Professional Documents
Culture Documents
1 Department of Communication Sciences and Disorders, University of South Carolina, Columbia, USA
2 Department of Neurology, Johns Hopkins University School of Medicine; Department of Cognitive Science, Johns Hopkins
University, Baltimore, USA
3 Department of Physical Medicine and Rehabilitation, Johns Hopkins School of Medicine; Department of Cognitive Science, Johns
Hopkins University, Baltimore, MD, USA
4 Cognitive Sciences, School of Social Sciences, University of California, Irvine, USA
5 Department of Psychology, University of South Carolina, Columbia, USA
6 Department of Neurology, Medical University of South Carolina, Charleston, South Carolina, USA
Correspondence to: Julius Fridriksson
Professor and Endowed Chair, SmartState
Department of Communication Sciences and Disorders
University of South Carolina
Columbia, USA
E-mail: jfridrik@sc.edu
Received January 28, 2017. Revised October 26, 2017. Accepted November 8, 2017. Advance Access publication January 17, 2018
ß The Author(s) (2018). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved.
For Permissions, please email: journals.permissions@oup.com
Anatomy of aphasia BRAIN 2018: 141; 848–862 | 849
Table 1 List of measures included in RLSM and CLSM analyses and the domain of communication assessed
AOS = apraxia of speech; ASRS = Apraxia of Speech Rating Scale; WAB = Western Aphasia Battery.
that tests requiring the involvement and coordination of dementia or other neurological problems (as per self/caregiver
several different processes would involve both streams. or medical report). All participants were recruited through
The methodology used here involved both region-wise local advertisement and were enrolled at the University of
lesion symptom mapping (RLSM) and connectome lesion- South Carolina or at the Medical University of South
Carolina. They provided informed consent to participate in
symptom mapping (CLSM; Yourganov et al., 2016). CLSM
this study, which was approved by the Institutional Review
uses the same statistical approach as traditional lesion-
Boards at the University of South Carolina and at the
symptom mapping methods with one major exception: in- Medical University of South Carolina.
stead of relating a given lesion location to impairment,
CLSM associates damage involving white matter connec-
tions between anatomical regions and behavioural impair- Speech and language testing
ment. Thus, rather than isolating lesion locations, CLSM The behavioural battery included 16 tests and tasks selected to
makes it possible to reveal cortical networks that are cru- reflect speech and language impairments typically included as
cial for performing a given task. For both RLSM and part of the clinical management of aphasia in the USA
CLSM, we carried out univariate and multivariate analyses (Table 1). It was not meant to reflect in-depth assessment of
to provide slightly different perspectives based on different linguistic processing or neurophysiology of speech, as the
data analysis approaches of the same data. intended audience here consists of clinicians and clinician-sci-
entists who might be more interested in the effects of stroke-
related brain damage on communication skills that might
reflect real-life situations (e.g. speech fluency and overall sever-
Materials and methods ity of aphasia). However, we selected tests and tasks that pri-
marily tax motor speech processes with relatively minimal
speech comprehension requirements as well as the opposite;
Participants tests that focus specifically on speech recognition and compre-
The data for this project were obtained from an archival data- hension with little or no input from motor speech. This
base in the Aphasia Lab, University of South Carolina and allowed us to compare and contrast cortical network damage
Medical University of South Carolina. Participants had sus- that primarily affects motor speech versus auditory compre-
tained a single-event stroke to the left hemisphere at least 6 hension, two aspects of processing that are commonly assessed
months prior to study inclusion and were tested either as part as part of a comprehensive aphasia work-up. In addition,
of an aphasia treatment study or strictly for the purpose of other typical subtests included on aphasia test batteries were
lesion-symptom mapping research. A lesion overlay map is used to test functions such as reading, writing, speech repeti-
included in Supplementary Fig. 1. The average time post- tion, verbal naming, and grammatical processing of sentences.
stroke was 36.4 months [standard deviation (SD) = 43.1]. Unfortunately, our dataset did not include a comprehensive
The total sample size was 159 chronic stroke survivors and measure of phonological input, something that is clearly a
the number of tested participants varied across the different focus of Hickok and Poeppel’s dual stream model.
assessment batteries used here to assess speech and language The speech and language battery included six tests or rating
impairment. All participants were native speakers of English scales from the Western Aphasia Battery (Kertesz, 1982):
with a mean age of 60.0 years (SD = 11.2), and 68 were Aphasia Quotient, a 0–100 point aphasia severity scale;
female. Participants were excluded if they had a history of Speech Fluency, a 10-point qualitative rating scale of speech
Anatomy of aphasia BRAIN 2018: 141; 848–862 | 851
production; Speech Repetition, a 60-point test of speech repe- included nine B = 0 s/mm2. Therefore, in total we acquired 60
tition of real words and progressively longer sentences; volumes with B = 1000, 60 with B = 2000 and 11 with B = 0.
