Image Analysis of Circulating Tumour Cell Clusters from Imaging Flow Cytometry Data
(2020) BMEM01 20202Department of Biomedical Engineering
- Abstract
- Circulating tumour cells (CTCs) are cancer cells that have entered the circulation of the body breaking free from their primary tumour and that can act as progenitors of metastasis. At the time of writing, a study on a novel method to detect and count CTCs using imaging flow cytometry (IFC) is being conducted at Lund University. In the study, a problem was found where CTCs clustered with normal white blood cells (WBCs) were not detected as CTC candidates. These CTCs were not detected because the analysis software treated clusters the same as single cells. The rarity of CTCs in blood means it is important to detect every single one in a sample.
This thesis aimed to develop an algorithm that could detect CTC-WBC clusters in IFC data of... (More) - Circulating tumour cells (CTCs) are cancer cells that have entered the circulation of the body breaking free from their primary tumour and that can act as progenitors of metastasis. At the time of writing, a study on a novel method to detect and count CTCs using imaging flow cytometry (IFC) is being conducted at Lund University. In the study, a problem was found where CTCs clustered with normal white blood cells (WBCs) were not detected as CTC candidates. These CTCs were not detected because the analysis software treated clusters the same as single cells. The rarity of CTCs in blood means it is important to detect every single one in a sample.
This thesis aimed to develop an algorithm that could detect CTC-WBC clusters in IFC data of prostate cancer patient samples. An algorithm that could automate the detection of CTC candidates would simplify the present process which suffer from excessive manual assessment. The main problem to be solved was to segment the different cells in the clusters from each other in the images.
An algorithm to detect CTC-WBC clusters in IFC data was proposed and was initially tested on three patient datasets. The algorithm showed stable segmentation results. The problem of segmenting cells was solved by using an Otsu threshold and watershed approach on images of cells stained with the nucleic fluorescent marker DAPI. The segmented regions could then be used to examine the fluorescent intensity of other stains within the regions.
The initial results of CTC detection were promising. The number of candidates to manually assess to find CTC-WBC clusters was greatly reduced and is now at a manageable level.
At the time of writing this, the program is deployed and ready for use in the continuation of the study. (Less) - Popular Abstract (Swedish)
- Detektion av tumörceller i blod med hjälp av bildanalys
Cirkulerande tumörceller är celler som har lämnat sin ursprungliga primärtumör och tagit sig till blodet. Dessa celler kan sedan ta sig vidare till andra delar av kroppen där de kan ge upphov till dottertumörer. Cirkulerande tumörceller är mycket viktiga att studera eftersom dottertumörer är den ledande orsaken till dödsfall i cancer.
Tyvärr är cirkulerande tumörceller extremt svåra att upptäcka. Sannolikheten att en cell i blodet hos en cancerpatient är en tumörcell kan vara mindre än en på miljonen, och då räknar vi inte ens med röda blodceller. I en studie på Lunds universitet undersöks just nu en ny metod att detektera cirkulerande tumörceller. I mitt examensarbete har jag... (More) - Detektion av tumörceller i blod med hjälp av bildanalys
Cirkulerande tumörceller är celler som har lämnat sin ursprungliga primärtumör och tagit sig till blodet. Dessa celler kan sedan ta sig vidare till andra delar av kroppen där de kan ge upphov till dottertumörer. Cirkulerande tumörceller är mycket viktiga att studera eftersom dottertumörer är den ledande orsaken till dödsfall i cancer.
Tyvärr är cirkulerande tumörceller extremt svåra att upptäcka. Sannolikheten att en cell i blodet hos en cancerpatient är en tumörcell kan vara mindre än en på miljonen, och då räknar vi inte ens med röda blodceller. I en studie på Lunds universitet undersöks just nu en ny metod att detektera cirkulerande tumörceller. I mitt examensarbete har jag löst ett problem för studien där vissa av tumörcellerna inte kunde detekteras.
I den nya metoden används en högteknologisk maskin, kallad bildflödescytometer, som kan ta bilder på celler en och en i ett blodprov. Därtill kan maskinen ta fluorescensbilder, vilket innebär att man kan färga in celler med olika markörer och urskilja celler med vissa egenskaper. Det är dock inte alltid som celler fotograferas exakt en och en. Ibland kan celler sitta ihop i kluster som i figur 1, speciellt om blodprovet kommer från en cancerpatient.
Problemet med metoden i den pågående studien var att analysprogrammet som användes inte gjorde skillnad på kluster och celler. Det gjorde att tumörceller som satt ihop med vita blodceller i detta fall automatiskt sorterades bort. Om man manuellt skulle försöka hitta dessa tumörceller skulle man behöva gå igenom uppemot 100 000 bilder. Det finns det inte många som har tid med.
Jag löste detta problem genom att skapa en bildanalysalgoritm som kunde skilja på celler i kluster. Detta kallas inom bildanalys för segmentering. Segmentering delar upp en bild i dess intressanta regioner. När jag hade löst segmenteringen av cellerna i klustren kunde jag analysera cellerna individuellt och upptäcka de tumörceller som tidigare låg gömda. Istället för att behöva gå igenom 100 000 bilder räcker det nu med att kolla igenom ett handfull bilder. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9033350
- author
- Berg, Filip LU
- supervisor
- organization
- course
- BMEM01 20202
- year
- 2020
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- MSc, Image Analysis, Segmentation, Otsu Threshold, Watershed, Circulating Tumour Cells, Imaging Flow Cytometry
- language
- English
- additional info
- 2020-21
- id
- 9033350
- date added to LUP
- 2021-01-22 15:08:26
- date last changed
- 2021-01-22 15:08:26
@misc{9033350,
abstract = {{Circulating tumour cells (CTCs) are cancer cells that have entered the circulation of the body breaking free from their primary tumour and that can act as progenitors of metastasis. At the time of writing, a study on a novel method to detect and count CTCs using imaging flow cytometry (IFC) is being conducted at Lund University. In the study, a problem was found where CTCs clustered with normal white blood cells (WBCs) were not detected as CTC candidates. These CTCs were not detected because the analysis software treated clusters the same as single cells. The rarity of CTCs in blood means it is important to detect every single one in a sample.
This thesis aimed to develop an algorithm that could detect CTC-WBC clusters in IFC data of prostate cancer patient samples. An algorithm that could automate the detection of CTC candidates would simplify the present process which suffer from excessive manual assessment. The main problem to be solved was to segment the different cells in the clusters from each other in the images.
An algorithm to detect CTC-WBC clusters in IFC data was proposed and was initially tested on three patient datasets. The algorithm showed stable segmentation results. The problem of segmenting cells was solved by using an Otsu threshold and watershed approach on images of cells stained with the nucleic fluorescent marker DAPI. The segmented regions could then be used to examine the fluorescent intensity of other stains within the regions.
The initial results of CTC detection were promising. The number of candidates to manually assess to find CTC-WBC clusters was greatly reduced and is now at a manageable level.
At the time of writing this, the program is deployed and ready for use in the continuation of the study.}},
author = {{Berg, Filip}},
language = {{eng}},
note = {{Student Paper}},
title = {{Image Analysis of Circulating Tumour Cell Clusters from Imaging Flow Cytometry Data}},
year = {{2020}},
}