Probably, today almost every user imagines the basic principle of storing and displaying graphic information on a computer. Nevertheless, let's say a few words about this so that the subsequent information about digital video (which is a dynamically changing sequence of images) would be clearer for us.

At first glance, a high-quality drawing, being displayed on the screen of a good monitor, differs little from ordinary photography. However, at the level of image presentation, this difference is enormous. While a photographic image is created at the molecular level (that is, its constituent elements are fundamentally indistinguishable by human vision, regardless of magnification), drawings on the monitor screen (and, we emphasize, in the computer's memory) are formed thanks to pixels (or pixels) - the elementary components of the image (most often) of a rectangular shape. Each pixel has its own specific color, however, due to their small size, individual pixels are (almost or completely) indistinguishable to the eye, and a large cluster of them creates the illusion of a continuous image for a person viewing the picture on a monitor screen (Fig. 1.2).

Note
Images on computer screens are formed using square pixels. Unlike computers, many television standards use rectangular rather than square pixels. The parameter characterizing the ratio of pixel sizes is the ratio of their horizontal and vertical sizes, or the pixel proportion ( pixel aspect ratio). You can learn more about this characteristic in lesson 4.

Figure: 1.2... Computer images are formed by pixels

Each pixel (by the way, the word pixel formed from the first two letters of English words picture element) represents information about some "average" intensity and color of the corresponding area of \u200b\u200bthe image. The total number of pixels that represent a picture determines its resolution. The more pixels create an image, the more natural it is perceived by the human eye, the higher its resolution is said to be (Fig. 1.3). Thus, the limit of the "quality" of a computer drawing is the size of the pixels that form it. The details of the computer drawing, smaller than pixels, are completely lost and, in principle, unrecoverable. If we look at such a drawing with a magnifying glass, then, as we zoom in, we will see only a blurry cluster of pixels (see Fig. 1.2), and not small details, as it would be in the case of a high-quality photograph.

Figure: 1.3... The total number of pixels (resolution) determines the image quality

It is worth making a reservation here that, firstly, we mean traditional (analog, not digital) photography (since the principle of digital photography is exactly the same as the discussed principle of forming an image from pixels), and secondly , even for her, speaking of image quality, one should always remember about the technology of photography itself. After all, the image on the film appears due to the passage of light through the camera lens, and its quality (in particular, the clarity and discrimination of small details) directly depends on the quality of the optics. Therefore, strictly speaking, the "infinite" clarity of the traditional photographic image, which we talked about, is somewhat of an exaggeration.

Note
In fact, modern digital cameras allow you to capture an image, the resolution of which is practically not inferior to analog (in the sense that it is now possible to digitize such a number of pixels that will "overlap" the resolution boundaries of the optics themselves). However, for the subject of our book, this fact does not play an important role, since currently digital video in the overwhelming majority of cases is transmitted precisely with a low resolution (relatively small total number of pixels) and it is simply necessary to take into account such a parameter as resolution.

So, simplifying a little, in order to digitize the drawing, you need to cover it with a rectangular grid of size MxN (M dots horizontally and N vertically). This is a combination of numbers MxN (e.g. 320x240, 800x600, etc.) and is called a resolution ( resolution) image, or frame size ( frame size). Then the data about the structure of the image within each pixel should be averaged and the corresponding information about each of the MxN pixels of the image should be written into the graphic file. For a color image, this will be information about the specific color of each pixel (the computer representation of color is described below in this section), and for black and white images, this will be information about the intensity of black. To explain a few more important parameters of the computer representation of images, let us dwell a little more on their last type - drawings made in shades of gray ( grayscale), i.e. in gradation from white to black.

Medical Radiology (TMP) technologists typically perform numerous computer manipulations to refine diagnostic images to aid in correct interpretation. While experienced technologists are usually aware of the visual implications of their manipulations, they cannot fully understand the mathematical and scientific principles behind one-click action. The principles can be complex for all but the most technologically savvy TMPs. In all likelihood, the mathematical processing of images in textbooks and articles intimidates, hinders, or is possibly uninteresting TMP. However, by overcoming resistance and understanding the underlying principles behind image processing, TMPs can enhance their ability to produce high quality diagnostic images.

