto exposure) system in the camera. The AE system is designed to provide a constant quantity of light to the film, and the exposure time therefore varies according to the recorded site. This irregularity can be decreased by setting the exposure time to a constant value, but cannot be eliminated completely by this operation. The total amount of light delivered to the film may influence the darkness.

These errors are hard to eliminate by refining the classical photographic process; merely equalizing the darkness would entail much cost and effort. However, these errors can now be compensated for with the aid of computer imaging techniques. We decided to refine this patchwork method with the aid of a computer to improve the micrograph quality.

	Materials and Methods

Top Summary Introduction Materials and Methods Results Discussion Literature Cited

Materials
A specimen of a large lesion, a photomicroscope (Olympus VANOX, Tokyo Japan), a film scanner, a personal computer (Apple Macintosh 8100/100AV), a photographic quality color printer, and image handling software (Adobe Photoshop v. 3.0) were used.

Digitization of the Micrographs
Photomicrographs of the specimen were taken in the scanning mode to cover the entire field (Figure 3). Some overlapping area was kept for reference for later operations. In addition, at least one micrograph with no specimen on the stage and one micrograph recording a blood cell counting plate (or some other object with a cross-hatched pattern) were taken. These images were recorded with a single roll of film whenever possible. If more than two rolls of film were needed, the emulsion numbers of the films were matched to minimize variation.

View larger version (11K): [in this window] [in a new window]   Figure 3. The micrographs of a large specimen are taken in the scanning mode to cover the entire field.

The developed films were scanned with the film scanner at a constant setting to transfer the images into 24-bit color image files. When a film scanner was not available, scanning printed micrographs on a flatbed scanner took the place of this process. The sampling resolution was set referring to the quality and size of the final output. For example, for a final output to a 300 ppi dye sublimination printer with a print area 10 inches long, with six individual fields composing the 10 inches, one field needs to be 500 pixels wide. A 1.4-inch-wide image on 35-mm film can be scanned in at 360 ppi. (In practice, the micrographs are overlapped to some degree, so the final output is smaller than estimated.)

The micrograph files were checked for the rotational distortion that is introduced in the scanning process. This was done with "Photoshop" software. The digitized micrographic image was displayed and magnified enough to see individual pixels. Then the image was checked to see whether its edge was overriding several columns. When the rotation was not negligible, it was compensated for by calling up the function "RotateArbitrary" from the "Image" menu. The angle of the image's rotation was measured by choosing the "Line" tool and drawing a line parallel to the edge of the image. The angle that appeared in the "Info" palette indicated the degree of rotation.

The micrographic images were captured at a constant position and size and were stored as picture files on a hard disk drive for later operations.

Compensation for Irregular Shading in Each Small Image (Figure 4)
The reference image taken with no object on the stage was opened first (Picture A). The function "Histogram" was called from the "Image" menu to check the distribution of the values of each color element in this image (Figure 5). The value at the higher end of the distribution pattern (i.e., the value that would be obtained without shading) was recorded in each color channel.

View larger version (21K): [in this window] [in a new window]   Figure 4. Outline of our compensation method for peripheral darkness. The difference between an ideal blank image and the actual one is calculated for adequate correction.

View larger version (32K): [in this window] [in a new window]   Figure 5. A histogram of the distribution of color values in the picture with no specimen on the stage. The higher end of the histogram is recorded as the reference for correction.

A new picture (Picture B) with the same size as Picture A was created. This picture was filled with the color values that were recorded in the first step (e.g., red = 243, Green = 237, Blue = 231).

Picture A was subtracted from Picture B to extract the components that were needed to compensate for the irregular shading in Picture A, as follows. First, Picture A and Picture B windows were opened, and a new picture (Picture C) of the same size as Picture A was created. Then the "Calculations" command was selected from the "Image" menu to open the dialogue box (Figure 6). Picture B was chosen as "Source 1" and Picture A as "Source 2." "Subtract" was chosen as "Blending," and the result box was set to save the result on Picture C. The channels of all pictures were set to the same color (e.g., red). This procedure was done for all color channels (red, green, and blue) with the same settings, and Picture C was saved on a hard disk drive.

