Application of the dual stream model to neurodegenerative disease: evidence from a multivariate classification tool in primary progressive aphasia

ABSTRACT

Background:

A clinical diagnosis of primary progressive aphasia relies on behavioral characteristics and patterns of atrophy to determine a variant: logopenic; nonfluent/agrammatic; or semantic. The dual stream model is a contemporary paradigm that has been applied widely to understand brain-behavior relationships; however, applications to neurodegenerative diseases like primary progressive aphasia are limited.

Aims

The primary aim of this study is to determine if the dual stream model can be applied to a neurodegenerative disease, such as primary progressive aphasia, using both behavioral and neuroimaging data.

Methods & Procedures:

We analyzed behavioral and neuroimaging data to apply a multivariate classification tool (support vector machines) to determine if the dual stream model extends to primary progressive aphasia. Sixty-four individuals with primary progressive aphasia were enrolled (26 logopenic variant, 20 nonfluent/agrammatic variant, and 18 semantic variant) and administered four behavioral tasks to assess three linguistic domains (naming, repetition, and semantic knowledge). We used regions of interest from the dual stream model and calculated the cortical volume for gray matter regions and white matter structural volumes and fractional anisotropy. We applied a multivariate classification tool (support vector machines) to distinguish variants based on behavioral performance and patterns of atrophy.

Outcomes & Results:

Behavioral performance discriminates logopenic from semantic variant and nonfluent/agrammatic from semantic variant. Cortical volume distinguishes all three variants. White matter structural volumes and fractional anisotropy primarily distinguish nonfluent/agrammatic from semantic variant. Regions of interest that contribute to each classification in cortical and white matter analyses demonstrate alignment of logopenic and nonfluent/agrammatic variants to the dorsal stream, while the semantic variant aligns with the ventral stream.

Conclusions:

A novel implementation of an automated multivariate classification suggests that the dual stream model can be extended to primary progressive aphasia. Variants are distinguished by behavioral and neuroanatomical patterns and align to the dorsal and ventral streams of the dual stream model.

Application of the dual stream model to PPA

KEYWORDS:

• (6): primary progressive aphasia
• dual stream
• atrophy
• diffusion tensor imaging (DTI)
• automated multivariate classification
• support vector machines (SVM)
• PPAprimary progressive aphasia
• lvPPA
• logopenic variant PPA
• nfaPPA
• nonfluent/agrammatic variant PPA
• svPPA
• semantic variant PPA
• roi
• region of interest
• wm
• white matter

Acknowledgments

The authors gratefully acknowledge participants and families who participated in this investigation.

Conflict of Interest: Argye E. Hillis, MD receives compensation from AHA for editorial activities for Stroke and Elsevier for editorial activities for Practice Update Neurology

Funding Sources: National Institutes of Health/National Institute on Deafness and Other Communication Disorders (NIH/NIDCD): R01 DC005375, P50 DC011739, National Institutes of Health/National Institute of Deafness and Communication Disorders (NIH/NIDCD) and the National Institute on Aging: R01 DC011317, National Institutes of Health/National Institute of Deafness and Communication Disorders (NIH/NIDCD): R00 DC015554.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed here.

Participants were asked to name 30 line drawings of objects orally on the short form of the BNT (score range 0–30). Objects ranged from high familiarity items, such as “bed,” to low frequency items, such as “sphinx.” If a participant experienced difficulty naming a pictured object, a phonemic cue was provided; however, these responses were not included in the total correct.

On the SRT, patients were asked to repeat a sentence exactly as the clinician says it (score range 0–5). One repetition was permitted if the patient did not hear the sentence and asked the examiner to repeat the sentence. The total score is the number of accurately repeated sentences.

Participants were asked to identify a line drawing of an object semantically related to target pictured objects on the PPTT (score range 0–14). Fourteen trios of pictures were presented. Participants were asked to point to the picture (of two) related to the stimulus. For example, the stimulus picture of eyeglasses (spectacles) was presented along with pictures depicting an eye and an ear. Participants were asked to decide which picture is more closely related in meaning to the stimulus picture (eyeglasses).

The next subtest, SWPM, evaluated spoken word recognition and assessed the frequency of semantic errors in single word comprehension. Participants were asked to identify a picture matching an orally presented stimulus (score range 0–20). Visual stimuli consisted of displays of four pictures of semantically related objects. For example, a set of four pictures, witch, ghost, pumpkin and bat, were presented and the participant was asked to point to the picture matching the stimulus “ghost.” Five sets of four picture displays are each presented four times (once for each picture as the target) for a total of 20 trials. The location of the target picture is counterbalanced across all of the trials and the order of presentation is pseudo-randomized across all trials.

Additional information

Funding

This work was supported by the National Institute on Aging [R01 DC011317]; National Institute on Deafness and Other Communication Disorders [P50 DC014664,R00 DC01554,R01 DC005375].