Limits...
Complexity measures in magnetoencephalography: measuring "disorder" in schizophrenia.

Brookes MJ, Hall EL, Robson SE, Price D, Palaniyappan L, Liddle EB, Liddle PF, Robinson SE, Morris PG - PLoS ONE (2015)

Bottom Line: These time-courses are modulated by cognitive tasks, with an increase in local neural processing characterised by localised and transient increases in entropy in the neural signal.We observe a direct but complex relationship between entropy and oscillatory amplitude, which suggests that these metrics are complementary.We demonstrate significantly increased task induced entropy change in patients (compared to controls) in multiple brain regions, including a cingulo-insula network, bilateral insula cortices and a right fronto-parietal network.

View Article: PubMed Central - PubMed

Affiliation: Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, University Park, Nottingham, United Kingdom.

ABSTRACT
This paper details a methodology which, when applied to magnetoencephalography (MEG) data, is capable of measuring the spatio-temporal dynamics of 'disorder' in the human brain. Our method, which is based upon signal entropy, shows that spatially separate brain regions (or networks) generate temporally independent entropy time-courses. These time-courses are modulated by cognitive tasks, with an increase in local neural processing characterised by localised and transient increases in entropy in the neural signal. We explore the relationship between entropy and the more established time-frequency decomposition methods, which elucidate the temporal evolution of neural oscillations. We observe a direct but complex relationship between entropy and oscillatory amplitude, which suggests that these metrics are complementary. Finally, we provide a demonstration of the clinical utility of our method, using it to shed light on aberrant neurophysiological processing in schizophrenia. We demonstrate significantly increased task induced entropy change in patients (compared to controls) in multiple brain regions, including a cingulo-insula network, bilateral insula cortices and a right fronto-parietal network. These findings demonstrate potential clinical utility for our method and support a recent hypothesis that schizophrenia can be characterised by abnormalities in the salience network (a well characterised distributed network comprising bilateral insula and cingulate cortices).

Show MeSH

Related in: MedlinePlus

Multi-scale-entropy, measured in the visual cortex (A), cingulo-insula cortex (B), left motor cortex (C) and right motor cortex (D).Graphs show change in MSE from rest in patients (red) and controls (blue) for the RM and Sternberg tasks. Bar charts show specific cases for scales 1 and 2. Note that there is general agreement between RVE (Fig 4) and MSE (Fig 5) in showing an increased entropy difference in patients relative to controls. Note also that this difference depends on temporal scale, and is maximum when entropy is measured on the very short timescale (i.e. scales of 1 and 2). * indicates p<0.05 corrected.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4401778&req=5

pone.0120991.g008: Multi-scale-entropy, measured in the visual cortex (A), cingulo-insula cortex (B), left motor cortex (C) and right motor cortex (D).Graphs show change in MSE from rest in patients (red) and controls (blue) for the RM and Sternberg tasks. Bar charts show specific cases for scales 1 and 2. Note that there is general agreement between RVE (Fig 4) and MSE (Fig 5) in showing an increased entropy difference in patients relative to controls. Note also that this difference depends on temporal scale, and is maximum when entropy is measured on the very short timescale (i.e. scales of 1 and 2). * indicates p<0.05 corrected.

Mentions: The difference in entropy between patients and controls, observed in Fig 7, is based upon RVE, which allows direct inference on the temporal evolution of entropy in the brain. In this RVE implementation, we assess temporal variation on a timescale of 3.4ms (taking every second time sample at a sample rate of 600Hz); entropy is calculated using a leaky integrator with a (1/e) window width of 0.3s. In other words, whilst the timecourses in Figs 5 and 7 look temporally smooth, RVE assesses temporal variability at a millisecond timescale. This said, entropy can be measured at any temporal scale and it proves instructive to assess multiple scales in order to test whether the observed differences persist when using a slower timeframe. To achieve this, we used MSE [26,27], which has been employed in previous schizophrenia measurements (e.g. [9]) and in addition to offering a means to assess the effect of temporal scale, also allows a means to compare our RVE methodology to others previously employed. Results are given in Fig 8, which shows the task induced change in MSE (ΔMSE) plotted against temporal scale in the visual cortex (A), cingulo-insula cortex (B), left motor cortex (C) and right motor cortex (D). Results are in agreement with our RVE metrics (Fig 7) in showing significantly increased task related entropy change in patients compared to controls. (p<0.05, corrected for multiple comparisons across the 12 regions). More importantly, Fig 8 also shows clear evidence that this observation is critically dependent on the timescale at which entropy is assessed, with significant results observed only at the most rapid timescales (scales of 1 and 2, corresponding to sampling at 1.7ms and 3.4ms scale respectively). These timescales are approximately equivalent to those used for RVE. Finally, note that whilst the primary results showing increased entropy in patients relative to controls is confirmed by MSE, no information on the temporal evolution of entropy is generated.


