Low-contrast lesion segmentation in advanced MRI experiments by time-domain Ricker-type wavelets and fuzzy 2-means
Identifikátory výsledku
Kód výsledku v IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00179906%3A_____%2F22%3A10443959" target="_blank" >RIV/00179906:_____/22:10443959 - isvavai.cz</a>
Nalezeny alternativní kódy
RIV/62690094:18450/22:50019071 RIV/00216208:11150/22:10443959
Výsledek na webu
<a href="https://verso.is.cuni.cz/pub/verso.fpl?fname=obd_publikace_handle&handle=_DE62HpImo" target="_blank" >https://verso.is.cuni.cz/pub/verso.fpl?fname=obd_publikace_handle&handle=_DE62HpImo</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1007/s10489-022-03184-1" target="_blank" >10.1007/s10489-022-03184-1</a>
Alternativní jazyky
Jazyk výsledku
angličtina
Název v původním jazyce
Low-contrast lesion segmentation in advanced MRI experiments by time-domain Ricker-type wavelets and fuzzy 2-means
Popis výsledku v původním jazyce
Automated suspicious region segmentation has become a crucial need for the experts dealing with numerous images containing contrast-based lesions in MRI. Not all solutions, however, are based on mathematical infrastructure or providing adequate flexibility. On the other hand, segmentation of low-contrast lesions is very challenging for researchers; therefore, advanced magnetic resonance imaging (MRI) experiments are not commonly used in researches. Given the need of repeatability and adaptability, we present an automated framework for intelligent segmentation of brain lesions by wavelet imaging and fuzzy 2-means. Besides the general use of the wavelets in image processing, which is edge detection; we employed the second-order Ricker-type wavelets as the core of our novel imaging framework for low-contrast lesion segmentation. We firstly introduced the mathematical basis of several Ricker wavelet functions, which are in symmetrical form satisfying finite-energy and admissibility conditions of mother wavelets. Afterwards, we investigated three types of Ricker wavelets to apply on our clinical dataset containing susceptibility-weighted (SW) and minimum intensity projection SW (mIP-SW) images with barely-visible lesions. Finally, we adjusted the system parameters of the wavelets for optimization and post-segmentation by fuzzy 2-means. According to the preliminary results of the clinical experiments we conducted, our framework provided 93.53% average dice score (DSC) for SWI by Ricker-3 and 92.56% for mIP-SWI by Ricker-2 wavelet, as the main performance criteria of segmentation. Despite the lack of SWI or mIP-SWI experiments in the public datasets, we tested our framework with BraTS 2012 training sets containing real images with visible lesions and achieved an average of 88.13% DSC with 11.66% standard deviation by re-optimized framework for whole lesion segmentation, which is one of the highest among other relevant researches. In detail, 87.52% DSC for LG datasets with 11.32% standard deviation; while 88.34% DSC for HG datasets with 11.77% standard deviation are calculated.
Název v anglickém jazyce
Low-contrast lesion segmentation in advanced MRI experiments by time-domain Ricker-type wavelets and fuzzy 2-means
Popis výsledku anglicky
Automated suspicious region segmentation has become a crucial need for the experts dealing with numerous images containing contrast-based lesions in MRI. Not all solutions, however, are based on mathematical infrastructure or providing adequate flexibility. On the other hand, segmentation of low-contrast lesions is very challenging for researchers; therefore, advanced magnetic resonance imaging (MRI) experiments are not commonly used in researches. Given the need of repeatability and adaptability, we present an automated framework for intelligent segmentation of brain lesions by wavelet imaging and fuzzy 2-means. Besides the general use of the wavelets in image processing, which is edge detection; we employed the second-order Ricker-type wavelets as the core of our novel imaging framework for low-contrast lesion segmentation. We firstly introduced the mathematical basis of several Ricker wavelet functions, which are in symmetrical form satisfying finite-energy and admissibility conditions of mother wavelets. Afterwards, we investigated three types of Ricker wavelets to apply on our clinical dataset containing susceptibility-weighted (SW) and minimum intensity projection SW (mIP-SW) images with barely-visible lesions. Finally, we adjusted the system parameters of the wavelets for optimization and post-segmentation by fuzzy 2-means. According to the preliminary results of the clinical experiments we conducted, our framework provided 93.53% average dice score (DSC) for SWI by Ricker-3 and 92.56% for mIP-SWI by Ricker-2 wavelet, as the main performance criteria of segmentation. Despite the lack of SWI or mIP-SWI experiments in the public datasets, we tested our framework with BraTS 2012 training sets containing real images with visible lesions and achieved an average of 88.13% DSC with 11.66% standard deviation by re-optimized framework for whole lesion segmentation, which is one of the highest among other relevant researches. In detail, 87.52% DSC for LG datasets with 11.32% standard deviation; while 88.34% DSC for HG datasets with 11.77% standard deviation are calculated.
Klasifikace
Druh
J<sub>imp</sub> - Článek v periodiku v databázi Web of Science
CEP obor
—
OECD FORD obor
30103 - Neurosciences (including psychophysiology)
Návaznosti výsledku
Projekt
<a href="/cs/project/EF18_069%2F0010054" target="_blank" >EF18_069/0010054: IT4Neuro(degeneration)</a><br>
Návaznosti
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)
Ostatní
Rok uplatnění
2022
Kód důvěrnosti údajů
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Údaje specifické pro druh výsledku
Název periodika
Applied Intelligence
ISSN
0924-669X
e-ISSN
1573-7497
Svazek periodika
52
Číslo periodika v rámci svazku
13
Stát vydavatele periodika
NL - Nizozemsko
Počet stran výsledku
22
Strana od-do
nestrankovano
Kód UT WoS článku
000767910400003
EID výsledku v databázi Scopus
2-s2.0-85126105197