Digital Image Processing

Image Processing – manipulation an image to enhance image or to extract useful information from it 

• Change images depending on what you need it for 

Analogue vs Digital 

Analogue – continuous tone images produced by analogue optical and electronic devices, vary continuously over all dimensions of the images 

To process/display analogue image, it must be converted into a computer-readable form or digital format 

Screen/Film Radiography 

Cassettes (aluminium or plastic) holding light sensitive photographic film (single use) 

Preparation of cassettes done in dark rooms 

Films passed through film processing unit (developers/fixers/etc.) - took some time and chance of ruining film 

Physical image produced 

Cannot process/change image if exposures/technique is wrong 

Problems: 

• Physical storage 

• Lack of post processing 

Digital Images – matrix of many small elements (pixels) 

Each pixel represented by a numerical value 

Pixel value related to brightness/colour seen when digital image converted into analogue image for display and viewing 

Digital Image Processing – use of digital computer to process digital images through an algorithm  

Can do more with a digital image because you can change the numerical value 

Purpose of early image processing was to improve quality of image – input is a low-quality image and output is image with improved quality 

Development of digital image processing affected by 

• Development on computers 

• Development of mathematic algorithms 

• Demand for a wide range of applications in environment, agriculture, military, industry and medical science 

Computer Language 

Binary: 1 or 0 – on/off, black/white 

Bit – binary digit 

Limitation of binary numbers: range of values that can be written with a specific number of bits 

Four bits can have 16 different values because there are 16 ways four bits can be placed 

• Range of possible values that can be written increased by using more bits 

• Range of possible values doubled for each additional bit 

• More bits require more processing power 

Pixel Bit Depth – number of bits available in digital system to represent each pixel in image 

With 4 bits, pixel limited to having only 16 different values – brightness levels/shades of grey 

Standard medical displays about 10 bit – latest displays are 24 bit 

• No need to have more than 10 bits as the eye cannot distinguish such subtle changes in grey 

Pixel size 

Smaller pixel sizes require more pixels to fill image matrix/portray patient anatomy 

Represents smaller amount of image – allows to represent more shades of grey 

Smaller pixel size increases image quality as resolution 

Reduce field of view but keep number of pixels (pixel size becomes smaller) increases detail able to be seen – keeps resolution but provides more detail 

Digital Image Post-Processing 

Film-based radiology obsolete – all imaging modalities produce digital images that can be post-processed and manipulated with relative ease 

Main aim of digital image post-processing is to alter/change image to enhance diagnostic interpretation  

Pre-Processing vs Post-Processing 

Pre-processing operations apply appropriate corrections to the raw data (done by system depending on modality) 

Post-processing changes image contrast, reduce image noise and enhance image sharpness of image displayed to enhance diagnostic interpretation 

Post-Processing 

Varies between imaging modalities – each uses specific operations best suited to enhance that specific image 

Image Post-Processing Categories 

Image restoration – improve quality of images with distortions/degradations 

Image analysis – allows measurements and statistics to be performed (image segmentation, feature extraction, classification of objects) 

Image synthesis – create images from other images or non-image data 

Image enhancement – generates image that is more pleasing to observer (contrast enhancement, spatial and frequency filtering, noise reduction) 

Image compression – reduce size of image to decrease transmission time and reduce storage space 

Limitations to how much an image can be improved 

Post-processing doesn’t make up for not optimal technique – should aim to acquire best possible/optimum image first and then post-process minimally 

Grey Scale Processing  

Most common methods to adjust contrast  

• Lookup Table 

• Windowing  

Lookup Tables – pre-calculated data (numerical information) stored in computer used to substitute new values for each pixel during processing 

• Provide quick and efficient way of enhancing image contrast 

Increases contrast between two structures or structures and the background 

Low contrast image changed into a high contrast image 

LUT Graphs/Histograms 

Plotting number values onto graph 

Starting point is straight line/linear graph – shows substituted number is same as original image pixel value 

