Download Prostate Microscopic Image Segmentation

Prostate Microscopic Image Segmentation
 Goals:
• To segment cancer cells from the prostate biopsy images.
• To extract object-level features such as mean intensity, area, perimeter, and
standard deviation of the mean intensity of the segmented histological objects.
 Brief Description:
The prostate cancer diagnosis done by pathologists is very subjective and relatively
slow since it heavily depends on the pathologists’ interpretation. In this project, we
try to use image processing technique to assist pathologists. Image segmentation is
the first step. Cancer cells are segmented from the biopsy images and then objectlevel features, such as mean intensity, area, perimeter, and standard deviation of
the mean intensity of the segmented histological objects, are extracted from
segmented histological objects. This can then be used together with metabolites as
features for tumor/non-tumor classification.
Prostate Microscopic Image Segmentation
 Heights of achievements throughout the semester (bullet points)
▫ Segmented the cancer cells from a few images.
▫ Features are extracted from the segmentation.
Prostate Microscopic Image Segmentation
 Representative Figures/Diagrams/Videos that highlight your research
methodology and results.
As an example, Figure 1 shows a prostate biopsy image with pathologist
circled cancer region.
Figure 1. A prostate biopsy image with pathologist circled cancer region.
Prostate Microscopic Image Segmentation
Then the region is selected in the Leica software as shown in Figure 2(a). Cancer
cells are segmented and numbered as shown in Figure 2(b).
Original ROI
(a)
Segmented ROI
(b)
Figure 2. (a) selected region of interest (ROI) from
the prostate biopsy image and (b) segmented ROI.
Prostate Microscopic Image Segmentation
A few object-level features, such as mean intensity, area, and perimeter of the
segmented histological objects, are extracted from the segmentation as shown in
Figure 3.
The value of each area
6000
4000
(a)
2000
0
2
4
6
8
10
12
14
16
18
20
The mean intensity of each area
170
160
150
(b)
140
130
2
4
6
8
10
12
14
16
18
20
The perimeter of each area
800
600
400
(c)
200
0
2
4
6
8
10
12
14
16
18
20
Figure 3. Object-level features,
(a) mean intensity, (b) area,
and (c) perimeter of the
segmented histological objects,
are extracted from the
segmentation.