Aphasia Neuroplasticity Review

Purcell et al. (2019)

Reference

Authors	Purcell JJ, Wiley RW, Rapp B
Title	Re-learning to be different: Increased neural differentiation supports post-stroke language recovery
Reference	NeuroImage 2019; 202: 116145
PMID	31479754
DOI	10.1016/j.neuroimage.2019.116145

Participants

Language	US English
Inclusion criteria	Chronic dysgraphia (acquired impairment in spelling)
Number of individuals with aphasia	21 (plus 4 excluded: 3 health reasons; 1 data acquisition error)
Number of control participants	0
Were any of the participants included in any previous studies?	No
Is age reported for patients and controls, and matched?	Yes (range 40-80 years)
Is sex reported for patients and controls, and matched?	Yes (males: 13; females: 8)
Is handedness reported for patients and controls, and matched?	Yes (right: 16; left: 3; other: 2)
Is time post stroke onset reported and appropriate to the study design?	Yes (range 14-209 months)
To what extent is the nature of aphasia characterized?	Comprehensive battery
Language evaluation	Spelling (PALPA 40 and 54, and other word lists), oral reading (PALPA 35), reading comprehension (PALPA 51), spoken word-picture matching and picture naming tests from Northwestern Naming Battery, PPT-P; note no generic aphasia battery, but fairly complete coverage of language domains
Aphasia severity	Spelling of untrained items range 51%-94%
Aphasia type	4 orthographic working memory deficit, 8 orthographic long-term memory deficit, 9 both types of deficit
First stroke only?	Yes
Stroke type	Not stated
To what extent is the lesion distribution characterized?	Lesion overlay
Lesion extent	Range 7.7-215.0 cc
Lesion location	L MCA with L ventral occipitotemporal cortex mostly intact
Participants notes	—

Imaging

Modality	fMRI
Is the study cross-sectional or longitudinal?	Longitudinal—chronic treatment
If longitudinal, at what time point(s) were imaging data acquired?	T1: pre-treatment/chronic; T2: post-treatment, 6-24 weeks later
If longitudinal, was there any intervention between the time points?	Spelling treatment, 60-80 minutes/day, 2 days/week, range 6-24 weeks
Is the scanner described?	No (not stated)
Is the timing of stimulus presentation and image acquisition clearly described and appropriate?	Yes
Design type	Event-related
Total images acquired	1232 (four runs distributed over two days)
Are the imaging acquisition parameters, including coverage, adequately described and appropriate?	Yes (cerebellum excluded)
Is preprocessing and intrasubject coregistration adequately described and appropriate?	Yes
Is first level model fitting adequately described and appropriate?	No* (moderate limitation) (not feasible to separate closely spaced instruction, word, and letter/response, especially when responses will be compared to rest)
Is intersubject normalization adequately described and appropriate?	Yes
Imaging notes	—

Conditions

Are the conditions clearly described?

Yes

Condition	Response type	Repetitions	All groups could do?	All individuals could do?
spelling probe (training items)	Button press	60	Yes	Unknown
spelling probe (known items)	Button press	60	Yes	Unknown
case verification	Button press	60	Yes	Unknown
rest	None	implicit baseline	N/A	N/A

Conditions notes

Condition 3 not used in any contrasts

Contrasts

Are the contrasts clearly described?

Yes

Contrast 1: spelling probe (training items) vs rest

Language condition	Spelling probe (training items)
Control condition	Rest
Are the conditions matched for visual demands?	No
Are the conditions matched for auditory demands?	No
Are the conditions matched for motor demands?	No
Are the conditions matched for cognitive/executive demands?	No
Is accuracy matched between the language and control tasks for all relevant groups?	N/A, tasks not comparable
Is reaction time matched between the language and control tasks for all relevant groups?	N/A, tasks not comparable
Behavioral data notes	—
Are control data reported in this paper or another that is referenced?	No
Does the contrast selectively activate plausible relevant language regions in the control group?	Unknown
Are activations lateralized in the control data?	Unknown
Control activation notes	Task comes from Rapp and Lipka (2011), who report lateralized activations for the contrast of spelling probes to case verification, but do not report results relative to fixation baseline
Contrast notes	—

Contrast 2: spelling probe (known items) vs rest

Language condition	Spelling probe (known items)
Control condition	Rest
Are the conditions matched for visual demands?	No
Are the conditions matched for auditory demands?	No
Are the conditions matched for motor demands?	No
Are the conditions matched for cognitive/executive demands?	No
Is accuracy matched between the language and control tasks for all relevant groups?	N/A, tasks not comparable
Is reaction time matched between the language and control tasks for all relevant groups?	N/A, tasks not comparable
Behavioral data notes	—
Are control data reported in this paper or another that is referenced?	No
Does the contrast selectively activate plausible relevant language regions in the control group?	Unknown
Are activations lateralized in the control data?	Unknown
Control activation notes	Task comes from Rapp and Lipka (2011), who report lateralized activations for the contrast of spelling probes to case verification, but do not report results relative to fixation baseline
Contrast notes	—

Analyses

Are the analyses clearly described?

