Individual analysis of molecular brain imaging data through automatic identification of abnormality patterns

TY - JOUR

T1 - Individual analysis of molecular brain imaging data through automatic identification of abnormality patterns

AU - Burgos, Ninon

AU - Samper-González, Jorge

AU - Bertrand, Anne

AU - Habert, Marie Odile

AU - Ourselin, Sébastien

AU - Durrleman, Stanley

AU - Cardoso, M. Jorge

AU - Colliot, Olivier

PY - 2017/1/1

Y1 - 2017/1/1

N2 - We introduce a pipeline for the individual analysis of positron emission tomography (PET) data on large cohorts of patients. This pipeline consists for each individual of generating a subject-specific model of healthy PET appearance and comparing the individual’s PET image to the model via a novel regularised Z-score. The resulting voxel-wise Z-score map can be interpreted as a subject-specific abnormality map that summarises the pathology’s topographical distribution in the brain. We then propose a strategy to validate the abnormality maps on several PET tracers and automatically detect the underlying pathology by using the abnormality maps as features to feed a linear support vector machine (SVM)-based classifier. We applied the pipeline to a large dataset comprising 298 subjects selected from the ADNI2 database (103 cognitively normal, 105 late MCI and 90 Alzheimer’s disease subjects). The high classification accuracy obtained when using the abnormality maps as features demonstrates that the proposed pipeline is able to extract for each individual the signal characteristic of dementia from both FDG and Florbetapir PET data.

AB - We introduce a pipeline for the individual analysis of positron emission tomography (PET) data on large cohorts of patients. This pipeline consists for each individual of generating a subject-specific model of healthy PET appearance and comparing the individual’s PET image to the model via a novel regularised Z-score. The resulting voxel-wise Z-score map can be interpreted as a subject-specific abnormality map that summarises the pathology’s topographical distribution in the brain. We then propose a strategy to validate the abnormality maps on several PET tracers and automatically detect the underlying pathology by using the abnormality maps as features to feed a linear support vector machine (SVM)-based classifier. We applied the pipeline to a large dataset comprising 298 subjects selected from the ADNI2 database (103 cognitively normal, 105 late MCI and 90 Alzheimer’s disease subjects). The high classification accuracy obtained when using the abnormality maps as features demonstrates that the proposed pipeline is able to extract for each individual the signal characteristic of dementia from both FDG and Florbetapir PET data.

UR - http://www.scopus.com/inward/record.url?scp=85029587569&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-67564-0_2

DO - 10.1007/978-3-319-67564-0_2

M3 - Conference paper

AN - SCOPUS:85029587569

SN - 0302-9743

VL - 10555 LNCS

SP - 13

EP - 22

JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

T2 - 5th International Workshop on Computational Methods for Molecular Imaging, CMMI 2017, 2nd International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2017 and 1st International Stroke Workshop on Imaging and Treatment Challenges, SWITCH 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017

Y2 - 14 September 2017 through 14 September 2017

ER -