w5 Flashcards
what is the difference between screening and diagnostics
A screening test is given to those who have no symptoms of the condition in question, but tests are performed to find any sign of abnormality associated with a disease.
A diagnostic test is used to find the probability of a suspected condition once initial testing has revealed its possibility. The medical expert concentrates on the region of suspected abnormality, and examines its characteristics to decide if the region exhibits signs related to a particular disease.
how is detection of regions of interest performed traditionally, and why is this evaluation not ideal
radiological images are interpreted by visual inspection of the images printed slice by slice on film and displayed against a light box.
Not ideal because it is subjective. The third dimension has to be reconstructed mentally by the radiologist by looking at adjacent image slices.
what are the different methods of digital deliniation of regions of interest
Manual delineation.
Automatic segmentation: bounding-box, edge based, region based, etc.
Semi-automated approaches.
what is the disadvantage of using manual delineation
inter-(between different people)
and intra-user (between same person) variance
what is Segmentation
the process that divides an image into its constituent parts, objects, or ROIs.
Segmentation is an essential step before the description, recognition, or classification of an image or its constituents
The two major approaches to image segmentation are based on the detection of what two digital characteristics
Similarity / region-based segmentation : Homogeneous parts (regions of similar color) are detected based on gray-level thresholding, region growing, and region splitting / merging.
Discontinuity / edge-based segmentation : Abrupt changes in gray level (corresponding to edges) are detected.
what is thresholding segmentation (binarization)
similarity-based. Image color is thresholded such that features of interest are set to black while other details are rejected, i.e. set to white.
what are the different categories of thresholds for thresholding segmentation?
Global: when T is a constant applicable over an entire image e.g. there is no light distortion across image.
Variable: when the value of T changes over an image.
Local or regional thresholding: if the value of T at any point (x,y) depends on properties of a neighborhood of (x,y) (e.g., the average intensity of the pixels in the neighborhood).
Dynamic or adaptive thresholding: if T depends on the values of the spatial coordinates (x,y) themselves.
what are the steps / algorithm for Basic Global Thresholding
Step 1. Select an initial estimate for the global threshold, T.
Step 2. Segment the image using T. This will produce two groups of pixels: G1 = all pixels with intensity values > T and G2 = pixels with values ≤ T.
Step 3. Compute the average (mean) intensity values m1 and m2 for the pixels in regions G1 and G2, respectively.
Step 4. Compute a new threshold: T = (m1 + m2) / 2.
Step 5. Repeat steps 2 - 4 until the difference between values of T in successive iterations is smaller than a predefined parameter ΔT.
in what types of images does the Basic Global Thresholding algorithm work best for
images where there are reasonably clear valleies between the modes of the histogram.
what characteristics affect the quality of the valleys of histograms
The separation (distance) between peaks.
The noise content in the image (the modes broaden as noise increases).
The relative sizes of objects and background.
The uniformity of the illumination sources.
The uniformity of the reflectance properties of the image (glare).
why would you want to smoothen a medical image
to make it more strongly multimodal / improves quality of peaks.
if illumination and reflectance cannot be controlled, what are three ways in which you can correct for these distortions?
To correct the shading pattern directly e.g. non-uniform (but fixed) illumination can be corrected by multiplying
the image by the inverse of the pattern, which can be obtained by imaging a flat surface of constant intensity.
To “work around” non-uniformities using variable thresholding e.g. using image partitioning.
To correct the global shading pattern e.g. using the top-hat transformation (erosion then dialation).
Basic adaptive thresholding via image partitioning is used to compensate for non-uniformities in illumination and/or relfectance.
How is Basic Adaptive Thresholding via Image Partitioning performed, and what type of image is this method good for
partition the image (sub-images are approximately uniform) and perform thresholding segmentation on each partition.
good for images when the objects of interest and the background occupy regions of reasonably comparable size. if many segments contain only objects or only backround, then the chances of falsely segmenting those pixels becomes higher.
what is Region-growing based segmentation
examines pixels in the neighborhood based on a pre-defined similarity criterion. The neighborhood pixels with similar properties are merged to form closed regions for segmentation.
can be extended to merging regions or region splitting.
