SECTION I - CHAPTER 2 DIGITAL IMAGING PROCESSING CONCEPTS

RADT 3463 - COMPUTERIZED IMAGING Section I: Chapter 2 RADT 3463 Computerized Imaging 1 SECTION I - CHAPTER 2 DIGITAL IMAGING PROCESSING CONCEPTS RADT 3463 COMPUTERIZED IMAGING Section I: Chapter 2 RADT 3463 Computerized Imaging 2 Section I: Chapter 2 RADT 3463 Computerized Imaging 3 1

NASA NASA used digital image processing to manipulate images beamed back to Earth to improve visualization of the surface of the moon. Benefits from the space program's research were applied to other areas such as photography, biology, forensics, defense, remote sensing, and medicine, including medical imaging. Section I: Chapter 2 RADT 3463 Computerized Imaging 4 Benefits of NASA http://techtran.msfc.nasa.gov/at_home.html Section I: Chapter 2 RADT 3463 Computerized Imaging 5 IMAGE FORMATION AND REPRESENTATION Think of images as a subset of all objects and that image set contains subsets within it, such as visible and invisible images, optical images, and mathematical images. (Castleman 1994) Section I: Chapter 2 RADT 3463 Computerized Imaging 6 2

IMAGE FORMATION AND REPRESENTATION This is what we see (visible image) This is what the computer see (invisible image) Section I: Chapter 2 RADT 3463 Computerized Imaging 7 IMAGE FORMATION AND REPRESENTATION Section I: Chapter 2 RADT 3463 Computerized Imaging 8 ANALOG IMAGES If the chest image is scanned from left to right using a light source (positioned in front of the image) and a photomultiplier tube (PMT-positioned behind the image) to detect the transmitted light, the light intensity will change continuously with respect to dark and bright spots on the image. Section I: Chapter 2 RADT 3463 Computerized Imaging 9 3

ANALOG IMAGES The photomultiplier tube generates an output signal The signal intensity varies continuously depending on the location of the light on the image This signal is called an analog signal An analog signal represents the image scanned by the light source and the PMT. Section I: Chapter 2 RADT 3463 Computerized Imaging 10 ANALOG IMAGES The chest image is called an analog image It is generated from a continuous function In radiology, images displayed on analog monitors for viewing and interpretation are analog images. Section I: Chapter 2 RADT 3463 Computerized Imaging 11 DIGITAL IMAGES A digital image is a numerical representation of patient's data. Section I: Chapter 2 RADT 3463 Computerized Imaging 12 4

DIGITAL IMAGES The ADC change the continuous analog signal into discrete digital data. To generate a digital image, a computer requires discrete data (0 and 1) for operation. Section I: Chapter 2 RADT 3463 Computerized Imaging 13 DIGITAL IMAGES The result is a digital image. The digital image is converted into a physical image, an analog image for human viewing. Section I: Chapter 2 RADT 3463 Computerized Imaging 14 IMAGE DOMAINS The digital medical images can be represented in two domains, based on how they are acquired: 1. the spatial location domain 2. the spatial frequency domain. Section I: Chapter 2 RADT 3463 Computerized Imaging 15 5

IMAGE DOMAINS Spatial Location Domain The digital/numerical image is arranged so the location of each number assigned to a pixel in the image can be identified using an X- Y coordinate system. X-axis describes the pixel s horizontal location Y-axis describes the pixel s vertical location Section I: Chapter 2 RADT 3463 Computerized Imaging 16 IMAGE DOMAINS Spatial Frequency Domain Frequency refers to the number of cycles per unit length; the number of times a signal changes per unit length. Small structures produce high frequencies that represent the detail in the image Large structures produce low frequencies and represent contrast information in the image. In digital radiography, Fourier Transform (FT) can transform one image domain into another image domain. Section I: Chapter 2 RADT 3463 Computerized Imaging 17 IMAGE DOMAINS Spatial Frequency Domain WHY? A reason to view the image in frequency domain is to enhance sharpness (suppress low frequency) or to smooth (suppress high frequency) the image. Section I: Chapter 2 RADT 3463 Computerized Imaging 18 6

IMAGE DOMAINS Spatial Frequency Domain The inverse Fourier Transform is used to transform an image in the frequency domain back to the spatial location domain when it needs to be viewed for interpretation. Section I: Chapter 2 RADT 3463 Computerized Imaging 19 Section I: Chapter 2 RADT 3463 Computerized Imaging 20 Contrast Low Freq Section I: Chapter 2 RADT 3463 Computerized Imaging 21 7