Auditory Word Recognition, a 60-point test where the partici-
pant is required to point to a picture or object that corres-
ponds to spoken words presented by a clinician (picture and
Image preprocessing
object targets are presented along with five distractors);
Lesions
Reading ability, a collection of nine subtests with a maximum
possible score of 20 points; and Writing ability, a collection of The chronic stroke lesion was demarcated on T2-weighted
seven subtests with a maximum possible score of 20 points. images by a neurologist (L.B.), who was blinded to the par-
Three scores were derived from the Philadelphia Naming Test ticipants’ language scores. The T2 image was co-registered to
(PNT), a 175-item test of picture naming (Roach et al., 1996): the T1 image, and these parameters were used to reslice the
correct naming; phonological errors; and semantic errors. lesion into the native T1 space. The resliced lesion maps were
Assessment of auditory sentence comprehension relied on a smoothed with a 3 mm full-width at half-maximum Gaussian
45-item test where participants were required to match a clin- kernel to remove jagged edges associated with manual draw-
attenuate the distorting effects of stroke-related necrotic analyses relied on conventional lesion symptom mapping:
changes on the brain anatomy and fibre tracking. The prepro- General Linear Model (GLM) (pooled variance t-test) with
cessing approach also excluded the lesion site from all tracto- P 5 0.05 (one-tailed) and control for multiple comparisons
graphy tracings (including cortical seeding, cortical waypoints used permutation threshholding (4000 permutations).
or white matter tracking regions) thus controlling for lesion Similarly, the univariate CLSM analyses also relied on GLM
influence on the final tract tracing. For this reason, steps to with P 5 0.05 and 4000 permutations to control for multiple
control for the overall lesion size were not included in subse- comparisons. A handful of studies have shown that aphasia
quent statistical analyses. These steps are described in more severity changes considerably among some patients, even in the
detail below. chronic phase (Naeser et al., 1998; Holland et al., 2017; Hope
To align the diffusion image with the lesion map, the T2- et al., 2017). To verify that time post-stroke would not influ-
weighted image (co-registered to match the T1-weighted image, ence the results, we examined the correlation between each of
and therefore in the same space as the native T1 image) was the 16 tasks and tests and time post-stroke. No statistically
normalized to match the non-diffusion weighted image from significant correlations were revealed. Therefore, the lesion-
the diffusion MRI sequence (the B0 image) and the resulting symptom analyses were not adjusted for time post-stroke.
Table 2 The top three brain regions predictive in the damaged link. Other functions such as ‘auditory syllable
univariate analyses of each test/task and the z-score discrimination’, ‘speech rate’ and ‘apraxia of speech’
associated with brain damage and each given region involved relatively few links. Not surprisingly, the number
of cortical regions implicated in the univariate RLSM ana-
Test/scale Region Z-Score
lyses was correlated with the number of links revealed in
Aphasia Quotient STG 7.84 the CLSM analyses, r = 0.86, P 5 0.0001. However, the
Posterior Ins 7.65
number of participants included in the analyses of each
Posterior STG 7.61
behavioural test or task was not correlated with the
Speech Fluency STG 7.74 number of regions revealed by RLSM (r = 0.38, P = 0.17)
IFG opercularis 7.73 or CLSM (r = 0.46, P = 0.075). To appreciate which cor-
Posterior Ins 7.71 tical regions were most often included as part of a damaged
Speech Repetition Posterior STG 7.78 link related to test performance, we counted how often
STG 7.69 each region of interest occurred in the CLSM results.
