Ever wondered what was going on in the brain of John Coltrane when he played the famous solo on his album Giant Steps?
Researchers at the National Institute of Information and Communications Technology (NICT), Japan, and Western University, Canada, have succeeded in visualizing how information is represented in a widespread area in the human cerebral cortex during a performance of skilled finger movement sequences.
Contrary to the common assumption, the researchers found that overlapping regions in the premotor and parietal cortices represent the sequences in multiple levels of motor hierarchy (e.g., chunks of a few finger movements, or chunks of a few chunks), whereas the individual finger movements (i.e., the lowest level in the hierarchy) were uniquely represented in the primary motor cortex.
These results uncovered the first detailed map of cortical sequence representation in the human brain.
The results may also provide some clue for locating new candidate brain areas as signal sources for motor BCI application or developing more sophisticated algorithm to reconstruct complex motor behavior.
The results were published online as Yokoi and Diedrichsen “Neural Organization of Hierarchical Motor Sequence Representations in the Human Neocortex” in Neuron on July 22, 2019.
Achievements
The best way to remember/produce long and complex motor sequences is to divide them into several smaller pieces recursively.
For example, a musical piece may be remembered as a sequence of smaller chunks, with each chunk representing a group of often co-occurring notes.
Such hierarchical organization has long been thought to underlie our control of motor sequences from the highly skillful actions, like playing music, to daily behavior, like making a cup of tea.
Yet, very little is known about how these hierarchies are implemented in our brain.
In a new study published in a journal Neuron, Atsushi Yokoi, Center for Information and Neural Networks (CiNet), NICT, and Jörn Diedrichsen, Brain and Mind Institute, Western Univ., provide the first direct evidence of how hierarchically organized sequences are represented through the population activity across the human cerebral cortex.
The researchers measured the fine-grained fMRI activity patterns, while human participants produced 8 different remembered sequences of 11 finger presses.
“Remembering 8 different sequences of 11 finger presses is a tough task, so you will definitely need to organize them hierarchically,” says Diedrichsen, the study’s senior author and a Western Research Chair for Motor Control and Computational Neuroscience at the Western University, Canada.
“To study a hierarchy, you would really need the sequences to have this much of complexity.
And currently it’s very hard to train animals to learn such sequences,” added Yokoi, the study’s lead author who is a former postdoctoral researcher at the Diedrichsen’s group since both were at the Institute of Cognitive Neuroscience, University College London in UK, and now a Researcher at the CiNet, NICT, Japan.
Through a series of careful behavioural analyses, the researchers could show that participants encoded the sequences in terms of a three-level hierarchy;
(1) individual finger presses,
(2) chunks consisting of two or three finger presses, and
(3) entire sequences consisting of four chunks.
They could then characterize the fMRI activity patterns with respect to these hierarchies using machine learning techniques.
As expected, the patterns in primary motor cortex, the area that controls finger movements, seemed to only depend on each individual finger moved, independent of it positioning in the sequence.
Activity in higher-order motor areas, such as premotor and parietal cortices, clearly could be shown to encode the sequential context at the level of chunks or entire sequence.
Thus, in contrast to primary motor cortex, these areas “know” what was played before and what comes after the ongoing finger press.
For the first time the study now allowed insights into the organization of these higher-order representations.
Surprisingly, different levels of sequence information overlapped greatly.
An unsupervised clustering approach further subdivided these areas into distinct clusters, each had a different mixing ratio of the representations, just like how one’s iPhone storage is used.
These results uncovered the first detailed map of cortical sequence representation in the human brain.
Study’s impact
One common assumption in the cognitive neuroscience has been that each level in the functional hierarchy would mirror the anatomical hierarchy, from the higher, association cortices (e.g., premotor or parietal cortices) down to the primary sensorimotor cortices. The mysterious coexistence of a clear anatomical separation (i.e., individual finger vs. other representations) and an overlap (i.e., chunk and sequence representations) sheds new light on the classical question of the correspondence between functional and anatomical hierarchies.
“It can be said the brain represents motor sequences in partly hierarchical, yet partly flat ways.”
Hierarchical organization of motor sequence. The image is credited to NICT.
“Although its functional role is still unclear, the anatomical overlap between chunk and sequence representations may suggest these representations in upper movement hierarchy may influence with each other to support flexible sequence production. This needs to be tested in the future study,” Yokoi concluded.
Future prospects
The study also suggests possible loci from which we can record brain signal to control neural prosthetics to make fluent movement sequences in potential BCI applications.