No* (moderate limitation) (see specific limitation(s) below)

Voxelwise analysis 1

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal change in aphasia
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	—
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Appear mismatched
Is reaction time matched across the second level contrast?	Appear mismatched
Behavioral data notes	See section S2, but main effects include known items also
Type of analysis	Voxelwise
Search volume	Appears to be restricted to voxels spared in all patients
Correction for multiple comparisons	Clusterwise correction based on 3dClustSim
Software	BrainVoyager QX 2.4 or SPM12
Voxelwise p	.01
Cluster extent	49 voxels (size not stated)
Statistical details	—
Findings	↑ L posterior cingulate ↑ R angular gyrus ↑ R posterior cingulate
Findings notes	—

ROI analysis 1

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on training items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Regions of interest (ROI)
ROI type	Functional
How many ROIs are there?	3
What are the ROI(s)?	(1) R AG; (2) L PCC; (3) R PCC
How are the ROI(s) defined?	Regions activated in SPM analysis 1
Correction for multiple comparisons	No correction
Statistical details	—
Findings	None
Findings notes	—

ROI analysis 2

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on untrained items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Regions of interest (ROI)
ROI type	Functional
How many ROIs are there?	3
What are the ROI(s)?	(1) R AG; (2) L PCC; (3) R PCC
How are the ROI(s) defined?	Regions activated in SPM analysis 1
Correction for multiple comparisons	No correction
Statistical details	—
Findings	None
Findings notes	—

ROI analysis 3

First level contrast	Spelling probe (training items) vs rest
Analysis class	Cross-sectional correlation with language or other measure
Group(s)	Aphasia T1
Covariate	Subsequent Δ spelling accuracy on training items (T2 vs T1)
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Somewhat (T1 behavioral measure should be included in model)
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Region of interest (ROI)
ROI type	Functional
How many ROIs are there?	1
What are the ROI(s)?	L ventral occipitotemporal cortex
How are the ROI(s) defined?	The region that showed an increase in Local-Hreg from T1 to T2
Correction for multiple comparisons	One only
Statistical details	—
Findings	None
Findings notes	—

ROI analysis 4

First level contrast	Spelling probe (training items) vs rest
Analysis class	Cross-sectional correlation with language or other measure
Group(s)	Aphasia with both timepoints T1 (n = 20)
Covariate	Subsequent Δ spelling accuracy on untrained items (T2 vs T1)
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Somewhat (T1 behavioral measure should be included in model)
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Region of interest (ROI)
ROI type	Functional
How many ROIs are there?	1
What are the ROI(s)?	L ventral occipitotemporal cortex
How are the ROI(s) defined?	The region that showed an increase in Local-Hreg from T1 to T2
Correction for multiple comparisons	One only
Statistical details	—
Findings	None
Findings notes	—

ROI analysis 5

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on training items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Region of interest (ROI)
ROI type	Functional
How many ROIs are there?	1
What are the ROI(s)?	L ventral occipitotemporal cortex
How are the ROI(s) defined?	The region that showed an increase in Local-Hreg from T1 to T2
Correction for multiple comparisons	One only
Statistical details	—
Findings	None
Findings notes	—

ROI analysis 6

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on untrained items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Region of interest (ROI)
ROI type	Functional
How many ROIs are there?	1
What are the ROI(s)?	L ventral occipitotemporal cortex
How are the ROI(s) defined?	The region that showed an increase in Local-Hreg from T1 to T2
Correction for multiple comparisons	One only
Statistical details	—
Findings	None
Findings notes	—

Complex analysis 1

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal change in aphasia
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	—
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Appear mismatched
Is reaction time matched across the second level contrast?	Appear mismatched
Behavioral data notes	See section S2, where Figures S1 and S2 appear to show differences; the main effects of time were not significant for accuracy or RT, but those analyses included known items also, which had smaller effects
Type of analysis	Complex
Statistical details	Local Heterogeneity Regression Analysis (Local-Hreg) was used to identify brain regions where the heterogeneity of timecourses between neighboring voxels, specifically for the trained condition, increased from T1 to T2. A voxelwise threshold of p < 0.05 was applied, followed by cluster correction based on permutation testing. The analysis appears to have been restricted to brain regions not damaged in any patients.
Findings	Other
Findings notes	Only in L ventral occipitotemporal cortex, there was a significant increase in Local-Hreg from T1 to T2 (p = 0.028, corrected).