To change contrast characteristics of image, must substitute numbers that are different from original pixel values 

Brightness/contrast enhanced in specific parts of image, not uniformly across the whole image 

Used in CR/DR systems for different types of examinations – chest/spine/pelvis/extremities 

• Radiographer should select appropriate LUT to match part being imaged 

Dynamic Range 

Digital radiographic detectors have wide exposure latitude – sensitive to large range of x ray intensities (allows wider range of exposure factors to be used, more room for error which will still produce a diagnostic image) 

Latitude (dynamic range) - range of receptor exposures over which an image and contrast will be formed 

Digital receptors respond to x ray exposure and produce digital data over a wide range of x ray exposure values 

Radiographic film has a limited dynamic range 

Exposure Index – measure of amount of exposure received by image receptor 

• Provides useful feedback about accuracy of exposure utilised 

• Vendor specific but there is an international standard for EI 

• Prevents over exposure of patient – image gained may still be diagnostic but cannot tell it is overexposed 

Image windowing – selecting segment of total value range and displaying pixel values within that segment over full brightness range from white to black 

Adjusts brightness and contrast of an image to visualise specific anatomy 

Windows allows the display and enhancement of contrast in selected segments of total pixel value range 

On film, full range of exposure displayed in one image and cannot be changed 

Windowing creates many displayed images – each one focuses on specific range of pixel values (used in CT) 

When window is set to cover the lower segment of total pixel value, good contrast seen in lighter areas e.g., mediastinum 

Setting window higher segment produces good contrast in darker areas e.g., lungs 

Spatial Frequency Processing 

Spatial resolution – detail seen 

Series of different algorithms used to post process image 

• Edge enhancement/sharpness to increase detail by sharpening edges 

• Unsharp Masking uses Digital Subtraction to enhance image sharpening 

• Smoothing/blurring to reduce noise/graininess - can compromise image quality 

Digital Subtraction Angiography – fluoroscopic technique used extensively in interventional radiology to visualise blood vessels 

Radiopaque structures (bones) eliminated digitally, allows accurate depiction of blood vessels without interfering shadows from overlapping tissues 

Provides clear view of vessels and allows for a lower dose contrast medium 

Geometrical Processing – techniques allowing user to change position/orientation of pixels in image rather than contrast/brightness of pixels 

Results in scaling/sizing/centering/cropping/rotation of images to enhance diagnosis 

Image Reconstruction (CT) - use of mathematical algorithms to create new image sets by processing original images  

Particularly used in CT 

• MPR (routine) 

• MIP (Maximum Intensity Projection) 

• Volume Rendering/3D Reconstruction (surgical planning) 

MIP 

Multiple images in a series, taking maximum attenuation (brightest parts) and stacking images up 

Combined slices into one image, all information from the slices represented in the image – decreases number of images but maintaining amount of information 

e.g., used in angiography  

Digital Breast Tomosynthesis (DBT) - allows volumetric reconstruction of whole breast from a finite number of low-dose 2D projections obtained by different x ray tube angles 

Image Registration/Image Fusion – process of mapping input images with help of reference image 

• Gives structural and functional information 

• e.g., CT and PET combined 

Goal is to match corresponding images based on certain features to assist in image fusion process 

Computer Aided Detection/Diagnosis (CAD) - use of computer-generated output as assisting tool for clinician to make a diagnosis 

Most common applications 

• Detection of breast cancer (mammography) 

• Detection of pulmonary nodules (chest CT) 

Image Compression  

Fixes digital storage problem – post processing creates more digital images that also must be stored 

Depends on how image is acquired 

Lossless/reversible - prevents loss of information, more storage used 

Lossy/irreversible security – not saving all of data, minimising quality to save space 

Storage has cost considerations for the hospital 

SUMMARY 

Image processing cannot increase amount of information available in input image 

Removing information not relevant can make it easier to interpret images 

Image processing always limited by quality of input data 

If imaging system provides data of unacceptable quality, better to try to improve imaging system rather than hope image processing will compensate for poor imaging