View larger version (48K): [in this window] [in a new window]   Figure 6. Calculation between two images was performed in each of the color channels to extract the components needed for background correction. The results were saved in a new image file.

In all the pictures waiting for reconstruction, their irregularity of darkness was compensated for by adding Picture C. This was done as follows. First, two windows were opened for the picture to be compensated for ("Picture X") and Picture C, and a new picture ("Picture N") with the same size as Picture X was created. Then the "Calculations" command was selected from the "Image" menu to open the dialogue box (Figure 6). Picture X was chosen as Source 1, and Picture C as Source 2. "Add" was chosen as "Blending," and the result box was set to save the result on Picture N. The channels of all pictures were set to the same color (e.g., red). This procedure was done for all color channels (red, green, and blue) with the same settings, and the compensated picture (Picture N) was saved.

Compensation for Distortion
The distortion produced through the photomicrographic system was checked by investigating the shape of the cross-hatched pattern. The reference image recording the cross-hatched pattern was opened first. If the distortion was not negligible, it was compensated for by calling up the function "EffectsDistort" from the "Image" menu. The parameters needed for this operation were obtained from the "Info" palette and were utilized for compensation of all pictures used in the reconstruction.

Equalizing Irregularity of Darkness Among the Small Images
Differences in darkness between adjacent images were checked and adjusted by comparing the darkness of the overlapping area. This operation was done on the basis of naked eye view utilizing the "ImageAdjustBrightness" command, but it can be done by measuring the average value of color elements for accurate compensation. The procedure is as follows. First, select the rectangular selection tool from the toolbox. Open the "Marquee Options" palette and set the style to "Fixed Size." Then place the selection tool on the existing image and set the "Width" and "Height" so as to cover the overlapping area. Call up the "Histogram" function from the "Image" menu and record the average value of each color channel in this area. After that, place the selection tool on the image to be attached and measure the mean values of the area including the same image. Calculate the difference between these values and record them. Select the picture to be attached and open the "Levels" box by choosing "Adjust- Levels" from the "Image" menu, and then select the channel to be compensated for. If increasing the brightness by level n is required, put (255 - n) in the right box of "Input Levels" and n in the left box of "Output Levels." Decreasing the brightness by level n can be done by putting n in the left box of "Input Levels" and (255 - n) in the right box of "Output Levels."

Reconstruction of a Large Image
Compensated images were reconstructed on the monitor screen utilizing the "Layer" function that enabled us to move the images independently and fit them with each other. First, one image was pasted on the background layer of a newly created image and the image to be attached was pasted on another layer with the "Edit-Paste Layer" command. Then the images were moved until they fit perfectly with each other. The two layers were combined to make a single layer by choosing "Merge Layers" from the control menu of the "Layer" palette. After that the next image was pasted on a new layer, and all the images were built up this way to make a large image. The reconstructed image was saved as one image file. The color and darkness of the reconstructed image were adjusted if needed, and the unsharp masking was done to make the picture clear by selecting "Sharpen-Unsharp Mask" from the "Filter" menu. The image was printed with a photographic quality full-color printer or a full-color digital film recorder.

	Results

Top Summary Introduction Materials and Methods Results Discussion Literature Cited

The pattern of shading in our micrographic system is shown in Figure 7. The picture is contrast-enhanced to show the irregularity of darkness visible. This irregularity is also reflected to the wide distribution of color values shown in Figure 5.

View larger version (161K): [in this window] [in a new window]   Figure 7. The contrast-enhanced image of the micrograph taken with no specimen on the stage. Irregularity of light distribution in our photomicrographic system is seen.

The reconstructed image is shown in Figure 8. The entire micrograph of the tumor, as well as the surrounding normal skin, is shown without loss of its fine structure. No seam lines or shading are visible.

