Limits...
Brain classification reveals the right cerebellum as the best biomarker of dyslexia.

Pernet CR, Poline JB, Demonet JF, Rousselet GA - BMC Neurosci (2009)

Bottom Line: The clinical relevance of this result was assessed by inquiring cognitive brain-based differences among dyslexic brain subgroups in comparison to normal readers' performances.Furthermore, cerebellar and lentiform grey matter volumes interacted in dyslexic subjects, so that lower and higher lentiform grey matter volumes compared to controls differently modulated the phonological and lexical performances.In addition, behavioural analyses suggest that these brain phenotypes relate to different deficits of automatization of language-based processes such as grapheme/phoneme correspondence and/or rapid access to lexicon entries.

View Article: PubMed Central - HTML - PubMed

Affiliation: SFC Brain Imaging Research Centre, SINAPSE Collaboration, University of Edinburgh, Edinburgh, UK. cyril.pernet@ed.ac.uk

ABSTRACT

Background: Developmental dyslexia is a specific cognitive disorder in reading acquisition that has genetic and neurological origins. Despite histological evidence for brain differences in dyslexia, we recently demonstrated that in large cohort of subjects, no differences between control and dyslexic readers can be found at the macroscopic level (MRI voxel), because of large variances in brain local volumes. In the present study, we aimed at finding brain areas that most discriminate dyslexic from control normal readers despite the large variance across subjects. After segmenting brain grey matter, normalizing brain size and shape and modulating the voxels' content, normal readers' brains were used to build a 'typical' brain via bootstrapped confidence intervals. Each dyslexic reader's brain was then classified independently at each voxel as being within or outside the normal range. We used this simple strategy to build a brain map showing regional percentages of differences between groups. The significance of this map was then assessed using a randomization technique.

Results: The right cerebellar declive and the right lentiform nucleus were the two areas that significantly differed the most between groups with 100% of the dyslexic subjects (N = 38) falling outside of the control group (N = 39) 95% confidence interval boundaries. The clinical relevance of this result was assessed by inquiring cognitive brain-based differences among dyslexic brain subgroups in comparison to normal readers' performances. The strongest difference between dyslexic subgroups was observed between subjects with lower cerebellar declive (LCD) grey matter volumes than controls and subjects with higher cerebellar declive (HCD) grey matter volumes than controls. Dyslexic subjects with LCD volumes performed worse than subjects with HCD volumes in phonologically and lexicon related tasks. Furthermore, cerebellar and lentiform grey matter volumes interacted in dyslexic subjects, so that lower and higher lentiform grey matter volumes compared to controls differently modulated the phonological and lexical performances. Best performances (observed in controls) corresponded to an optimal value of grey matter and they dropped for higher or lower volumes.

Conclusion: These results provide evidence for the existence of various subtypes of dyslexia characterized by different brain phenotypes. In addition, behavioural analyses suggest that these brain phenotypes relate to different deficits of automatization of language-based processes such as grapheme/phoneme correspondence and/or rapid access to lexicon entries.

Show MeSH

Related in: MedlinePlus

Illustration of the advantage and robustness of the bootstrap procedure over the one-sample t-test confidence interval. From A to D, histograms of the grey matter values across subjects are plotted for the maximum of the average image, the standard deviation image, the kurtosis image and the skewness image. As illustrated (data distribution in blue), data did not conform well (bias) to the normal distribution. This resulted in an over-estimation of the CI size using a one sample t-test (red dotted lines on the CI evolution graphics). By contrast, bootstrapped CI sizes were narrower (blue lines on the CI evolution graphics). This is illustrated over the whole brain on brain renders E and F. Overall, bootstrapped and t-test CI were similarly distributed (E) but bootstrapped CI were in general narrower (warm colours in F). The average one-sample t-test CI size was 0.0486 ± 0.0096 (min 0.0198, max 0.096, median 0.0474) vs. 0.0461 ± 0.0091 (min 0.0188, max 0.0895, median 0.045) for the 5000 resamples bootstrap CIs. This difference was statistically significant (t(251572) = 65.25 p < .00001), illustrating the advantage of the bootstrap approach, even if the majority of brain voxels have a close to Normal distributions (G – Lilliefors test > .05 – arrows and circles indicates the few 'non-normal' voxels). Graphics titled 'CI size' illustrate the evolution of the bootstrapped CI size with the number of resamples. The vertical dotted line mark the 3000th resample, from which CI size tend to be stable. On the right hand side, graphics titled 'Means' show the distribution of the data means after 5000 resample (to compare with the original 'data distribution') from which the last bootstrapped CIs were obtained.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2713247&req=5

