Deep Learning Pipeline for Chromosome Segmentation

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26220%2F22%3APU145357" target="_blank" >RIV/00216305:26220/22:PU145357 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/9764950" target="_blank" >https://ieeexplore.ieee.org/document/9764950</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/RADIOELEKTRONIKA54537.2022.9764950" target="_blank" >10.1109/RADIOELEKTRONIKA54537.2022.9764950</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Deep Learning Pipeline for Chromosome Segmentation

Popis výsledku v původním jazyce

Chromosome segmentation is a challenging and time-consuming part of karyotyping and requires a high level of expertise. Computer segmentation algorithms still require the assistance of cytologists in more complicated cases with overlapping or touching chromosomes. Deep learning models have the potential to make the segmentation process completely automated, and their applications are currently actively re-searched. This paper proposes a segmentation pipeline by using deep learning models and traditional computer vision algorithms. This process can be split into four steps, in which we use U-Net architecture to remove any background noises of the metaphase image. Next, we use thresholding and skeletonization to extract and classify single chromosomes and chromosome clusters. As a final step, we use Mask R-CNN, for instance, segmentation on the overlapping and touching chromosomes, and apply test-time augmentation to improve the model's precision.

Název v anglickém jazyce

Deep Learning Pipeline for Chromosome Segmentation

Popis výsledku anglicky

Chromosome segmentation is a challenging and time-consuming part of karyotyping and requires a high level of expertise. Computer segmentation algorithms still require the assistance of cytologists in more complicated cases with overlapping or touching chromosomes. Deep learning models have the potential to make the segmentation process completely automated, and their applications are currently actively re-searched. This paper proposes a segmentation pipeline by using deep learning models and traditional computer vision algorithms. This process can be split into four steps, in which we use U-Net architecture to remove any background noises of the metaphase image. Next, we use thresholding and skeletonization to extract and classify single chromosomes and chromosome clusters. As a final step, we use Mask R-CNN, for instance, segmentation on the overlapping and touching chromosomes, and apply test-time augmentation to improve the model's precision.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20204 - Robotics and automatic control

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2022

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

2022 32nd International Conference Radioelektronika (RADIOELEKTRONIKA)

ISBN

978-1-7281-8686-3

ISSN

—

e-ISSN

—

Počet stran výsledku

5

Strana od-do

197-201

Název nakladatele

IEEE

Místo vydání

neuveden

Místo konání akce

Košice

Datum konání akce

21. 4. 2022

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000856002200041