View larger version (88K): [in this window] [in a new window]   Figure 8. Micrograph of the same specimen as in Figure 1 made by our modified patchwork method. There is no irregular shading and no seam lines are visible. Bar = 1 cm.

	Discussion

Top Summary Introduction Materials and Methods Results Discussion Literature Cited

Full-color images provide a considerable amount of information. Ten years ago, a huge computer was needed to handle such images, but now personal computers (PCs) have enough performance to process these images (Brown 1994 ; Clements 1994 ). Quantitative analysis and processing of images, including programmed deformation, can be executed with the aid of computer imaging techniques.

The shading that occurred in an actual micrographic system was irregular in degree and distribution (Figure 7). It was impossible to neutralize such shading by simple peripheral brightening or manual retouching. Compensation of the light values of each pixel with an accurate correction table was needed. We obtained such a table by analyzing a blank picture, and we used it to correct the shading.

In contrast to the effort required for compensation of the shading, there was almost no need to correct the distortion in the study with the film scanner. This is due in part to the absence of the enlarging and printing step, in which the images are considerably distorted.

Although our method is relatively easy, great care should be taken in the digitization step because the pictures are compensated by applying the correction table created from the reference image. The position and size of capturing must be the same for the reference image and the images requiring compensation. Errors introduced during the scanning step must also be avoided because it is almost impossible to correct them afterwards. A drum should be used, if possible, to minimize the distortion and color changes at the digitizing step. The flatbed scanner is cheaper and easier to use but has some disadvantages inherent in its design. In addition to the errors introduced by the principles of digitization, the scanning process itself can introduce some distortions. Among these is the rotational distortion that occurs because one cannot place the object to be scanned exactly perpendicular to the scanning axis. This distortion can be reduced with the use of film scanners, but never eliminated.

As a solution to these problems, direct digitization of micrographs through a color charge-coupled device (CCD) camera can be considered as an alternative method. The technology of CCD cameras has progressed enough to produce fine pictures (Brown 1994 ). Direct digital images obviate the need for scanning and eliminate one link in the image chain. This yields a better quality end product and does away with the extra time needed to scan in micrographs.

The patchwork method used in this study was refined enough to provide a high-quality seamless image with the aid of computer imaging techniques. Some of the steps can be automated by making corresponding plug-ins of the software "Photoshop," and we intend to make them in the future. Our method is simple and does not require special hardware or software. No manual retouching is needed in the entire process. The final products are of sufficient quality that fine structures can be visualized. The maximal resolving power and the maximal handling size of the specimen are, in theory, unlimited.

We propose our method as a new way of making widefield photomicrographs.

	Literature Cited

Top Summary Introduction Materials and Methods Results Discussion Literature Cited

Brown S (1994) Digital imaging in clinical photography, Part 1. J Audiovis Media Med 17:53-65

Clements C (1994) Photo CD in practice. J Audiovis Media Med 17:21-22

CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    Technorati    What's this?

This article has been cited by other articles:

				A. C. Johansson, E. Visse, B. Widegren, H.-O. Sjogren, and P. Siesjo Computerized Image Analysis as a Tool to Quantify Infiltrating Leukocytes: A Comparison Between High- and Low-magnification Images J. Histochem. Cytochem., September 1, 2001; 49(9): 1073 - 1080. [Abstract] [Full Text] [PDF]

				H.-A. Lehr, C. M. van der Loos, P. Teeling, and A. M. Gown Complete Chromogen Separation and Analysis in Double Immunohistochemical Stains Using Photoshop-based Image Analysis J. Histochem. Cytochem., January 1, 1999; 47(1): 119 - 126. [Abstract] [Full Text]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Ikeda, I. Articles by Ono, T. Search for Related Content PubMed PubMed Citation Articles by Ikeda, I. Articles by Ono, T. Social Bookmarking                      What's this?

Guidelines | Subscriptions | About | exPRESS - Current - Archive | Business Information | Contact