Kaushiki Roy presents Automated segmentation of overlapping cells from fluorescence and HE-stained image
On 2025-03-04 11:00:00 at G205, Karlovo náměstí 13, Praha 2
THE DATE HAS CHANGED - OTHERWISE THIS IS A REPETITION OF AN EARLIER
ANNOUNCEMENT
Cell segmentation plays a critical role in developing automated systems for
microscopic image analysis. However, cancer cells tend to grow in confluent
cell
colonies and separating individual cells is difficult. Thus, in the seminar, I
will talk about a system that we developed to segment overlapping cells from
confluent cell colonies. Our framework was tested on multiple cancer cell lines
namely HeLa, MDA-MB-468, MDA-MB231, T-47D and a public dataset to infer its
generalization ability. The next part of our task was to study the effect of
drugs in these segmented cells. Cells from different cancer cell lines were
treated with various drugs or combination of drugs and our objective was to
decide the percentage of the cells that died by centromere fragmentation
phenotype. Centromere fragmentation is a common mode of cell death or apoptosis
following drug treatment and greater the percentage of cell death, the more
efficient a drug is in targeting a particular cancer cell line. We developed an
automated system for classifying cells that had undergone centromere
fragmentation and hence distinguish it from the healthy cells. Our framework
worked successfully for most cell lines.
The following papers will be discussed
1. Roy, Kaushiki, Debapriya Banik, Gordon K. Chan, Ondrej Krejcar, and Debotosh
Bhattacharjee. "2pClPr: A Two-Phase Clump Profiler for Segmentation of Cancer
Cells in Fluorescence Microscopic Images." IEEE Transactions on Instrumentation
and Measurement 72 (2023): 1-14.
2. Roy, Kaushiki, Cody W. Lewis, Gordon K. Chan, and Debotosh Bhattacharjee.
"Automated classification of mitotic catastrophe by use of the centromere
fragmentation morphology." Biochemistry and Cell Biology 99, no. 2 (2021):
261-271.
Kaushiki Roy has completed a PhD from Jadavpur University, India. She is
currently a post-doc in the BIA group of Jan Kybic.
ANNOUNCEMENT
Cell segmentation plays a critical role in developing automated systems for
microscopic image analysis. However, cancer cells tend to grow in confluent
cell
colonies and separating individual cells is difficult. Thus, in the seminar, I
will talk about a system that we developed to segment overlapping cells from
confluent cell colonies. Our framework was tested on multiple cancer cell lines
namely HeLa, MDA-MB-468, MDA-MB231, T-47D and a public dataset to infer its
generalization ability. The next part of our task was to study the effect of
drugs in these segmented cells. Cells from different cancer cell lines were
treated with various drugs or combination of drugs and our objective was to
decide the percentage of the cells that died by centromere fragmentation
phenotype. Centromere fragmentation is a common mode of cell death or apoptosis
following drug treatment and greater the percentage of cell death, the more
efficient a drug is in targeting a particular cancer cell line. We developed an
automated system for classifying cells that had undergone centromere
fragmentation and hence distinguish it from the healthy cells. Our framework
worked successfully for most cell lines.
The following papers will be discussed
1. Roy, Kaushiki, Debapriya Banik, Gordon K. Chan, Ondrej Krejcar, and Debotosh
Bhattacharjee. "2pClPr: A Two-Phase Clump Profiler for Segmentation of Cancer
Cells in Fluorescence Microscopic Images." IEEE Transactions on Instrumentation
and Measurement 72 (2023): 1-14.
2. Roy, Kaushiki, Cody W. Lewis, Gordon K. Chan, and Debotosh Bhattacharjee.
"Automated classification of mitotic catastrophe by use of the centromere
fragmentation morphology." Biochemistry and Cell Biology 99, no. 2 (2021):
261-271.
Kaushiki Roy has completed a PhD from Jadavpur University, India. She is
currently a post-doc in the BIA group of Jan Kybic.