in both lesion maps. The multivariate RLSM analysis of Aphasia Battery, is associated with extensive cortical net-
components 1 and 2 also revealed regions that mostly work damage, mostly involving the dorsal stream and, to a
loaded onto the dorsal and ventral stream regions of inter- lesser extent, the ventral stream. This is perhaps not sur-
est with only one region, middle frontal gyrus, included in prising as the aphasia quotient is heavily weighted for
both lesion maps. For both the univariate and multivariate speech production where three out of the four factors
RLSM analyses, the precentral gyrus was the strongest pre- that comprise aphasia quotient (speech fluency, speech
dictor of component 1 and the posterior STG was the repetition, and naming) rely on speech production. Given
strongest predictor of component 2. Somewhat less conver- that the rating of ‘speech fluency’ and ‘speech repetition’
gence across the univariate and multivariate analyses was ability are two of the subtests that comprise aphasia quo-
revealed for the CLSM results. A link between the precen- tient, it stands to reason that there would be considerable
tral gyrus and the middle frontal gyrus was the best pre- overlap among the regions and connections that predict
dictor of component 1, whereas two different links were each of these factors. Aphasia can involve different degrees
identified as the strongest predictor of component 2 in the of impairment to multiple speech and language processes
univariate (angular gyrus $ SMG) and multivariate ana- that subserve communication functioning. As such, even
lyses (posterior STG $ MTG). Both the univariate and relatively smaller strokes that affect the cortical language
multivariate CLSM analyses for component 1 mainly high- network, especially if the affected areas involve links be-
lighted links between dorsal stream regions. Although the tween the dorsal and ventral streams, seem likely to cause
multivariate CLSM analysis mostly revealed links across aphasia lasting beyond the subacute stage.
ventral stream regions, this was not the case for the uni- Whereas traditional RLSM can reveal damage that pre-
variate results, which included links between both dorsal dicts a given speech or language impairment, CLSM pro-
and ventral stream regions. vides complementary information highlighting links
between regions that, when damaged, have a particularly
detrimental effect on functioning. CLSM and RLSM are
Discussion inter-related since CLSM is also dependent on lesion loca-
tion: white matter pathways related to areas of cortical
The current study demonstrates that overall aphasia sever- damage are more commonly lesioned. Nonetheless, the spe-
ity, as reflected by the ‘aphasia quotient’ on the Western cific white matter projections from the areas of cortical
856 | BRAIN 2018: 141; 848–862 J. Fridriksson et al.
Table 3 All regions surviving multivariate lesion ana- damage are not systematically mapped by RLSM. CSLM
lyses for each test/task, along with corresponding R2 can thus define the anatomy of subcortical networks that
and R2 change values are related to behaviour. Importantly, CLSM can also dis-
close the relationship between cortical disconnection and be-
Test/scale Region R2 R2
change haviour. Post-stroke cortical damage can be understood as
the combination of direct vascular injury (necrosis or gliosis)
Aphasia Quotient STG 0.43 0.43
and disconnection (Bonilha et al., 2014a, b). However, dis-
PrCG 0.48 0.05
Posterior STG 0.50 0.02
connected areas that are seemingly preserved are not
Posterior Ins 0.51 0.01 mapped by RLSM. By disclosing crucial pathways extending
beyond the stroke lesion, CLSM can identify which remotely
Speech Fluency IFG opercularis 0.40 0.40
disconnected areas are directly related to cognitive function.
Posterior STG 0.53 0.13
In this context, it is worth noting that cortical damage seen
Putamen 0.56 0.02
on clinical scans does not only show a given lesion location
Auditory Word Posterior STG $ MTG 0.17 0.17 Syllable AG $ SMG 0.26 0.26
Recognitiona Identificationa
IFG triangularis $ IFG 0.21 0.04 GP $ IFG orbitalis 0.38 0.12
opercularis ITG $ MTG 0.46 0.09
IFG orbitalis $ MFG 0.25 0.03 STG $ PrCG 0.53 0.07
Posterior MTG $ ITG 0.28 0.03 Ins $ MTG 0.59 0.06
MTG pole $ MFG 0.30 0.03 GP $ STG 0.64 0.04
Posterior Ins $ MTG pole 0.33 0.03
Sentence Ins $ ITG 0.26 0.26
Posterior STG $ PrCG 0.36 0.03
Comprehensiona
Posterior MTG $ IFG 0.38 0.02
Posterior MTG $ PoCG 0.34 0.08
orbitalis
Putamen $ Ins 0.40 0.07
Readinga IFG triangularis $ IFG 0.32 0.32 Putamen $ STG pole 0.45 0.05
opercularis SMG $ IFG triangularis 0.43 0.02
Posterior Ins $ GP 0.39 0.07
MTG pole $ IFG 0.50 0.07
Posterior MTG $ pSTG 0.44 0.04 triangularis
Putamen $ MFG 0.49 0.05
IFG triangularis $ IFG 0.49 0.00 Non-Canonicala STG pole $ SMG 0.12 0.12
orbitalis GP $ Putamen 0.18 0.07
ITG $ STG pole 0.54 0.05
AOSa PrCG $ MFG 0.16 0.16
(continued)
(continued)
860 | BRAIN 2018: 141; 848–862 J. Fridriksson et al.