Complexity measures in magnetoencephalography: measuring "disorder" in schizophrenia.

Brookes MJ, Hall EL, Robson SE, Price D, Palaniyappan L, Liddle EB, Liddle PF, Robinson SE, Morris PG - PLoS ONE (2015)

Multi-scale-entropy, measured in the visual cortex (A), cingulo-insula cortex (B), left motor cortex (C) and right motor cortex (D).Graphs show change in MSE from rest in patients (red) and controls (blue) for the RM and Sternberg tasks. Bar charts show specific cases for scales 1 and 2. Note that there is general agreement between RVE (Fig 4) and MSE (Fig 5) in showing an increased entropy difference in patients relative to controls. Note also that this difference depends on temporal scale, and is maximum when entropy is measured on the very short timescale (i.e. scales of 1 and 2). * indicates p<0.05 corrected.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4401778&req=5

pone.0120991.g008: Multi-scale-entropy, measured in the visual cortex (A), cingulo-insula cortex (B), left motor cortex (C) and right motor cortex (D).Graphs show change in MSE from rest in patients (red) and controls (blue) for the RM and Sternberg tasks. Bar charts show specific cases for scales 1 and 2. Note that there is general agreement between RVE (Fig 4) and MSE (Fig 5) in showing an increased entropy difference in patients relative to controls. Note also that this difference depends on temporal scale, and is maximum when entropy is measured on the very short timescale (i.e. scales of 1 and 2). * indicates p<0.05 corrected.
Mentions: The difference in entropy between patients and controls, observed in Fig 7, is based upon RVE, which allows direct inference on the temporal evolution of entropy in the brain. In this RVE implementation, we assess temporal variation on a timescale of 3.4ms (taking every second time sample at a sample rate of 600Hz); entropy is calculated using a leaky integrator with a (1/e) window width of 0.3s. In other words, whilst the timecourses in Figs 5 and 7 look temporally smooth, RVE assesses temporal variability at a millisecond timescale. This said, entropy can be measured at any temporal scale and it proves instructive to assess multiple scales in order to test whether the observed differences persist when using a slower timeframe. To achieve this, we used MSE [26,27], which has been employed in previous schizophrenia measurements (e.g. [9]) and in addition to offering a means to assess the effect of temporal scale, also allows a means to compare our RVE methodology to others previously employed. Results are given in Fig 8, which shows the task induced change in MSE (ΔMSE) plotted against temporal scale in the visual cortex (A), cingulo-insula cortex (B), left motor cortex (C) and right motor cortex (D). Results are in agreement with our RVE metrics (Fig 7) in showing significantly increased task related entropy change in patients compared to controls. (p<0.05, corrected for multiple comparisons across the 12 regions). More importantly, Fig 8 also shows clear evidence that this observation is critically dependent on the timescale at which entropy is assessed, with significant results observed only at the most rapid timescales (scales of 1 and 2, corresponding to sampling at 1.7ms and 3.4ms scale respectively). These timescales are approximately equivalent to those used for RVE. Finally, note that whilst the primary results showing increased entropy in patients relative to controls is confirmed by MSE, no information on the temporal evolution of entropy is generated.

Bottom Line: These time-courses are modulated by cognitive tasks, with an increase in local neural processing characterised by localised and transient increases in entropy in the neural signal.We observe a direct but complex relationship between entropy and oscillatory amplitude, which suggests that these metrics are complementary.We demonstrate significantly increased task induced entropy change in patients (compared to controls) in multiple brain regions, including a cingulo-insula network, bilateral insula cortices and a right fronto-parietal network.

View Article: PubMed Central - PubMed

Affiliation: Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, University Park, Nottingham, United Kingdom.

ABSTRACT
This paper details a methodology which, when applied to magnetoencephalography (MEG) data, is capable of measuring the spatio-temporal dynamics of 'disorder' in the human brain. Our method, which is based upon signal entropy, shows that spatially separate brain regions (or networks) generate temporally independent entropy time-courses. These time-courses are modulated by cognitive tasks, with an increase in local neural processing characterised by localised and transient increases in entropy in the neural signal. We explore the relationship between entropy and the more established time-frequency decomposition methods, which elucidate the temporal evolution of neural oscillations. We observe a direct but complex relationship between entropy and oscillatory amplitude, which suggests that these metrics are complementary. Finally, we provide a demonstration of the clinical utility of our method, using it to shed light on aberrant neurophysiological processing in schizophrenia. We demonstrate significantly increased task induced entropy change in patients (compared to controls) in multiple brain regions, including a cingulo-insula network, bilateral insula cortices and a right fronto-parietal network. These findings demonstrate potential clinical utility for our method and support a recent hypothesis that schizophrenia can be characterised by abnormalities in the salience network (a well characterised distributed network comprising bilateral insula and cingulate cortices).

Show MeSH
Related in: MedlinePlus