Paul R. Montague, CRA FOPS
Department of Ophthalmology and Visual Sciences
The Univesity of Iowa Hospitals and Clinics
Iowa City, Iowa
In the second half of the 19th century, elaborate drawings sketched by skilled medical illustrators, artists, and ophthalmologists were required to record the appearance of disorders of the eye. With the advent of photography, a more precise method of documentation became available.
Optical artifact that was not resolved until the late 1960's.
Current developments in computer hardware, electronic cameras, and imaging software have stimulated further evolution toward imaging without film. Digital Imaging is a broad term applied to the recording of images electronically, conversion of those images into a set of numbers, storage of those numbers in a computer, and manipulation with computer programs. Since the images are represented as numeric data, they can be transmitted over phone lines, satellite, or computer networks. All of the photographs on the OPS web pages are in digital form.
This technology provides a wide variety of new and different techniques which have greatly expanded the horizons of ophthalmic photography. The highly sensitive electronic cameras can successfully record the very low infra-red light emissions of Indocyanine Green dye, making this adjunct to fluorescein angiography practical.
Fluorescein and Indocyanine Green images of the same patient.
Photographs by Kirby R. Miller
Comparing Film with Digital Images
Imaging with film and digital imaging are vastly different technologies which may appear on the surface to produce similar results. In order to make meaningful comparisons, it is necessary to have some understanding of each.
How does film work?
Black and white photographic film is made up of a thin, light sensitive emulsion coated on a flexible tri-acetate base. The light sensitive substance within the emulsion is made up of one or more of the silver halides:
There are about 86,400,000,000 sliver halide crystals in a 35mm photographic image. When one of these silver halide crystals is struck by a photon of light, a tiny spec of solid silver is formed. More light creates more specs of solid silver, none of which can be seen by the naked eye. When the film is exposed in a camera, dark areas in the scene produce few silver specs, and lighter areas produce more silver specs, creating a latent image on the film.
Once a latent image has been captured on the film, it is placed in a liquid developer which reduces the silver halide to metallic silver. Because the development reaction is strongly catalyzed by the presence of solid silver the areas of the film which were struck by the greatest amount of light are converted to silver first. The development process is halted by a stop bath before unexposed silver halide can be converted to metallic silver, the remaining silver halide is dissolved away with a fixing bath, and a black and white photographic negative remains.
Film emulsions can be made in different thickness' and with different combinations of halides that make them more sensitive or less sensitive to light. The ISO of the film is a rating of that sensitivity. A film with an ISO of 200 is twice as sensitive to light as a film with an ISO of 100. The ISO is often indicated as part of the film name, as in Ektachrome® 64. The sensitivity of a film to light is referred to as the film speed.
The silver in the processed film tends to clump together into small particles called grain. The grain pattern can be seen when photographs are projected or enlarged. Generally speaking, films with finer grain patterns are capable of resolving more detail than films with coarse grain patterns. Granularity usually increases as film speed increases.
A low speed, ISO 32 film produces this fine grained image. The right image is a small portion of the whole negative on the left.
A high speed, ISO 400 film produces a coarser grain with less apparent sharpness in the image.
One measure of a film's ability to record fine detail is it's resolving power. Film is exposed to a test object composed of alternating light and dark lines of equal width. Resolution is expressed as the number of these line pairs that can be distinguished in 1 millimeter of film. In order for the resolution number to be meaningful, the contrast of the original image must also be stated, since resolution decreases as the subject contrast decreases. A typical resolution might be stated as 125 line pairs / mm at 1000:1.
Although imperfect, resolution measured in line pairs per millimeter can be used in comparisons of film with digital images.
Capturing Digital Images
How Does Digital Imaging Work?
Simply defined, digital imaging is the representation of images as a set of numbers. In practice, it encompasses the electronic capture of images, their conversion to numeric data, the storage and retrieval of those data, and the manipulation, view and printing of the images.
Assigning numbers to tonal values in a black and white photograph is a relatively simple concept. Assume that the brightest white is assigned a value of 255, and that the darkest black is assigned a value of 0. The gray tones between white and black are now divided into 255 equal steps.
The numbers are assigned to the average gray tone in a square area of the image. These areas of gray tone are called picture elements, or pixels. To represent an image in very low resolution, a small number of large pixels would be used.
A photograph is divided into six pixels, producing a digital image so low in resolution that the there is no useful information.
Increasing the number of pixels increases the resolution until very fine detail in the image can be identified. In a black and white image, each pixel is represented by one number. Each number occupies a unit of computer storage called a byte. The greater the number of pixels, the greater the amount of computer storage required to store the image. A typical frame of a digital angiogram is captured at 1024 x 1024 pixels, requiring 1,048,576 bytes of storage space. Only one of these black and white images can be stored on a 3 1/2" floppy disk.