Mathematics cannot be left out of the discussion on image processing and filtering. This article will describe the principles behind a number of general procedures. This description should be acceptable to technologists of various levels of mathematical knowledge. The first procedures to be discussed are simple procedures related to static images. Further, more complex procedures related to dynamic images. Much of the image processing and filtering occurs with physiologically gated images and SPECT (single photon emission computed tomography) images. Unfortunately, the complexity of these questions does not give detailed description here.

Static image processing

Still images that have been transferred directly to film in real time are presented in analog format. This data can have an infinite range of values \u200b\u200band can produce images that accurately reflect the distribution of radionuclides in organs and tissues. While these images can be of very high quality if captured correctly, real-time collection of information provides only one opportunity to acquire data. Due to human error or other errors, it may be necessary to repeat the acquisition of the image and, in some cases, to repeat the whole examinations.

Still images transferred to a computer for storage or enhancement are presented in digital format. This is done electronically with an analog-to-digital converter. In older cameras, this transformation took place through a series of resistor networks that contain the signal strengths from several photomultiplier tubes and produce a digital signal proportional to the radiation energy of the events.

Regardless of the method used to digitize images, the digital output assigns a discrete value to the processed analog data. The result is images that can be stored and processed. However, these images are only approximations of the original analog data. As you can see in Figure 1, the digital representation is approximate, but does not duplicate analog signals.

Figure 1 - Analog curve and its digital representation

The digital images of radiological medicine are composed of a matrix selected by the technologist. Some common matrices used in radiological medicine are 64x64, 128x128 and 256x256. In the case of a 64x64 matrix, the computer screen is divided into 64 cells horizontally and 64 vertically. Each square as a result of this division is called a pixel. Each pixel can contain limited quantity data. In a 64x64 matrix, there will be a total of 4096 pixels on the computer screen, a 128x128 matrix gives 16384 pixels, and a 256x256 matrix gives 65536 pixels.

High pixel count images are more like the original analog data. However, this means that the computer must store and process more data, which requires more hard disk space and higher demands on random access memory... Most static images obtained for visual inspection by a radiological medicine physician, so they usually do not require significant statistical or numerical analysis. A number of common static imaging techniques are commonly used for clinical purposes. These techniques are not necessarily unique to static image processing, and can be used in some applications for dynamic, physiologically gated, or SPECT images. These are the following methods:

Image scaling;

Subtraction of the background;

Smoothing / filtering;

Digital subtraction;

Normalization;

Profile picture.

Image scaling

When viewing digital images for visual inspection or for recording images, the technologist must select the correct image scaling. Image scaling can be done either in black and white with intermediate shades of gray or in color. The simplest gray scale would be a scale with two shades of gray, namely white and black. In this case, if the pixel value exceeds the value set by the user, a black dot will appear on the screen, if the value is less, then white (or transparent in the case of X-ray images). This scale can be inverted at the user's discretion.

The most commonly used scale is 16, 32 or 64 shades of gray. In these cases, the pixels containing the most full information look like dark shadows (black). Pixels containing the least information appear as the lightest shades (transparent). All other pixels will appear as grayscale based on the amount of information they contain. The relationship between the number of points and shades of gray can be defined linearly, logarithmically, or exponentially. It is important to choose the right shade of gray. If too many shades of gray are selected, the image may appear washed out. If it is too small, the image may look too dark (Fig. 2).

Figure 2 - (A) images with many grayscale, (B) image with few grayscale, (C) image with correct grayscale

The color format can be used to scale the image, in which case the process is the same as the gray scale manipulation. However, instead of displaying the data in grayscale, the data is displayed in different colors depending on the amount of information contained in the pixel. While color images are attractive to beginners and more descriptive for public relations purposes, color images add little to the film's interpretability. Therefore, many doctors still prefer to view grayscale images.

Background subtraction

There are numerous unwanted factors in radiological medicine images: background, Compton scatter, and noise. These factors are unusual in radiological medicine in relation to the localization of radiopharmaceuticals within a single organ or tissue.