The researchers also hope that it could also contribute in developing a new decoding algorithm that effectively combines the information in different hierarchies to reproduce movements.
Funding: The study was conducted under an international collaboration between NICT (Japan), UCL (UK), and Western University (Canada).
Although we traditionally understand voluntary motor movements to stem from the cortex within the hemisphere contralateral to a moving limb, there is increasing evidence that the ipsilateral hemisphere also plays an active role in the execution of voluntary motor movements. Across a variety of modalities in both human subjects and animal models, ipsilateral cortical activations have been observed during unilateral limb movements (Tanji et al., 1988; Aizawa et al., 1990; Wisneski et al., 2008; Ganguly et al., 2009; Buetefisch et al., 2014; Hotson et al., 2014). Similarly, ipsilesional motor deficits have been observed in human patients following unilateral cortical injuries (Baskett et al., 1996; Sunderland, 2000; Schaefer et al., 2007, 2009a,b, 2012). Although this evidence supports the idea that the ipsilateral hemisphere may be involved in the execution of voluntary motor movements, the exact role of the ipsilateral hemisphere remains uncertain.
Defining how the brain encodes motor kinematics (i.e., speed, velocity, and position) is essential to understanding the cortical dynamics that underpin motor control in humans. With regards to ipsilateral motor activations, previous studies have demonstrated that some limited movement kinematics can be decoded from the ipsilateral hemisphere (Ganguly et al., 2009; Hotson et al., 2012, 2014). The extent and detail of the information that is encoded ipsilateral to a moving limb, however, is currently unknown. Additionally, how this ipsilateral kinematic information compares to contralateral kinematic encoding also remains largely unresolved because previous studies have reached conflicting conclusions. Some investigators found that ipsilateral motor activations occur at distinct frequencies (35–50 Hz) and locations (premotor) relative to contralateral motor activations (Wisneski et al., 2008), whereas others have found that contralateral and ipsilateral movement-related activations are similar (Fujiwara et al., 2017; Haar et al., 2017). These discrepant results underscore the need to define, to the highest level possible, how a given hemisphere represents multidimensional ipsilateral and contralateral kinematics. Specifically, encoding detailed movement parameters such as kinematics, joint angles, or muscle activations is a necessary condition for a given hemisphere to play a role in planning and executing voluntary motor movements. Further, understanding the link between cortical physiology and movement is essential to crafting more informed rehabilitation strategies in the setting of brain injuries and movement disorders.
In this study, motor-intact humans implanted with unilateral intracranial electrocorticographic (ECoG) electrodes performed a 3D center-out reaching task with each arm (Fig. 1B–D). We initially hypothesized that unilateral ECoG signals would enable us to decode movement kinematics of both limbs with distinct features distinguishing each arm. We found that ECoG signals could be used to decode 3D kinematics of ipsilateral limb movements. Additionally, we found that ECoG representations of reaching movements are conserved between contralateral and ipsilateral limb movements. Together, these findings support that the ipsilateral hemisphere plays an active role in unilateral arm movement and that kinematic information has substantial bihemispheric representation.
Figure 1.
Study methodology. Patients implanted with electrocorticography arrays completed a 3D center-out reaching task. A, Electrode locations were based upon the clinical requirements of each patient and were localized to an atlas brain for display. B, Patients were seated in the semirecumbent position and completed reaching movements from the center to the corners of a 50 cm physical cube based upon cues from LED lights located at each target while hand positions and ECoG signals were simultaneously recorded. Each patient was implanted with electrodes in a single cortical hemisphere and performed the task with the arm contralateral (C) and ipsilateral (D) to the electrode array in separate recording sessions. E, The task incorporated a center hold period (Hold-A), planning delay, movement period, and exterior hold period (Hold-B). To decode kinematics of contralateral and ipsilateral reaching movements, a hierarchical PLS regression that incorporated a logistic regression classification of movement and rest periods to switch the predicted output between the output of two PLS regression models was used. The first PLS model was trained using data from the rest periods to predict speed and velocity during rest periods and the second PLS regression model was trained using data from the movement periods to predict speed and velocity during movement periods (F). (E adapted from Bundy et al., 2016 under terms of the CC BY license).
Source:
NICT
Media Contacts:
Sachiko Hirota – NICT
Image Source:
The image is credited to NICT.
Original Research: Closed access
“Neural Organization of Hierarchical Motor Sequence Representations in the Human Neocortex”.Atsushi Yokoi, Jörn Diedrichsen.
Neuron. doi:10.1016/j.neuron.2019.06.017