Figure 5: Illustration of the advantage and robustness of the bootstrap procedure over the one-sample t-test confidence interval. From A to D, histograms of the grey matter values across subjects are plotted for the maximum of the average image, the standard deviation image, the kurtosis image and the skewness image. As illustrated (data distribution in blue), data did not conform well (bias) to the normal distribution. This resulted in an over-estimation of the CI size using a one sample t-test (red dotted lines on the CI evolution graphics). By contrast, bootstrapped CI sizes were narrower (blue lines on the CI evolution graphics). This is illustrated over the whole brain on brain renders E and F. Overall, bootstrapped and t-test CI were similarly distributed (E) but bootstrapped CI were in general narrower (warm colours in F). The average one-sample t-test CI size was 0.0486 ± 0.0096 (min 0.0198, max 0.096, median 0.0474) vs. 0.0461 ± 0.0091 (min 0.0188, max 0.0895, median 0.045) for the 5000 resamples bootstrap CIs. This difference was statistically significant (t(251572) = 65.25 p < .00001), illustrating the advantage of the bootstrap approach, even if the majority of brain voxels have a close to Normal distributions (G – Lilliefors test > .05 – arrows and circles indicates the few 'non-normal' voxels). Graphics titled 'CI size' illustrate the evolution of the bootstrapped CI size with the number of resamples. The vertical dotted line mark the 3000th resample, from which CI size tend to be stable. On the right hand side, graphics titled 'Means' show the distribution of the data means after 5000 resample (to compare with the original 'data distribution') from which the last bootstrapped CIs were obtained.

Mentions: At variance with a two-sample t-test where most patients' data have to be located below or above the CIs (sample homogeneity) to observe differences, our method made it possible to identify areas in which patients differ from controls even if patients' data were distributed below and above the CIs (sample inhomogeneity – see Fig. 2A and 2B). Concretely, using a two-sample t-test, we could not observe a significant difference between controls and patients in the cerebellum (t(75) = -.48 p = .6), or in the lentiform nucleus (t(75) = -0.9 p = .37). Similarly, because the classification was performed on bootstrapped CIs that have a better control on type I error and narrower intervals than one sample t-test CIs (Fig. 5), the classification performed using the 95% one sample t-test CI failed to show any areas with 100% of differences. Using the classical approach, the classification performed on cerebellar and the lentiform nucleus voxels only reached a maximum of 94% difference.


Brain classification reveals the right cerebellum as the best biomarker of dyslexia.

Pernet CR, Poline JB, Demonet JF, Rousselet GA - BMC Neurosci (2009)

Illustration of the advantage and robustness of the bootstrap procedure over the one-sample t-test confidence interval. From A to D, histograms of the grey matter values across subjects are plotted for the maximum of the average image, the standard deviation image, the kurtosis image and the skewness image. As illustrated (data distribution in blue), data did not conform well (bias) to the normal distribution. This resulted in an over-estimation of the CI size using a one sample t-test (red dotted lines on the CI evolution graphics). By contrast, bootstrapped CI sizes were narrower (blue lines on the CI evolution graphics). This is illustrated over the whole brain on brain renders E and F. Overall, bootstrapped and t-test CI were similarly distributed (E) but bootstrapped CI were in general narrower (warm colours in F). The average one-sample t-test CI size was 0.0486 ± 0.0096 (min 0.0198, max 0.096, median 0.0474) vs. 0.0461 ± 0.0091 (min 0.0188, max 0.0895, median 0.045) for the 5000 resamples bootstrap CIs. This difference was statistically significant (t(251572) = 65.25 p < .00001), illustrating the advantage of the bootstrap approach, even if the majority of brain voxels have a close to Normal distributions (G – Lilliefors test > .05 – arrows and circles indicates the few 'non-normal' voxels). Graphics titled 'CI size' illustrate the evolution of the bootstrapped CI size with the number of resamples. The vertical dotted line mark the 3000th resample, from which CI size tend to be stable. On the right hand side, graphics titled 'Means' show the distribution of the data means after 5000 resample (to compare with the original 'data distribution') from which the last bootstrapped CIs were obtained.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2713247&req=5