The resolution of a digitized image is increased by increasing the number of pixels in the image.
It should be obvious that images recorded on silver film are constructed very differently from images recorded digitally. Neither lines per millimeter nor modulation transfer functions can be applied easily to digital images, and pixels per millimeter can not be applied to silver images.
Since silver-based imaging has been applied to ophthalmology for many years, it is reasonable to use the fine vessel detail which we obtain through conventional angiography as a de facto "standard" against which digital images can be measured. The question is "How many pixels of resolution must a digital image have in order to offer the same information as a silver image of the same subject?".
The answer to this question is highly subjective. At 512 x 512 pixels, most observers would agree that fine detail in the capillaries defining the avascular zone is lost. But in many cases, sufficient information is captured at this resolution to permit accurate diagnosis and to formulate an effective course of treatment with the noted exception of treatment of parafoveal neovascular membranes.
At 1024 x 1024, all but the finest capillary detail is recorded. This is the resolution employed by most of the companies who make digital imaging equipment for applications in retinal photography.
Resolutions of 2048 x 2048 an beyond are possible, but the size of the data files and the cost of the equipment becomes prohibitive. A black and white image at 1024 pixels requires about one million characters of storage space. At 2048, the same image would require four million characters of space.
Digital Images from Video Cameras
One common means of image capture is through the use of a standard video camera. In the United States, video images are made up of 525 horizontal scan lines as defined by NTSC. A narrow beam is passed across the subject from left to right, constructing one scan line. That process is repeated until all 525 lines have been scanned. The final picture is acutally made up of two sets of interlaced pictures, but discussion of interlacing complicates the model and adds nothing to the basic concept.
As the beam passes along one scan line, it produces voltage which varies with the brightness of the image. The beam produces higher voltage as it passes over bright objects, and lower voltage as it passes over darker ones. This electrical signal is called analog output, which can be used to produce a picture on a conventional television.
A computer component called an Analog to Digital Converter is used to change these voltages into numbers, creating a digital image. If a value of 0 is assigned to the lowest voltage produced representing the darkest area of the image, and a value of 255 is assigned to the highest voltage, then the numbers from 1 to 254 can be assigned to intermediate voltages representing shades of gray from black to white. This is a 256 level gray scale.
For the sake of simplicity, assume that it takes 1 second for the beam to scan one line (The time is actually much less). If the A/D board averaged the voltages produced for the first 1/10th second and assigned a gray value, then the next 1/10th second, continuing to the end of the line, that line would have been divided into ten separate numbers. A sample time of 1/20th second produces 20 values per line. A sample time of 1/100th second produces 100 values per line. The faster the sampling is done, the more horizontal resolution there is in the digitized image. The vertical resolution is fixed by the number of scan lines at 525.
Digital Images from CCD Cameras
A Charge Coupled Device (CCD) works differently from a standard video camera. The CCD consists of a matrix of capacitors which can store an electrical charge sandwiched with a light sensitive photo cell. When an image is focused on this matrix, each square will produce a voltage relating to the average brightness of the light falling on the square. The output is analog, just as it was with the video camera, but the time slicing is not necessary. CCD cameras are often used on digital fundus cameras.
Another kind of CCD device uses only a row of light sensitive devices instead of a 2 dimensional matrix. In order to capture an entire image, the top of the subject is presented to the device and the first row is captured. The device is then moved down slightly, and the second row is captured. Then the third, and forth rows, and so on, until the entire subject has been scanned. This method is considerably slower than the CCD matrix, but the cost of the device is much lower. This type of CCD device is most often used in slide and flat-bed scanners.
Black and white digital images are images which have been divided into small square picture elements called pixels, each pixel being assigned a numeric value based upon its average gray level.
The most common way of viewing digital images is to display them on a computer monitor. The computer reads the image file. The first part of the file might contain information like the number of pixels in the image, whether it is black and white or color, and the method of image compression used, if any. After the file header is read, the program knows how to interpret the actual image data.
For an uncompressed black and white image, the value for a pixel is read from the image file, and the imaging software instructs the monitor to display a gray square on the screen with it's shade of gray depending upon the numeric value. A value of 0 will produce black, a value of 255 will produce white, and intermediate numbers will produce shades of gray.
Image displayed at 173 x 196 pixels and an enlargement of the left
eye revealing the gray pixels which make up the image.
The resolution that can be achieved is dependent upon the type of monitor, the type of video adapter card in the computer, and the amount of video memory installed. A high resolution Ultra VGA monitor can display up to 1024 x 1280 pixels.
