Data-driven algorithm for the diagnosis of behavioral variant frontotemporal dementia

Abstract

Abbreviations FTD: frontotemporal dementia GRN: progranulin MAPT: microtubule-associated protein tau C9orf72 : chromosome 9 open reading frame 72 bvFTD: behavioural variant of frontotemporal dementia MRI: magnetic resonance imaging CNCs: cognitively normal controls DBM: deformation-based morphometry FTLDNI: frontotemporal lobar degeneration neuroimaging initiative T1-w: T1 weighted GENFI: Genetic frontotemporal dementia initiative MMSE: Mini mental state examination MOCA: Montreal cognitive assessment FTLD-CDR: Frontotemporal lobar degeneration Clinical Dementia Rating score CGI: Clinical global impression FRS: Frontotemporal dementia rating scale FDR: False Discovery Rate PCA: Principal component analysis PCs: Principal components SF: Semantic fluency ROC: Receiver operating characteristic curves AUC: Area under the curve LR+: positive likelihood ratio LR-: negative likelihood ratio Abstract INTRODUCTION: Brain structural imaging is paramount for the diagnosis of behavioral variant of frontotemporal dementia (bvFTD), but it has low sensitivity leading to erroneous or late diagnosis. METHODS:A total of 515 subjects from two different bvFTD databases (training and validation cohorts) were included to perform voxel-wise deformation-based morphometry analysis to identify regions with significant differences between bvFTD and controls. A random forest classifier was used to individually predict bvFTD from morphometric differences in isolation and together with bedside cognitive scores. RESULTS:Average ten-fold cross-validation accuracy was 89% (82% sensitivity, 93% specificity) using only MRI and 94% (89% sensitivity, 98% specificity) with the addition of semantic fluency. In a separate validation cohort of genetically confirmed bvFTD, accuracy was 88% (81% sensitivity, 92% specificity) with MRI and 91% (79% sensitivity, 96% specificity) with added cognitive scores. DISCUSSION:The random forest classifier developed can accurately predict bvFTD at the individual subject level.

Diagnosis of Alzheimer's disease and behavioural variant frontotemporal dementia with machine learning‐aided neuropsychological assessment using feature engineering and genetic algorithms

“…Many studies applying machine learning in the diagnosis of AD and related disorders consider only dichotomous diagnoses (e.g., 40–42 ). As shown in this study, this approach allows reaching a higher diagnostic accuracy.…”

Ventricular features as reliable differentiators between bvFTD and other dementias

“…Further, all of the image processing tools used in this study have been established and validated for use in multi-center and multi-scanner datasets 15,[25][26][27][28][29] , and have been designed to minimize such differences. In addition, all the voxel-wise analyses (which were most likely to be affected by such differences) compared the patient groups versus normal control participants from the same study; i.e.…”

“…MMSE 30 [29][30] 25 [22][23][24][25][26][27] 26 [22][23][24][25][26][27][28] 27 [21][22][23][24][25][26][27][28] 29 [29][30] 28 [27][28][29] 24 [21][22][23][24][25] <0.001…”

“…MoCA 28 [25][26][27][28][29] 19 [12][13][14][15][16][17][18][19][20][21][22][23] 21 [17][18][19][20][21][22] 21 [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25] 26 [25][26][27][28] 23 [21][22][23][24][25][26] 19 [14][15][16][17][18]…”