Detail High Freq Section I: Chapter 2 RADT 3463 Computerized Imaging 22 CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Five fundamental classes of operations can be used in digital image processing 1. Image Enhancement 2. Image Restoration 3. Image Analysis 4. Image Compression 5. Image Synthesis Section I: Chapter 2 RADT 3463 Computerized Imaging 23 CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Enhancement Certain characteristics such as contours and shapes can be enhanced to improve the overall quality of the image. The operations include: Contrast Enhancement Edge Enhancement Spatial And Frequency Filtering, Image Combining Noise Reduction Section I: Chapter 2 RADT 3463 Computerized Imaging 24 8

CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Restoration Designed to improve the quality of images that have distortions or degradations. Common in spacecraft imagery. Images sent to Earth from camera systems on spacecrafts suffer distortions/ degradations that must be corrected for proper viewing. Blurred images, for example, can be filtered to make them sharper. Section I: Chapter 2 RADT 3463 Computerized Imaging 25 Face on Mars Section I: Chapter 2 RADT 3463 Computerized Imaging 26 CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Analysis Allows measurements and statistics to be performed, as well as image segmentation, feature extraction, and classification of objects. Section I: Chapter 2 RADT 3463 Computerized Imaging 27 9

CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Analysis (continued) The object measures are used to classify the objects into specific categories. Segmentation operations are used in 3D medical imaging. Section I: Chapter 2 RADT 3463 Computerized Imaging 28 CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Compression Used to reduce the image size image to decrease transmission time or reduce storage space Two forms of image compression are lossy and lossless compression Section I: Chapter 2 RADT 3463 Computerized Imaging 29 CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Compression (continued) Lossless compression no information is lost when the image is decompressed. Lossy compression - some loss of image details when the image is decompressed. Used exact original image detail is necessary Section I: Chapter 2 RADT 3463 Computerized Imaging 30 10

CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Synthesis Used to create images from other images or non-image data. Used when a desired image is either physically impossible or impractical to acquire, or does not exist in a physical form at all Reconstruction techniques are the basis for the production of CT and MR images, and 3D visualization techniques, which are based on computer graphics technology. Section I: Chapter 2 RADT 3463 Computerized Imaging 31 CLASSES OF DIGITAL IMAGE PROCESSING OPERATIONS Image Synthesis (continued) - Figure 2-5 textbook Section I: Chapter 2 RADT 3463 Computerized Imaging 32 CHARACTERISTICS OF THE DIGITAL IMAGE Matrix Pixels Voxels Bit Depth Appearance of digital Images Section I: Chapter 2 RADT 3463 Computerized Imaging 33 11

MATRIX A digital image is displayed as a combination of rows and columns make up of a 2D array of numbers called a matrix. The matrix consists of columns (M) and rows (N) that define small square regions called picture elements, or pixels. The dimension of the image is described by M, N and the size of the image is given by the relationship M x N x K bits Section I: Chapter 2 RADT 3463 Computerized Imaging 34 MATRIX When M = N, the image is square. The radiographer selects the matrix size for the examination, sometimes referred to as the fieldof-view (FOV). Larger images require more processing time and more storage space, as well more time to be transmitted to remote locations. Section I: Chapter 2 RADT 3463 Computerized Imaging 35 MATRIX Digital Imaging Modality Matrix Size and Typical Bit Depth Nuclear Medicine 128 x 128 x 12 Magnetic Resonance Imaging 256 x 256 x 16 Computed Tomography 512 x 512 x 16 Digital Subtraction Angiography 1024 x 1024 x 10 Computed Radiography 3520 x 4280 x 12 Digital Radiography (Flat-Panel Imagers) 3000 x 3000 x as-16 Digital Mammography 4096 x 4096 x 12 Section I: Chapter 2 RADT 3463 Computerized Imaging 36 12

MATRIX For a given image size, the larger the matrix size, the greater the number of smaller individual pixels. Increasing the number of pixels will improve the image quality. Upper left = 64 x 64 Upper right = 215 X 215 Lower left = 1024 x 1024 Lower right = 2048 x 2048 Section I: Chapter 2 RADT 3463 Computerized Imaging 37 Pixels make up the matrix. Generally, they are square. Each pixel contains a number (discrete value) that represents brightness level which reflects the tissue characteristics being imaged. PIXELS Section I: Chapter 2 RADT 3463 Computerized Imaging 38 PIXELS Calculate pixel size using the relationship: Pixel Size = FOV/Matrix Size Better spatial resolution is obtained with a larger matrix size, smaller the pixel size for the same FOV. The effect of the matrix size on image clarity is seen in the images from textbook, page 29. Section I: Chapter 2 RADT 3463 Computerized Imaging 39 13

