Added value of multimodal MRI to the clinical diagnosis of primary progressive aphasia variants. (April 2019)

• Record Type: Journal Article
• Authors: Canu, Elisa; Agosta, Federica; Imperiale, Francesca; Fontana, Andrea; Caso, Francesca; Spinelli, Edoardo Gioele; Magnani, Giuseppe; Falini, Andrea; Comi, Giancarlo; Filippi, Massimo
• Is Part Of: Cortex. Volume 113(2019)
• Page Start: 58
• Page End: 66
• Publication Date: 2019-04
• Subjects: Primary progressive aphasia (PPA) -- Nonfluent PPA variant -- Logopenic PPA variant -- Cortical thickness -- White matter tract damage; Neuropsychology -- Periodicals; Nervous system -- Periodicals; Neurology -- Periodicals; Psychophysiology -- Periodicals; Behavior -- Periodicals; Neurology -- Periodicals; 612.825
• DOI: 10.1016/j.cortex.2018.11.025
• Languages: English
• ISSNs: 0010-9452

Abstract

Abstract: Objective: To determine the added value of multimodal structural magnetic resonance imaging (MRI) to language assessment in the differential diagnosis of primary progressive aphasia (PPA) variants. Methods: 59 PPA patients [29 nonfluent (nfvPPA), 15 semantic (svPPA), 15 logopenic (lvPPA)] and 38 healthy controls underwent 3D T1-weighted and diffusion tensor (DT) MRI. PPA patients also performed a comprehensive language assessment. Cortical thickness measures and DT MRI indices of white matter tract integrity were obtained. A random forest analysis identified MRI features associated with each clinical variant. Using ROC curves, the discriminatory power of the language features alone ("language model") and the added contribution of multimodal MRI variables were assessed ("language + MRI model"). Results: The 'language model' alone was able to differentiate svPPA from both nfvPPA and lvPPA patients with high accuracy (area under the curve [AUC] = .95 and .99, respectively). When left inferior parietal cortical thickness and DT MRI metrics of the genu of the corpus callosum and left frontal aslant tract were added to the "language model", the ability to discriminate between nfvPPA and lvPPA cases increased from AUC .82 ("language model" only) to .94 ("language + MRI model"). Conclusions: Language measures alone are able to distinguish svPPA from the other two PPA variants with the highest accuracy. Multimodal structural MRI improves the distinction of nfvPPA and lvPPA, which is challenging in the clinical practice.