MEDICAL IMAGE PROCESSING
DETECTION OF REGIONS OF INTEREST
CHAPTER 4
MEDICAL IMAGE PROCESSING
- In Computer Aided Detection/Diagnosis (CAD)
systems, one of image processing’s roles would be to
detect regions of interest (ROIs) for a given, specific,
screening or diagnostic application then characterize and
classify them into one of several categories
- ROIs in biomedical imaging and image analysis:
+ Organs on the image such as breast, lung or bone
region…
+ Abnormal signs on the image as such calcifications or
masses on mammogram, lung nodules on lung CT
image…
-2-
MEDICAL IMAGE PROCESSING
-3-
- Segmentation: the process that divides an image R into
its constituent parts, objects or ROIs Ri
MEDICAL IMAGE PROCESSING
-4-
- Ri should satisfy
+ Homogeneity: pixels within a region should be
similar in an aspect
+ Connectivity: always exist a connection between any
2 pixels in the region
4.1. Thresholding and Binarizaion MEDICAL IMAGE PROCESSING
0 if f(m,n) T
g(m,n) 255 if f(m,n) T
=
5
- When object’s gray levels are known from prior
knowledge or can be determined from the histogram
30000x TEM image of a ligament sample demonstrating collagen
fibres (left) and thresholded image with T=180 (right)
