1/41
Looks like no tags are added yet.
Pixel
smallest unit of a digital 2D image
Digital xrays are a grid of pixels representing xray energy
increase pixel = increase spatial resolution
Voxel
smallest unit of 3D image
CT or MRI scan are grid of them which represent volume of tissue
increase in slice thickness = decrease in spatial resolution
Intensities
amount of energy or signal detected from region
Dynamic range
range of possible intensity values that an image can display
higher = better differentiation between tissues
Spatial Resolution
ability of imaging system to differentiate between two closely spaced objects
higher = better ability to detect finer structures (tumors, lung nodules)
Tradeoffs in Medical Imaging
increase in resolution means increase in file size, increase in processing time, and increase in computational resources and increase in noise
Binary Image
each pixel density is either 0 (black) or 1 (white)
Greyscale Image
single channel, each pixel intesity corresponds to a specific shade of grey (0 to 255)
Xray, CT, and MRI
RGB Images
uses 3 color channels (red, green, blue)
less common in diagnostic imaging but used in certain fields like pathology and retinal imaging
Cardiac doppler imaging
example of RGB imaging that uses sound waves to create images of blood through the heart and blood vessels
Bit Depth
the number of bits used to reepresent each pixel’s intensity values
increase = increase in shades of grey
8 bit = 2^8 = 256 shades, 12 bit = 4096 shades of grey, 16 bit = 65536 shades of grey
Segmentation
processing of images into meaningful regions or segments either manually or automatically (computer algorithms)
Segmentation Challenges
interobserver variability - inconsitencies due to segmentations performed by different radiologists
time consuming
sample size - small regions of pathology are often underrepresented in datasets making segmentation more difficult
Large data set needed for automated segmentation
Thresholding
primary method for segmentation using grey values contributing to a certain structure (0-255
Region Growing
user defines seed point and neighboring pixels are added based on similarity of intensity (added to a region) or dissimiliarity (not added = stop region grow)
6 connectivity
all voxels share a face
26 connectivity
all voles that share face and edges
Dynamic region growing
combines thresholding and region growing by picking a starting voxels with a corresponding gray value and evaluates selection and connected voxels
Draw profile line
threshold technique that shows how the gray value changes along a line within an image when GV is unknown
3D live wire
choose automatic contour in one plane then the boundaries are highlighted in other views
Gradiant magnitude - 0% contour attraccted to darker regions while 100% contour is attracted to lighter regions
Attraction - (-3) means all small inclusions included in masks while (3) means inclusions are excluded from automatic contour
Local Threshold
improve already existing segmentation by threshold zone directly attached to already existing mask and good for seperating extra anatomy
Smart Brush
edits existing mask are create new one by marking slices on the 2D images with pixel intensity sensitive brush which allows you to see preview of voxels that will be added or removed from mask by hovering over image data
Crop mask
selects region of interest and restricting segmentation to it
Split mask
splits mask into multiple smaller masks using combination of pixel intensities and edge detection by only marking a few slices
Edit mask
modify mask on a single slice to add or remove, also by threshold
Multiple slice edit
uses interpolation to edit masks with minimal input saving time but only on one view
3D interpolate
same as multiple slice edit but used on 2+ planes
Nearest neighbor (expands the number) and bilinear (divides equally among spaces)
Segment thin bone
helps you close holes that appear within cortical bone due to scanner resolution by drawing a line
Boolean operation
minus, unite, intersection
Minus
a resulting new mask is equal to the difference between sets of pixels of Masks A and B with threshold Value of Mask A
Unite
a resulting mask includes pixels of both masks with lower threshold equal to lowest threshold of A and B, upper threshold equal to the highest threshold of A and B
Intersection
a resulting mask equal to the common set of pixels shared by both masks with the lower threshold equal to higher lower threshold of 2 masks, upper threshold equal to lowest higher threshold
Morphology Operation
Erode, dilate, open, close
Dialate
pixels added to edges, uniformly grows in all directions
Erode
pixels removed from outer edge and followed by region grow
Open
erosion then dilation which removes small edges, or opens internal holes
Close
dilation then erosion, closes cavities and gaps on edges
Cavity fill
fills internal gaps of selected mask and add region to mask
Smart fill
fills holes by global and local filling by defining hole closing distance parameter
Smart expand
expands initial mask until it finds an edge in images according to gray value gradients
Smooths mask
smooths mask by integrating pixels (opposite of interpolating) using mode filter, median filter, mean filter, or gaussian filter