Figure 5: Illustration of the advantage and robustness of the bootstrap procedure over the one-sample t-test confidence interval. From A to D, histograms of the grey matter values across subjects are plotted for the maximum of the average image, the standard deviation image, the kurtosis image and the skewness image. As illustrated (data distribution in blue), data did not conform well (bias) to the normal distribution. This resulted in an over-estimation of the CI size using a one sample t-test (red dotted lines on the CI evolution graphics). By contrast, bootstrapped CI sizes were narrower (blue lines on the CI evolution graphics). This is illustrated over the whole brain on brain renders E and F. Overall, bootstrapped and t-test CI were similarly distributed (E) but bootstrapped CI were in general narrower (warm colours in F). The average one-sample t-test CI size was 0.0486 ± 0.0096 (min 0.0198, max 0.096, median 0.0474) vs. 0.0461 ± 0.0091 (min 0.0188, max 0.0895, median 0.045) for the 5000 resamples bootstrap CIs. This difference was statistically significant (t(251572) = 65.25 p < .00001), illustrating the advantage of the bootstrap approach, even if the majority of brain voxels have a close to Normal distributions (G – Lilliefors test > .05 – arrows and circles indicates the few 'non-normal' voxels). Graphics titled 'CI size' illustrate the evolution of the bootstrapped CI size with the number of resamples. The vertical dotted line mark the 3000th resample, from which CI size tend to be stable. On the right hand side, graphics titled 'Means' show the distribution of the data means after 5000 resample (to compare with the original 'data distribution') from which the last bootstrapped CIs were obtained.
Mentions: At variance with a two-sample t-test where most patients' data have to be located below or above the CIs (sample homogeneity) to observe differences, our method made it possible to identify areas in which patients differ from controls even if patients' data were distributed below and above the CIs (sample inhomogeneity – see Fig. 2A and 2B). Concretely, using a two-sample t-test, we could not observe a significant difference between controls and patients in the cerebellum (t(75) = -.48 p = .6), or in the lentiform nucleus (t(75) = -0.9 p = .37). Similarly, because the classification was performed on bootstrapped CIs that have a better control on type I error and narrower intervals than one sample t-test CIs (Fig. 5), the classification performed using the 95% one sample t-test CI failed to show any areas with 100% of differences. Using the classical approach, the classification performed on cerebellar and the lentiform nucleus voxels only reached a maximum of 94% difference.

Bottom Line: The clinical relevance of this result was assessed by inquiring cognitive brain-based differences among dyslexic brain subgroups in comparison to normal readers' performances.Furthermore, cerebellar and lentiform grey matter volumes interacted in dyslexic subjects, so that lower and higher lentiform grey matter volumes compared to controls differently modulated the phonological and lexical performances.In addition, behavioural analyses suggest that these brain phenotypes relate to different deficits of automatization of language-based processes such as grapheme/phoneme correspondence and/or rapid access to lexicon entries.

View Article: PubMed Central - HTML - PubMed

Affiliation: SFC Brain Imaging Research Centre, SINAPSE Collaboration, University of Edinburgh, Edinburgh, UK. cyril.pernet@ed.ac.uk

ABSTRACT

Background: Developmental dyslexia is a specific cognitive disorder in reading acquisition that has genetic and neurological origins. Despite histological evidence for brain differences in dyslexia, we recently demonstrated that in large cohort of subjects, no differences between control and dyslexic readers can be found at the macroscopic level (MRI voxel), because of large variances in brain local volumes. In the present study, we aimed at finding brain areas that most discriminate dyslexic from control normal readers despite the large variance across subjects. After segmenting brain grey matter, normalizing brain size and shape and modulating the voxels' content, normal readers' brains were used to build a 'typical' brain via bootstrapped confidence intervals. Each dyslexic reader's brain was then classified independently at each voxel as being within or outside the normal range. We used this simple strategy to build a brain map showing regional percentages of differences between groups. The significance of this map was then assessed using a randomization technique.

Results: The right cerebellar declive and the right lentiform nucleus were the two areas that significantly differed the most between groups with 100% of the dyslexic subjects (N = 38) falling outside of the control group (N = 39) 95% confidence interval boundaries. The clinical relevance of this result was assessed by inquiring cognitive brain-based differences among dyslexic brain subgroups in comparison to normal readers' performances. The strongest difference between dyslexic subgroups was observed between subjects with lower cerebellar declive (LCD) grey matter volumes than controls and subjects with higher cerebellar declive (HCD) grey matter volumes than controls. Dyslexic subjects with LCD volumes performed worse than subjects with HCD volumes in phonologically and lexicon related tasks. Furthermore, cerebellar and lentiform grey matter volumes interacted in dyslexic subjects, so that lower and higher lentiform grey matter volumes compared to controls differently modulated the phonological and lexical performances. Best performances (observed in controls) corresponded to an optimal value of grey matter and they dropped for higher or lower volumes.

Conclusion: These results provide evidence for the existence of various subtypes of dyslexia characterized by different brain phenotypes. In addition, behavioural analyses suggest that these brain phenotypes relate to different deficits of automatization of language-based processes such as grapheme/phoneme correspondence and/or rapid access to lexicon entries.

Show MeSH
Related in: MedlinePlus