The number of pixels in an image is not the only factor affecting image quality. Due to reduced video memory, some computer video modes restrict the levels of gray that can be displayed. The number of gray levels is referred to as the depth of the image. An image with pleasing and fairly accurate display characteristics at 256 levels of gray may appear uneven at 16 levels of gray. In the extreme case, only two levels are displayed, rendering the image in pure black and white with no levels of gray.
256 level image demonstrates good tonal transitions.
16 level image demonstrates harsher tonal transitions.
2 level image renders the image as pure black and white
Enlargement emphasizes the effects of reduced gray levels.
The effects of reduced depth may not appear significant in subjects of very high contrast as demonstrated in this fluorescein angiogram digitized at 1024 x 1024 pixels.
Enhancing Digital Images
In ophthalmology, retinal images are used in making a diagnosis, in determining a course of treatment, and in following changes over time. This scientific application requires standardization in the way images are acquired, processed, and presented to ensure reliability of interpretation.
Enhancement of silver photographic images is a matter of routine. Image contrast can be controlled through the choice of developers and printing papers, tone can be altered, and small parts of an image can be made darker or lighter during the printing process.
Digital imaging provides a much more extensive collection of image manipulation possibilities. Since digital images are nothing more than a set of numbers, corrections can be made based on pixel statistics and enhancements can be standardized so that all images are enhanced in exactly the same way.
Before considering image enhancement, it is important to realize that the original image, obtained before any enhancement is performed, contains the most accurate information available. Alterations to that original image are performed to make characteristics of that image more apparent to human observers. Resolution is never increased through the enhancement process, and care must be taken not to enhance an image to the extent that features are identified which did not exist in the original subject.
With a film image, the original negative always contains more information than a positive image made from the negative. This loss is due to imperfections in the photographic process. With digital imaging, photographs can be changed at will from negative to positive with no loss of information.
There appears to be reduced contrast in the unaltered image. There are no areas which are pure white or pure black. Of the total available range of 256 gray values, only values between 40 and 160 are present. The contrast can be stretched, moving the 160 values up to 255. The result is an image with higher contrast.
Further enhancement is possible by stretching the low value of 40 down to 0. It is possible to apply this technique to the point where near-white areas in the image become white, and near-black areas become black. At this point, image information is being lost.
A more sophisticated enhancement technique can make the image appear sharper. Let's say a white book lying on a black desk is photographed. If rendered perfectly, the edge of the paper would appear as a line of white pixels next to a line of black pixels. If the image were out of focus, examination of the edge would reveal a smooth transition from white to black several pixels wide.
White on black image.
In-focus and out-of-focus enlargements of one edge.
A computer program can be employed that steepens the transition between dark and light areas in the image. The effect is that edge appears to be sharp in the enhanced photograph. Careful examination revels that the sharp delineation in the original has not been exactly reproduced.
Out-of-focus image after sharpening.
These techniques can be applied to fundus photographs. The observer must realize that the enhanced image will not exactly match one which was precisely in focus at the start.
The amount of sharpening can be varied. A program might offer 6 levels of sharpening, level 1 being minimal and level 6 being maximum. One difficulty with sharpening and edge detection programs is that they can be applied to the extent that they identify and display edges which do not exist in the original image.
Manufacturers of digital fundus cameras are keenly aware of the possibility of over-enhancement. The original image is always stored, without enhancement, so it can be recalled at any time. The enhancement tools which are provided with these instruments are chosen to allow manipulation within conservative limits. It is still important that the observer pay close attention to the enhanced image, comparing it with the original, to confirm that the results are reasonable.
Ethics of Altering Images
The use of digital imaging programs to enhance ophthalmic images has been described in Enhancing Digital Images. Manufacturers of ophthalmic imaging equipment limit computer programs to those which will be safely useful in manipulating medical images. However, images collected by diagnostic equipment can be exported and altered by a vast collection of commercially available imaging programs which were designed to give the user maximum flexibility in image alteration. This kind of image processing leads to a number of ethical questions surrounding digital imaging.
Assume that a patient is seen at a baseline visit, appropriate drawings and charting are performed, but retinal photographs are not obtained. On follow-up, a highly unusual change occurs. Retinal photography is performed, but the lack of baseline photographs greatly reduces the chances that the case can be used for teaching or publication. Is it possible to make alterations to the follow-up photographs to illustrate the appearance of the retina at baseline?
The original fluorescein angiogram on the left was modified with a digital darkroom program, removing some of the retinal vascular abnormalities. In concept, this is no different than preparing an illustration of the retinal appearance at baseline. But the two images appear so similar that they might both be considered photographic documentation, even if a proper disclaimer were made.
Digital imaging presents the field of ophthalmic photography with a wide variety of powerful tools that can be applied easily to aid in the diagnostic and treatment process, to standardize documentation, and to prepare teaching and publication materials. The application of these tools requires scrutiny and care to preserve the scientific integrity of the images we produce.