This laboratory sets out to identify research paths that can effectively lead to the development of an artificial system capable of interacting with the ambient under cerebral control. Its distinctive characteristic lies in the fact that it integrates different approaches, making it particularly suitable for a research environment such as that of the Italian Institute of Technology (IIT). A peculiar feature of the proposed approach is the acquisition of in vivo data. This is because, though the project is considered of as a source of fundamental data for basic research, it is and remains an applicative project incorporating major technological challenges. As all the stages of the research project carried ou in the BMI lab are applicable to the man, this predetermines the choices of the materials and procedures required for its realisation. Although the majority of the research is performed within the Department of Robotics, Brain and Cognitive Science, the project benefits also from the contribution and collaboration of various other branches of the IIT, such as Neuroscience and Nanotechnology. Specifically, as regards those project skills that cannot be found within the IIT, use has been made of a network of collaborators including some leading Italian research centres. Among them, the Politecnico of Milan Electronics (POLIMI-ele) and Physics (POLIMI-phys) Departments, the Neurophysiology Sections of the Universities of Ferrara (UNIFE) and Modena (UNIMO), the Neurosurgery Department of the Udine Hospital “S.M. della Misericordia” (NSG), the SISSA in Trieste, and the Northwestern University of Chicago (NW).
The main effort in this area is devoted to the study of chronically implantable Brain Machine Communication devices in humans. In particular the main focus of this long term goal is to implement bidirectional and “ad-hoc” interfaces. By this we mean interfaces that can be adapted to the residual functional abilities and the morphology of individual patients and that can support bidirectional flow of information between the nervous system and the artificial device.
Initial attempts to realise brain-machine interfaces by means of intracortical microelectrodes focused on recording from the primary motor cortex or, better put, from the precentral convexity. In fact, the primary motor cortex of primates lies mainly inside the central sulcus (anterior bank) and is therefore not easy to access. Apart from very few teams, such as that lead by Richard A. Andersen at CALTECH, the main research teams involved in BMI have always looked in the cortex for the functional correlate to movement. On the contrary, we believe that movements are much less represented in the cortex and that, rather, the premotor neurons (as well as the motor ones) activate in a specific way in relation to a determined action. The difference between action and movement consists in the presence of a specific objective (goal). We shall explain this concept more fully by using an example. Consider the action of recovering a small piece of food inside a groove. If the groove is too narrow, the only way to achieve that goal is to introduce the index finger into the groove and bend to try and obtain the food. In this case, the goal is clear (recover the food by bending the index finger) but it is very different when the index finger is bend in order to scratch one's skin, even though the same movement is involved. Recent neurophysiology tells us that the neurons in the premotor cortex that are activated during the first action (recovering food) are different from those activated during the second (scratching) despite the fact that more or less the same muscles are used. Thus, in its neurophysiological sense, the term “action” defines a movement made in order to achieve a goal. The goal, therefore, is the fundamental property of the action. There are actions aiming to reach and manipulate objects, actions aimed towards oneself, communicative actions.
The BMI Lab at IIT is working on two main streams:
The first one is more technology-related and aims at optimize the recording and the analysis of the brain signals through activities in the following areas:
- Carbon nanotubes and conductive polymers composites for high performance, moldable electrodes for in-vivo neural recording
- Microelectronics for signal conditioning and processing of bidirectional brain signals
- Extraction of optimal information from the brain
- Near infrared spectroscopy (NIRS)
The second stream is more neurophysiology-related, and develops along the following lines of research:
- Mapping of motor/premotor functions in individual brains
- Study and implementation of in-vivo techniques for bidirectional communication with the nervous system.
- fMRI studies to study the structural and functional organization of cortical and subcortical motor areas
- Study of the bio-compatibility of implanted devices (in collaboration with NBT)
Carbon nanotubes and conductive polymers composites for high performance, moldable electrodes for in-vivo neural recordings
The aim of the research is to develop a new class of in-vivo neural recording electrodes that do not rely only on metals or silicon based materials, opening the way to the realization of long term neuro-electric interfaces. Particular attention is devoted to significantly improve the signal-to-noise ratio, to explore the possibility of epicortical mid-impedance recordings, to improve the biocompatibility of implants and therefore their temporal stability.