Such anomalous values \u200b\u200b(counts) contribute significantly to image degradation. Samples collected from overlapping and overlapping sources are the background. Compton spread is caused by a photon deviating from its path. If the photon has been deflected from the gamma camera, or has lost enough energy to be distinguishable by the electronics camera, this is not so important. However, there are times when a photon is deflected towards the camera and its energy loss can be large enough for the camera to detect as a scatter. Under these conditions, the Compton scatter can be recorded by the camera, which originated from sources other than the areas of interest. Noise is a random fluctuation in an electronic system. Under normal circumstances, noise does not contribute to unwanted emissions to the same extent as background and Compton scatter. However, like background and Compton scatter, noise can degrade image quality. This can be especially problematic for research in which quantitative analysis plays an important role in the final interpretation of the research. Background problems, Compton scatter, and noise can be minimized through a process known as background subtraction. Typically, the technologist uses an area of \u200b\u200binterest (ROI) suitable for background subtraction, but in some cases, the area of \u200b\u200binterest is computer generated (Figure 3).

Figure 3 - Image of the heart. Demonstration of correct placement of background ROI subtraction (arrow)

Regardless of the method, the technologist is responsible for the correct placement of the ROI background. The background of regions with a higher number of regions can capture too many parameters from the organ or tissue in the region of interest. On the other hand, the background of regions with extremely low areas will remove too few parameters from the image. Both errors can lead to misinterpretation of the study.

Background subtraction is determined by adding the number of samples in background ROI and division by the number of pixels that the background ROI contains. The resulting number is then subtracted from each pixel in the organ or tissue. For example, suppose the ROI background is 45 pixels and contains 630 samples. Average background:

630 counts / 45 pixels \u003d 14 counts / pixel

Anti-aliasing / filtering

The purpose of anti-aliasing is to reduce noise and improve the visual quality of an image. Anti-aliasing is often called filtering. There are two types of filters that can be useful in radiation medicine: spatial and temporal. Spatial filters are applied to both static and dynamic images, while temporal filters are applied only to dynamic images.

In the very simple method anti-aliasing uses a 3-x-3 pixel square (nine in total) and also determines the value in each pixel. The pixel values \u200b\u200bin the square are averaged, and this value is assigned to the center pixel (Fig. 4). At the discretion of the technologist, the same operation can be repeated for the entire computer screen or a limited area. Similar operations can be performed from 5-x-5 or 7-x-7 squares.

Figure 4 - 9-pixel simple anti-aliasing circuit

A similar but more complex operation involves creating a filter kernel by weighting the pixel values \u200b\u200bthat surround the center pixel. Each pixel is multiplied by the corresponding weighted values. Next, the values \u200b\u200bof the filter kernel are summed up. Finally, the sum of the filter kernel values \u200b\u200bis divided by the sum of the weighted values, and the value is assigned to the center pixel (Figure 5).

Figure 5 - 9-pixel anti-aliasing circuit with a weighted filter kernel

The disadvantage is that with anti-aliasing, although the image may be more visually appealing, the image may be blurry and there is a loss in image resolution. An end use of the filter kernel includes weighting with negative values \u200b\u200balong peripheral pixels with a positive value at the center of the pixel. This weighting method tends to increase the amount of discrepancy between adjacent pixels and can be used to increase the likelihood of detecting organ or tissue boundaries.

Digital subtraction and normalization

A common problem in radiological medicine is to prevent ongoing activity from hiding or masking abnormal areas of tracer accumulation. Many of these difficulties have been overcome through the use of SPECT technology. However, smarter methods are needed to get relevant information from a flat image. One of these methods is digital subtraction. Digital subtraction involves subtracting one image from another. It is based on the premise that some radiopharmaceuticals are localized in normal and abnormal tissues, making it difficult for the clinician to interpret correctly. To help differentiate between normal and abnormal tissue, a second radiopharmaceutical is administered only within healthy tissue. The image of the distribution of the second radiopharmaceutical is subtracted from the image of the first, leaving only the image of the abnormal tissue. It is imperative that the patient remains motionless between the first and second injection.

When the technologist subtracts the high-quality second image from the low-quality first image, sufficient values \u200b\u200bcan be removed from the abnormal tissue to make it appear “normal” (Figure 6).

Figure 6 - Digital subtraction without normalization

Images must be normalized to avoid false negative results. Normalization is a mathematical process in which scattered samples between two images are matched. To normalize the image, the technologist needs to isolate a small area of \u200b\u200binterest near the tissue that is considered normal. The number of counts in a region in the first image (with a low number) is divided into graphs in the same region of the second (with a high number). This will give a multiplication factor, counting all the pixels that make up the first image. In Figure 7, the "normal zone", in the calculation, this will be the upper left pixel. This number in the "normal area" (2) divided by the corresponding pixel of the second image (40) gives a multiplication factor of 20. All pixels in the first image are then multiplied by a factor of 20. Finally, the second image will be subtracted from the number in the first image.

