Premium
P2‐387: A NOVEL DIFFUSION TENSOR IMAGING METHOD TO CLASSIFY FRONTO‐TEMPORAL DEMENTIA SUBTYPES WITH MACHINE LEARNING
Author(s) -
Torso Mario,
Bozzali Marco,
Jenkinson Mark,
Chance Steven A.
Publication year - 2019
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 - 10.1016/j.jalz.2019.06.2794
Subject(s) - frontotemporal dementia , cohort , semantic dementia , neuroimaging , diffusion mri , medicine , dementia , primary progressive aphasia , artificial intelligence , audiology , psychiatry , pathology , disease , computer science , magnetic resonance imaging , radiology
Background: Fronto-temporal dementia (FTD) is one of the most common types of pre-senile dementia characterized by an heterogeneous clinical presentation that includes three main subtypes: behavioural variant (BV), semantic variant (SV) and progressive non fluent aphasia (PNFA). A correct diagnosis is important to better understand the different subtypes and to develop specific treatments. The aim of the study was to test the discrimination power of a novel set of Diffusion Tensor Imaging measures (DTI), on FTD subtypes.Methods: Two cohorts were included; a “training cohort” acquired in Rome and a “test cohort” from the frontotemporal lobar degeneration neuroimaging initiative (FTLDNI). A total of sixtysix patients (24 + 42) with probable FTD (20 bvFTD, 31 SV and 15 PNFA), and fifty-four normal healthy adults (30 + 24) were studied. A machine learning algorithm was selected from a range of seven main models and the “best” was applied to estimate the classification accuracy of a set of new features. The features selected were trained on the “training cohort” and tested on the “test cohort”. Results: The new DTI features distinguished between controls and patients (all variants combined) with an accuracy of 85% and distinguished between FTD subgroups with an accuracy of 75.75 %. Conclusions: These results suggest that DTI measures could be important with MRI playing a more central role in the process of differential diagnosis, improving diagnostic power and potentially aiding the discovery of drug targets, specific for each FTD subgroup.