VOXELS Pixels represent the information contained in a volume of tissue in the patient. This volume is referred to as a voxel (volume element). Section I: Chapter 2 RADT 3463 Computerized Imaging 40 VOXELS Tissue voxel information is converted into numerical values and expressed in the pixels, and these numbers are assigned brightness levels. Section I: Chapter 2 RADT 3463 Computerized Imaging 41 BIT DEPTH Pixel bit depth is the number of bits that determines the precision with which the exit radiation is recorded and thus controls the exact pixel brightness that can be specified. The number of bits per pixel is the bit depth. The binary number system uses the base 2 k bits = 2k. Each pixel will have 2k gray levels. Section I: Chapter 2 RADT 3463 Computerized Imaging 42 14

BIT DEPTH In an image with a bit depth of 2, each pixel will have 2 2 (4) gray levels (density). Similarly, a bit depth of 8 implies that each pixel will have 2 8 (256) gray levels or shades of gray. The higher the bit depth, the greater the shades of gray. Section I: Chapter 2 RADT 3463 Computerized Imaging 43 APPEARANCE OF DIGITAL IMAGES The bit depth has an effect on the number of shades of gray, hence the contrast resolution of the image. Section I: Chapter 2 RADT 3463 Computerized Imaging 44 APPEARANCE OF DIGITAL IMAGES Contrast resolution is the differences in image densities. Long scale = more gray shades Low contrast Short scale = fewer gray shades High contrast Section I: Chapter 2 RADT 3463 Computerized Imaging 45 15

APPEARANCE OF DIGITAL IMAGES The characteristics of a digital image are Matrix Size Pixel Size Bit Depth These can affect the appearance of the digital image, particularly: Spatial Resolution (matrix/pixel size) Contrast Resolution (bit depth) Section I: Chapter 2 RADT 3463 Computerized Imaging 46 APPEARANCE OF DIGITAL IMAGES Matrix size effects the detail or spatial resolution The larger the matrix size (for the same FOV), the smaller the pixel size - the better the appearance of detail. If FOV decreases without a change in matrix size, pixel size decreases (pixel size = FOV/matrix size) and detail improves. The radiographer selects a larger matrix size when imaging larger body parts to show small details in the anatomy. Section I: Chapter 2 RADT 3463 Computerized Imaging 47 STEPS IN DIGITIZING AN IMAGE There are three steps to digitizing an image Scanning, Sampling Quantization. Section I: Chapter 2 RADT 3463 Computerized Imaging 48 16

STEPS IN DIGITIZING AN IMAGE Scanning The image is first divided into an array of pixels. Sampling Involves measuring the brightness level of each of the pixels using devices such as a photomultiplier tube (PMT). The signal from the PMT is an analog signal (voltage waveform) that must be converted into a digital image for computer processing. Section I: Chapter 2 RADT 3463 Computerized Imaging 49 STEPS IN DIGITIZING AN IMAGE Quantization The process where brightness levels obtained from sampling are assigned an integer (zero or a negative or positive number) called a gray level. The image is now made up of a range of gray levels. The total number of gray levels is called the gray scale. Section I: Chapter 2 RADT 3463 Computerized Imaging 50 STEPS IN DIGITIZING AN IMAGE Analog-Digital-Converter (ADC) Important in the process of converting an analog signal into digital data for input into a computer. Consists of several components that will divide up the analog signal into equal parts. Section I: Chapter 2 RADT 3463 Computerized Imaging 51 17