Figure 7 - Background subtraction with normalization

Profiling image

Image profiling is a simple procedure that is used to quantify various parameters in a static image. To profile the image, the technologist opens the appropriate application on the computer and positions the line on the computer screen. The computer will consider the pixels indicated by the line and plot the number of samples contained in the pixels. The profile picture has several uses. For static myocardial perfusion studies, a profile is taken through the myocardium to aid in determining the degree of myocardial perfusion (Figure 8). In the case of examination of the sacroiliac region, the profile is used to assess the uniformity of bone absorption of the sacroiliac joint agent in the image. Finally, the image profiles can be used as a control to analyze camera contrast.

Figure 8 - Myocardial profile picture

Dynamic image processing

A dynamic image is a collection of static images taken sequentially. Thus, the previous discussion on the composition of analog and digital still images is applicable to dynamic images. Dynamic images obtained in digital format consist of matrices chosen by the technologist, but, as a rule, these are matrices of 64-x-64 or 128-x-128 sizes. While these matrices can compromise image resolution, they require significantly less memory and RAM than 256 x 256 matrices.

Dynamic images used to assess the rate of accumulation and / or the rate of excretion of RFP from organs and tissues. Some procedures, such as a three-phase scan of bone and gastrointestinal bleeding, require only a visual examination by a physician to make a diagnostic conclusion. Other tests, such as nephrogram (Fig. 9), gastric emptying studies, and hepatobiliary ejection fraction, require quantification as part of the physician's diagnosis.

This section discusses a number of general dynamic image processing techniques used in clinical practice. These techniques are not necessarily unique to dynamic imaging, and some will have applications for physiologically gated or SPECT images. These are the methods:

Summation / addition of images;

Time filter;

Activity time curves;

Image stacking / addition

Image summation and addition are interchangeable terms that refer to the same process. This article will use the term image stacking. Image summation is the process of summing the values \u200b\u200bof multiple images. Although there may be circumstances in which the stacked images are quantitative, this is more the exception than the rule. Because the image stacking reason is rarely used for quantitative purposes, it is not worth performing image stacking normalization.

Study images can be summarized either partially or completely to obtain a single image. An alternative method involves compressing the dynamic image into fewer frames. Regardless of the method used, the main advantage of image stacking is cosmetic. For example, sequential images with a low number of examinations will be stacked to visualize the organ or tissue of interest. Obviously, further processing of images of visualization of organs and tissues will be facilitated by the technologist, which will help the doctor in the visual interpretation of the research results (Fig. 9).

Figure 9 - (A) nephrogram before and (B) after summation

Temporary filtering

The purpose of filtering is to reduce noise and improve the visual quality of the image. Spatial filtering, often known as anti-aliasing, is applied to static images. However, since dynamic images are sequentially located static images, it is advisable to apply spatial filters for dynamic ones as well.

Various types of filters, time filter, applied for dynamic studies. Pixels in consecutive dynamic analysis frames are unlikely to experience huge fluctuations in accumulated samples. However, small changes in one frame from the previous one can lead to "flickering". Timing filters successfully reduce flicker while minimizing significant statistical fluctuations in the data. These filters use a weighted average technique in which a pixel is assigned a weighted average of identical pixels from the previous and subsequent frames.

Activity time curves

The quantitative use of dynamic imaging to assess the rate of accumulation and / or the rate of elimination of RP from organs or tissues is ultimately related to the activity time curve. Activity time curves are used to show how the counts in the area of \u200b\u200binterest will change over time. Physicians may be interested in the rate of collection and elimination of samples (eg, nephrogram), the rate of excretion (eg, hepatobiliary ejection fraction, gastric emptying), or simply the change calculated over time (eg, radioisotope ventriculography).

