Premium
MRI‐based automated volumetric segmentation tool in the detection of early Alzheimer’s disease
Author(s) -
Liu Wanting,
Au Lisa W.C.,
Abrigo Jill,
Luo Yishan,
Wong Adrain,
Kwan Pauline,
Ma Alison Hon Wing,
Ng Anthea Yee Tung,
Chen Sirong,
Leung Eric Y.L.,
Ho Chi Lai,
Chu Winnie C.W.,
Ko Ho,
Shi Lin,
Mok Vincent C.T.
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.042340
Subject(s) - positron emission tomography , magnetic resonance imaging , atrophy , nuclear medicine , medicine , cognitive impairment , alzheimer's disease , segmentation , pathology , radiology , disease , artificial intelligence , computer science
Background Cognitive unimpaired (CU) and mild cognitive impairment (MCI) subjects harboring beta‐amyloid (A+) and tau (T+) are at high‐risk for incident cognitive decline. Fluoro‐deoxyglucose (FDG) positron emission tomography (PET) likely represents both cumulative loss of the neuropil and functional impairment of neurons while magnetic resonance imaging (MRI) indicates cumulative loss and shrinkage of the neuropil. As PET and CSF analysis in detecting beta‐amyloid and tau are not easily accessible (either costly or invasive nature or radiation), we compared the performance of MRI‐based automated segmentation tool with FDG PET in identifying CU and MCI subjects harboring A+T+. Method A total of 62 subjects (CU=37, MCI=25) underwent MRI, FDG PET, 11 C‐ PIB and 18 F‐T807 PET. MRIs were processed by an automated segmentation tool to obtain an Alzheimer’s Disease (AD) ‐ resemblance atrophy index (AD‐RAI) which was derived from a machine learning method (AccuBrain®) that analyzed multiple brain regions relevant to AD. Result AD‐RAI yielded the same sensitivity (0.73) among all subjects with FDG PET while they had similar specificity. AD‐RAI performed the highest sensitivity (0.91) in MCI subjects over FDG PET (0.82) with an acceptable specificity (0.79) which was slightly lower than that of FDG PET. However, both AD‐RAI and FDG PET were suboptimal in CU group. Detailed results are demonstrated in the Table. Conclusion The volumetric segmentation tool achieves a good diagnostic profile similar to FDG PET in identifying early AD. It has the potential to be more widely used as a tool to diagnose early AD for receiving treatment or recruitment into clinical trials.