STEPS IN DIGITIZING AN IMAGE Analog-Digital-Converter (ADC) The greater the bits, the more accurate the signals from the detectors can be digitized for a faithful reproduction of the original signal. Image quality is better with higher-bit ADCs compared to lower-bit ADCs Section I: Chapter 2 RADT 3463 Computerized Imaging 52 STEPS IN DIGITIZING AN IMAGE Example: A 2-bit ADC will convert he analog signal into 4 (2 2 ) equal parts, resulting in 4 gray levels. An 8-bit ADC will divide up the analog signal into 256 (2 8 ) parts, resulting in 256 gray levels. Digital imaging modalities have 12 to 32-bit ADCs. Section I: Chapter 2 RADT 3463 Computerized Imaging 53 DIGITAL IMAGE PROCESSING OPERATIONS Point Processing Operations Local Processing Operations Global Processing Operations Geometric Operations Image Processing Operations Intended to change the intensity values of the pixels in the input image and display the resulting changes in the output image with the goal of changing the characteristics of the image to enhance diagnosis Section I: Chapter 2 RADT 3463 Computerized Imaging 54 18

POINT PROCESSING OPERATIONS The value of the one (point) input image pixel is mapped onto the corresponding output image pixel; that is, the output image pixel value at the same location as on the input image matrix depends on the value of the input image pixel. Section I: Chapter 2 RADT 3463 Computerized Imaging 55 POINT PROCESSING OPERATIONS The operation (algorithm) allows the entire input image matrix to be scanned pixel by pixel, using a "pixel point process" until the entire image is transformed. One common point processing operation is Gray-level Mapping or Gray-scale Processing Other terms to describe gray-level mapping are: Contrast Stretching Contrast Enhancement Histogram Modification Histogram Stretching Windowing Section I: Chapter 2 RADT 3463 Computerized Imaging 56 POINT PROCESSING OPERATIONS Windowing is the most commonly used image processing operation in digital imaging, including computed radiography, digital radiography using flat-panel detectors, CT, and MRI. Image contrast and brightness transformations are done using a variety of processing techniques The two common methods used in digital radiology are Look-up-table (LUT) Method Windowing Method. Section I: Chapter 2 RADT 3463 Computerized Imaging 57 19

POINT PROCESSING OPERATIONS Histogram A histogram is a graph of the number of pixels in the entire image or part of the image having the same gray levels (density values) plotted as a function of the gray levels Section I: Chapter 2 RADT 3463 Computerized Imaging 58 POINT PROCESSING OPERATIONS Histogram Changing the histogram of the image can alter its output image matrix brightness and contrast. The operation is called histogram modification or histogram stretching. A wide histogram implies more contrast and a narrow histogram will show less contrast. If the values of the histogram are concentrated in the lower end of the range of values, the image appears dark; conversely, the image appears bright at the higher end of the range. Section I: Chapter 2 RADT 3463 Computerized Imaging 59 POINT PROCESSING OPERATIONS Look-up Table This illustration shows a low-contrast numerical image with a contrast difference, (that is, the object relative to the background) of 10 (40-30) where 40 represents the background contrast and 30 represents the object contrast. continued Section I: Chapter 2 RADT 3463 Computerized Imaging 60 20

POINT PROCESSING OPERATIONS Look-up Table The LUT is then used to change the low-contrast numerical image to a high-contrast image by assigning numbers to the input values 40 and 30 that will subsequently change them into 90 and 10, respectively. The contrast difference for the new output image (on the right) is now 80 (90-10) and therefore this image appears as a high-contrast image. continued Section I: Chapter 2 RADT 3463 Computerized Imaging 61 POINT PROCESSING OPERATIONS Look-up Table During digital image processing, the LUT determines the numbers assigned to the input pixel values that change them into output pixel values, resulting in a change in contrast and brightness of the image continued Section I: Chapter 2 RADT 3463 Computerized Imaging 62 POINT PROCESSING OPERATIONS Look-up Table Digital radiographic imaging systems utilize a wide range of LUTs stored in the system for the different types of examinations (chest, spine, pelvis. extremities). The radiographer should select the appropriate LUT to match the part being imaged. continued Section I: Chapter 2 RADT 3463 Computerized Imaging 63 21

POINT PROCESSING OPERATIONS Look-up Table An important point is that since digital radiographic detectors have wide exposure latitude and a linear response, the image displayed without processing may appear as a low-contrast image. This illustrates the effect on a chest image of using an LUT to convert a low-contrast image into a high-contrast image. Section I: Chapter 2 RADT 3463 Computerized Imaging 64 POINT PROCESSING OPERATIONS Windowing Windowing is a digital image processing technique that also changes the contrast and brightness of an image. The illustration shows the range of pixel values (gray levels) and displayed image contrast range on a digital image. Section I: Chapter 2 RADT 3463 Computerized Imaging 65 POINT PROCESSING OPERATIONS Windowing A digital image is made up of numbers. The range of number is the window width (WW) The center of the range is defined as the window level (WL). The WW controls image contrast. The WL controls the brightness. Section I: Chapter 2 RADT 3463 Computerized Imaging 66 22

