Automatic Segmentation of Multi-Contrast MRI Using Statistical Region Merging

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26220%2F14%3APU110262" target="_blank" >RIV/00216305:26220/14:PU110262 - isvavai.cz</a>

Nalezeny alternativní kódy

RIV/68081731:_____/14:00432483

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Automatic Segmentation of Multi-Contrast MRI Using Statistical Region Merging

Popis výsledku v původním jazyce

Several methods have been developed for segmentation of MR images. Some of them are fully automated and some of them rely on an expert's assistance, such as determination of a starting point etc. The fully automated methods are usually based on prior knowledge of a given object and can be used only for particular problem. The purpose of the proposed method is a fully automatic segmentation for general MR images independent on the number of tissues present. The proposed method is based on Statistical Region Merging (SRM) algorithm developed by Richard Nock and Frank Nielsen in 2004. The suitable MR contrasts for this algorithm, as it was confirmed during the test phase, are T1, T2 and FLAIR images. The segmentation process divides to image into regions according the properties in the area, but it does not consider the unconnected areas. For this reason, the algorithm is repeated for created segments without a joint border condition. The algorithm was tested on 5000 axial images with resolution 256x256 pixels. In 2256 slices, the tumor was present. Since the proposed method is fully automatic and independent of image intensities, each image of the database can be considered as unique and independent of others. The Dice coefficient for tissue segmentation varies for particular tissues. The best average result was achieved for grey matter, where the dice coefficient reached value 0.84. The results show the suitability of SRM method for multi-contrast MRI segmentation.

Název v anglickém jazyce

Automatic Segmentation of Multi-Contrast MRI Using Statistical Region Merging

Popis výsledku anglicky

Several methods have been developed for segmentation of MR images. Some of them are fully automated and some of them rely on an expert's assistance, such as determination of a starting point etc. The fully automated methods are usually based on prior knowledge of a given object and can be used only for particular problem. The purpose of the proposed method is a fully automatic segmentation for general MR images independent on the number of tissues present. The proposed method is based on Statistical Region Merging (SRM) algorithm developed by Richard Nock and Frank Nielsen in 2004. The suitable MR contrasts for this algorithm, as it was confirmed during the test phase, are T1, T2 and FLAIR images. The segmentation process divides to image into regions according the properties in the area, but it does not consider the unconnected areas. For this reason, the algorithm is repeated for created segments without a joint border condition. The algorithm was tested on 5000 axial images with resolution 256x256 pixels. In 2256 slices, the tumor was present. Since the proposed method is fully automatic and independent of image intensities, each image of the database can be considered as unique and independent of others. The Dice coefficient for tissue segmentation varies for particular tissues. The best average result was achieved for grey matter, where the dice coefficient reached value 0.84. The results show the suitability of SRM method for multi-contrast MRI segmentation.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2014

Kód důvěrnosti údajů

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

Název statě ve sborníku

PIERS 2014 Guangzhou Proceedings

ISBN

978-1-934142-28-8

ISSN

1559-9450

e-ISSN

—

Počet stran výsledku

5

Strana od-do

1865-1869

Název nakladatele

Neuveden

Místo vydání

Guangzhou

Místo konání akce

Guangzhou

Datum konání akce

25. 8. 2014

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000393225900413