Regardless of the procedure, the activity time curves begin with the ROI around an organ or tissue. The technologist can use a light pen or mouse to draw the ROI. However, there are some computer programsthat automatically make selections through contour analysis. The low amount of research can be a problem for technologists, as organs and tissues can be difficult to understand. Proper ROI isolation may be required by the technologist to summarize or compress until the boundaries of the organ or tissue are readily distinguishable. For some studies, the ROI will remain the same throughout the study (eg, nephrogram), while in other studies, the ROI may vary in size, shape, and location (eg, gastric emptying). In quantitative studies, it is imperative that the background be corrected.

Once counted, the ROI is determined for each frame and the background is subtracted from each image, usually to plot the data over time along the X axis and computed along the Y axis (Figure 10).

Figure 10 - Simulation of the activity time curve

As a result, the time curve will be visually and numerically comparable to the established norm for each specific study. In almost all cases, the rate of accumulation or excretion, as well as the overall shape of the curve from the normal study, are used for comparison to determine the final interpretation of the study results.

Conclusion

A number of procedures that apply to static rendering can also be applied to dynamic rendering. The similarity is due to the fact that dynamic images are a sequential series of static images. However, the number of dynamic procedures does not have static equivalents. Some manipulations of static and dynamic images are not quantitative. Many procedures are aimed at improving the image image. However, the lack of quantitative results does not make the procedure less important. This suggests that a picture is worth a thousand words. Besides, high quality, computer enhancement of diagnostic images, through correct interpretation, can play a role in improving the quality of human life.

List of used literature

1. Bernier D, Christian P, Langan J. Nuclear Medicine: Technology and Techniques. 4th ed. St. Louis, Missouri: Mosby; 1997: 69.
2. Early P, Sodee D. Principles and Practices of Nuclear Medicine. St. Louis, Missouri: Mosby; 1995: 231.
3. Mettler F, Guiberteau M. Essentials of Nuclear Medicine Imaging, 3rd ed. Philadelphia, Penn: W.B. Saunders; 1991: 49.
4. Powsner R, Powsner E. Essentials of Nuclear Medicine Physics. Malden, Mass .: Blackwell Science; 1998: 118-120.
5. Faber T, Folks R. Computer processing methods for nuclear medicine images. J Nucl Med Technol. 1994; 22: 145-62.

Alphanumeric characters (BCS) and texts

BCS are the most important component of presentation images, therefore special attention should be paid to their implementation. Scientific research has proven that the accuracy and speed of reading these symbols from the screen depends on their style and visual conditions of observation.

First factorone thing to consider is the placement of the image field on the screen. The dimensions of the screen itself can be determined by the optics setting, which provides a uniform acceptable resolution over the entire screen area without distortion around the edges. Labels, texts and other important information should be placed within "Safe" image area, the boundaries of which are 5-10% of the corresponding linear size from the edges of the screen. Therefore, the most important text should be placed in the center of the screen.

Secondly, in the production of type headings, introductory and explanatory titles, an orderly and balanced arrangement of the text of the splash should be sought, taking into account the experience of broadcast television. At the same time, hyphenation of words is extremely undesirable in the credits. It is possible to use direct and reverse contrast, namely - dark BCS on a light background, and vice versa in the second. When the room is well lit, it is better to use direct contrast, and in low light, the opposite. The change of contrasts during the demonstration should not be frequent, which tires the eyesight, but the judicious use of this technique can contribute to the development of a certain dynamics of the presentation, break its monotony.

When using colored symbols, it is necessary to consider their combination. However, in any case, the background of the inscription should not have a saturated bright color.

Psychologists have experimentally established the presence of "edge effects", which consist in the fact that characters at the ends of a string (or even single ones) are recognized faster and more accurately than characters inside a string, and the string is read faster if it is isolated. This suggests that the text, consisting of several lines, should be increased in height of the letters, and short single inscriptions should be made out with a typical font applied to the entire presentation style.

Static images

The effectiveness of a particular type of graphic construction depends on the choice of form elements and their organization. The wrong choice of elements, poverty or an excessive variety of the alphabet of pictorial means reduce the information content of the illustrations.

In a graphic message, as in any other, semantic and aesthetic parts can be distinguished. When they are displayed on the screen, of course, semantic accuracy must be ensured, which determines the error-free reading of information.

The aesthetics of illustrations also deserves the closest attention, as it affects the speed of reading and creates a positive emotional background that contributes to the successful perception and assimilation of information. This is especially important where the quality of homemade illustrations is not yet very high.