E-infrastructure, cortical mesh segmentation, quality control environment : a red thread for neuroscientists

Abstract : Neuroscience entered the “big data” era. Individual desktop computers are no longer suitable to analyse terabyte, and potentially petabytes, of brain images. To fill in the gap between data acquisition and information extraction, e-infrastructures are being developing in North America, Canada, and Europe. E-infrastructures allow neuroscientists to conduct neuroimaging experiments using dedicated computational resources such as grids, high-performance computing (HPC) systems, and public/private clouds. Today, e-infrastructures are the most advanced and the best equipped systems to support the creation of advanced multimodal and multiscale models of the AD brain (chapter 2) or to validate promising imaging biomarkers with sophisticated pipelines, as for cortical thickness, (chapter 3). Indeed, imaging analyses such as those described in chapter 2 and 3 expand the amount of post-processed data per single study. In order to cope with the huge amount of post-processing data generated via e-infrastructures, an automatic quality control environment (QCE) of the cortical delineation algorithms is proposed (chapter 4). QCE is a machine learning (ML) classifier with a supervised learning approach based on Random Forest (RF) and Support Vector Machine (SVM) estimators. Given its scalability and efficacy, QCE fits well in the e-infrastructures under development, where this kind of sanity check service is still lacking. QCE represents a unique opportunity to process data more easily and quickly, allowing neuroscientists to spend their valuable time do data analysis instead of using their resources in manual quality control work.
Document type :
Theses
Complete list of metadatas

Cited literature [158 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02182658
Contributor : Abes Star <>
Submitted on : Saturday, July 13, 2019 - 1:01:21 AM
Last modification on : Saturday, July 13, 2019 - 1:11:39 AM

File

74523_REDOLFI_2017_diffusion.p...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02182658, version 1

Collections

STAR | CEA

Citation

Alberto Redolfi. E-infrastructure, cortical mesh segmentation, quality control environment : a red thread for neuroscientists. Medical Imaging. Université Paris-Saclay, 2017. English. ⟨NNT : 2017SACLS191⟩. ⟨tel-02182658⟩

Share

Metrics

Record views

209

Files downloads

18