In conclusion, our findings reveal that clinical tests typ- Caviness V, Makris N, Montinaro E, Sahin N, Bates J, Schwamm L,
ically used to assess aphasia recruit a cortical network com- et al. Anatomy of stroke, Part I an MRI-based topographic and
volumetric system of analysis. Stroke 2002; 33: 2549–56.
posed of the dorsal and ventral streams underlying Cloutman L, Gottesman R, Chaudhry P, Davis C, Kleinman JT,
phonological form-to-articulation (dorsal stream) and Pawlak M, et al. Where (in the brain) do semantic errors come
phonological form-to-meaning (ventral stream), respect- from? Cortex 2009; 45: 641–9.
ively. Speech production is impaired primarily as a result DeLeon J, Gottesman RF, Kleinman JT, Newhart M, Davis C, Heidler-
of damage to the dorsal stream whereas speech comprehen- Gary J, et al. Neural regions essential for distinct cognitive processes
underlying picture naming. Brain 2007; 130 (Pt 5): 1408–22.
sion is more likely associated with ventral stream damage. den Ouden DB, Saur D, Mader W, Schelter B, Lukic S, Wali E, et al.
Nevertheless, many clinical tests of aphasia involve multiple Network modulation during complex syntactic processing.
processes that rely on both streams, which can result in Neuroimage 2012; 59: 815–23.
poor performance due to damage affecting different sec- Dick AS, Bernal B, Tremblay P. The language connectome: new path-
tions of the cortical speech and language network. ways, new concepts. Neuroscientist 2014; 20: 453–67.
Dronkers NF, Plaisant O, Iba-Zizen MT, Cabanis EA. Paul Broca’s
Damage to specific cortical hubs such as Broca’s area,
Hickok G, Okada K, Serences JT. Area Spt in the human planum Ogar JM, Baldo JV, Wilson SM, Brambati SM, Miller BL, Dronkers
temporale supports sensory-motor integration for speech processing. NF, et al. Semantic dementia and persisting Wernicke’s aphasia:
J Neurophys 2009; 101: 2725–32. linguistic and anatomical profiles. Brain Lang 2011; 117: 28–33.
Hickok G, Poeppel D. Dorsal and ventral streams: a framework for Ojemann GA, Whitaker HA. Language localization and variability.
understanding aspects of the functional anatomy of language. Brain Lang 1978; 6: 239–60.
Cognition 2004; 92: 67–99. Poeppel D, Emmorey K, Hickok G, Pylkkänen L. Towards a new
Hickok G, Poeppel D. The cortical organization of speech processing. neurobiology of language. J Neurosci 2012; 32: 14125–31.
Nat Rev Neurosci 2007; 8: 393–402. Rapcsak SZ, Henry ML, Teague SL, Carnahan SD, Beeson PM. Do
Hillis AE, Kleinman JT, Newhart M, Heidler-Gary J, Gottesman R, dual-route models accurately predict reading and spelling perform-
Barker PB, et al. Restoring cerebral blood flow reveals neural re- ance in individuals with acquired alexia and agraphia?
gions critical for naming. J Neurosci 2006; 26: 8069–73. Neuropsychologia 2007; 45: 2519–24.
Hillis AE, Wityk RJ, Tuffiash E, Beauchamp NJ, Jacobs MA, Barker PB, Rauschecker JP, Scott SK. Maps and streams in the auditory cortex:
et al. Hypoperfusion of Wernicke’s area predicts severity of semantic nonhuman primates illuminate human speech processing. Nat
deficit in acute stroke. Ann Neurol 2001; 50: 561–6. Neurosci 2009; 12: 718–24.
Hodges JR, Patterson K, Oxbury S, Funnell E. Semantic dementia. Rauschecker JP, Tian B. Mechanisms and streams for processing of