Complex analysis 2

First level contrast	Spelling probe (known items) vs rest
Analysis class	Longitudinal change in aphasia
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	—
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Yes, matched
Is reaction time matched across the second level contrast?	Yes, matched
Behavioral data notes	See section S2, main effects were not significant and effects appear smaller for known than trained
Type of analysis	Complex
Statistical details	Local Heterogeneity Regression Analysis (Local-Hreg) was used to identify brain regions where the heterogeneity of timecourses between neighboring voxels, specifically for the known condition, increased from T1 to T2. A voxelwise threshold of p < 0.05 was applied, followed by cluster correction based on permutation testing. The analysis appears to have been restricted to brain regions not damaged in any patients.
Findings	None
Findings notes	—

Complex analysis 3

First level contrast	Spelling probe (training items) vs rest
Analysis class	Cross-sectional correlation with language or other measure
Group(s)	Aphasia T1
Covariate	T1 spelling accuracy on training items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Somewhat (training items were selected for individual patients, so training item accuracy is not an appropriate measure of spelling ability)
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between Local-Hreg at T1 in the L ventral occipitotemporal region previously identified and T1 spelling accuracy of training items. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	Other
Findings notes	There was a significant positive relationship between T1 Local-Hreg and T1 spelling accuracy on training items.

Complex analysis 4

First level contrast	Spelling probe (training items) vs rest
Analysis class	Cross-sectional correlation with language or other measure
Group(s)	Aphasia T1
Covariate	Subsequent Δ spelling accuracy on training items (T2 vs T1)
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Somewhat (T1 behavioral measure should be included in model)
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between Local-Hreg at T1 in the L ventral occipitotemporal region previously identified and subsequent improvement in spelling accuracy of training items from T1 to T2. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	Other
Findings notes	There was a significant positive relationship between T1 Local-Hreg and subsequent improvement in spelling accuracy on training items from T1 to T2.

Complex analysis 5

First level contrast	Spelling probe (training items) vs rest
Analysis class	Cross-sectional correlation with language or other measure
Group(s)	Aphasia with both timepoints T1 (n = 20)
Covariate	Subsequent Δ spelling accuracy on untrained items (T2 vs T1)
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Somewhat (T1 behavioral measure should be included in model)
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between Local-Hreg at T1 in the L ventral occipitotemporal region previously identified and subsequent improvement in spelling accuracy of untrained items from T1 to T2. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	Other
Findings notes	There was a significant positive relationship between T1 Local-Hreg and subsequent improvement in spelling accuracy on untrained items from T1 to T2.

Complex analysis 6

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on training items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between change in Local-Hreg in the L ventral occipitotemporal region previously identified and change in spelling accuracy of training items. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	Other
Findings notes	There was a significant negative relationship between change in Local-Hreg and change in spelling accuracy on training items.

Complex analysis 7

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on untrained items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between change in Local-Hreg in the L ventral occipitotemporal region previously identified and change in spelling accuracy of untrained items. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	Other
Findings notes	There was a significant negative relationship between change in Local-Hreg and change in spelling accuracy on untrained items.

Complex analysis 8

First level contrast	Spelling probe (training items) vs rest
Analysis class	Cross-sectional correlation with language or other measure
Group(s)	Aphasia with both timepoints T2 (n = 20)
Covariate	T2 spelling accuracy on training items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between Local-Hreg at T2 in the L ventral occipitotemporal region previously identified and T2 spelling accuracy of training items. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	None
Findings notes	—

Complex analysis 9

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Previous T1 Local-Hreg in L ventral occipitotemporal ROI
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	No (the ROI was defined based on change in Local-Hreg, so spurious findings could arise in the absence of a real effect)
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between change in Local-Hreg in the L ventral occipitotemporal region previously identified and T1 Local-Hreg. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	Other
Findings notes	There was a significant negative relationship between change in Local-Hreg and T1 Local-Hreg.

Complex analysis 10

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on training items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between change in Local-Hreg in the R AG, L PCC, and R PCC and change in spelling accuracy of training items. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	None
Findings notes	—

Complex analysis 11

First level contrast	Spelling probe (training items) vs rest
Analysis class	Longitudinal correlation with language or other measure
Group(s)	Aphasia with both timepoints (n = 20) T2 vs T1
Covariate	Δ spelling accuracy on untrained items
Is the second level contrast valid in terms of the group(s), time point(s), and measures involved?	Yes
Is accuracy matched across the second level contrast?	Unknown, not reported
Is reaction time matched across the second level contrast?	Unknown, not reported
Behavioral data notes	—
Type of analysis	Complex
Statistical details	A linear mixed effects model was used to investigate the relationship between change in Local-Hreg in the R AG, L PCC, and R PCC and change in spelling accuracy of untrained items. A complex model was used in which every voxel for every patient was considered an observation, with random effects of voxel and patient, but this is not described in detail.
Findings	None
Findings notes	—

Notes

Excluded analyses

(1) confirmatory voxelwise analyses in section S4.1 and S4.2; (2) additional analyses accounting for spelling deficit type and auditory comprehension deficits described in 3.3.3; (3) relationship between overall BOLD and local heterogeneity described in 3.4.3, because not related to aphasia recovery