POINT PROCESSING OPERATIONS Windowing The displayed WW and WL values are always shown on the image. Narrow WW provides higher image contrast (short-scale contrast) and a wide WW will show an image with less contrast. Section I: Chapter 2 RADT 3463 Computerized Imaging 67 POINT PROCESSING OPERATIONS Windowing If the WL is increased, the image becomes darker since more of the lower numbers will be displayed. Section I: Chapter 2 RADT 3463 Computerized Imaging 68 LOCAL PROCESSING OPERATIONS This shows a local processing operation. The output image pixel values are from a small area of pixels around the corresponding input pixel. Because of small area of pixels or group of pixels is used, these operations are also referred to as area group processes. Spatial frequency filtering is an example. Section I: Chapter 2 RADT 3463 Computerized Imaging 69 23

LOCAL PROCESSING OPERATIONS Spatial Location Filtering-Convolution Behind the scenes.accomplished automatically during Fourier transformation The convolution process is implemented by: Placing a filter mask array (matrix) over the image array in the memory A filter mask is a square array or section with numbers The size of the mask is determined by the manufacturer of the equipment continued Section I: Chapter 2 RADT 3463 Computerized Imaging 70 LOCAL PROCESSING OPERATIONS Spatial Location Filtering-Convolution (continued) Pixels directly above or below the mask are averaged to obtain a discrete value for each pixel Performed primarily to attenuate the higher frequencies within the image The result of the process is a blurring of the higher frequencies or intensities within the image Section I: Chapter 2 RADT 3463 Computerized Imaging 71 LOCAL PROCESSING OPERATIONS Spatial Location Filtering-Convolution (continued) Section I: Chapter 2 RADT 3463 Computerized Imaging 72 24

LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: High-Pass Filtering High-pass Filtration Edge Enhancement Produces edge-sharpening Remarkably high-contrast Useful in demonstrating small structures Section I: Chapter 2 RADT 3463 Computerized Imaging 73 LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: High-Pass Filtering Original Tokyo Airport High-pass filter Original LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: High- Pass Filtering Edge Enhancement More Edge Enhancement 25

LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: High-Pass Filtering LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: Low-Pass Filtering Smoothing Section I: Chapter 2 RADT 3463 Computerized Imaging 77 LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: Low-Pass Filtering Smoothing Original Low-pass filter 26

LOCAL PROCESSING OPERATIONS Spatial Frequency Filter: Masking The blurred image produced from the low-pass filtering process is subtracted from the original image to produce a sharp image. Section I: Chapter 2 RADT 3463 Computerized Imaging 79 GLOBAL PROCESSING OPERATIONS Global Processing Operations Implies all the pixels in the entire input image are used to change the value of a pixel in the output image. Uses the FT in filtering images in the frequency domain rather than the spatial location Process images for edge enhancement, image sharpening, and image restoration. Section I: Chapter 2 RADT 3463 Computerized Imaging 80 GEOMETRIC OPERATIONS Geometric Operations Image processing operations sometimes used in digital radiology. Allow the user to change the position or orientation of pixels in the image rather that the brightness of the pixels. Results in the scaling. sizing, rotation, and translation of images, once again, to enhance diagnosis. Section I: Chapter 2 RADT 3463 Computerized Imaging 81 27

IMAGE POST-PROCESSING: AN ESSENTIAL TOOL FOR TECHNOLOGISTS DIGITAL IMAGE POSTPROCESSING PROCESS Annotation Window and Level Magnification Image flip Image inversion Subtraction (DSA) Pixel shift Region of interest Label the image RESULTS Expand the digital grayscale to visible Improve visualization and spatial resolution Reorient image presentation Make white-black and black-white Improve image contrast Reregister an image to correct for patient motions Determining average pixel value for use in quantitative imaging Section VI: Chapter 29 RADT 3463 Computerized Imaging 82 SECOND LOOK: Sprawls Educational Foundation http://www.sprawls.org/ppmi2/imgchar/ Section I: Chapter 1 RADT 3463 Computerized Imaging 83 QUESTIONS?? Service to others is the rent you pay for your room here on earth. Mohammed Ali Section I: Chapter 2 RADT 3463 Computerized Imaging 84 28