By exploiting the ability to control down to the molecular level the manifold properties exhibited by polymers (such as variable mechanical stiffness, actuation, electrical conductivity and bio-functionalization) together with the outstanding electrical and mechanical properties of carbon nanotubes, we envisage a totally new way of manufacturing nano, micro and macro scale neural probes with very low anatomical impact and high efficiency. In the long run, an all "organic" implantable system (i.e. including signal conditioning) can be foreseen.
In an effort to create fully implantable neural recording/stimulating devices, scientists of this lab, in collaboration with the "Politecnico di Milano", are interested in combining multi-electrode arrays with integrated complementary metal-oxide-semiconductor (CMOS) electronics. The miniaturization of these systems presents significant circuit design challenges. Neurons communicate with each other by means of small amplitude voltages pulses (typically between 10 and 500 µV), known as spikes, which must be amplified before they can be processed further. Since multi-electrode chronic recording systems must be entirely implanted within the skull, incorporating a large number of neural signal amplifiers (one for each electrode), ultra-low noise and ultra-low power design techniques are needed to get clean neural signal recordings as well as to prevent damages of the surrounding tissue which might be caused by excessive heating.
Although a number of studies evidenced that information is mainly carried out by spikes, it is important to facilitate neuroscience research by recording and transmitting these signals with the largest as possible resolution, without any on-chip processing. As a result, multichannel neural recording systems potentially produce an extremely high data rate that must be continuously transmitted. Such approach, although useful for neuroscience research, is not feasible for chronic implantable large multi-electrode systems because of its extremely high power consumption; hence, some form of on-chip processing as well as data compression are subject of study, in order to provide the same amount of information while saving power consumption.
Some Application-Specific-Integration-Circuits (ASICs) are being designed within this lab, in order to be used in the development of brain-machine-interfaces. A first prototype, comprising 8 low-power and low-noise neural signal amplifiers was designed and successfully tested in-vivo with rats. Two distinct solutions comprising respectively 16 and 64 channels, with on-chip A to D conversion and spike detection are currently under test. Wireless communication solutions, as well as stimulation circuits to provide sensory feedbacks are currently subject of study.
The research of the Brain Signal Analysis Laboratory aims at improving and optimizing the way to extract as much information as possible from brain recordings. The signals we analyze can be achieved both by intracortical and epicortical recordings. Different methods of analysis can be used, from the study of frequency spectrum, to that of the correlation between various recording sites, to information analysis of signals, even of different nature, but simultaneously acquired from the brain.
The figure shows the simultaneous recordings from 64 microcontacts performed with an epicortical array (2x1 cm) designed in our laboratory and positioned over the motor cortex. Different colors represent the averaged potentials recorded during movements of the index finger (green), thumb (red) and middle finger (blue) on a temporal scale of 600 ms. The center of abscissas indicates the beginning of the movement. Note how the various signals vary on the basis of their spatial localization.
Near infrared spectroscopy (NIRS) is a new non-invasive technique which uses diffused light to investigate human cerebral functions. The functional mapping of the human brain using the fNIRS technique is a promising though challenging activity. The fNIRS can be used to investigate local changes in concentrations of oxyhaemoglobin (O2Hb) and deoxyhaemoglobin (HHb) following cerebral activation. During the last 6-7 years, tests have been performed on two-wavelength multichannel NIRS-CW (Continuous Wave) systems in order to plot spatial maps of oxygenation changes in the frontal, temporal, parietal and visual cortical areas in response to various stimuli. Although interesting data has been obtained with multichannel CW tools, this approach is not sufficient to carry out accurate studies on the focal changes that take place in the cortex in response to functional stimulation. The Department of Physics of the Politecnico of Milan (POLIMI_PHYS) has worked with time-domain imaging systems (NIRS-TR, Time Resolved). The key advantages of NIRS-TR spectroscopy are: discrimination between absorption and scattering coefficients, increasing penetration depth and spatial resolution with respect to other optical techniques, the possibility of quantifying the contribution of the brain tissue by naturally exploiting time-coded information. The system is based on pulsed semiconductor lasers, a compact multichannel photomultiplier and a PC board for time correlated single photon counting. A collaboration between IIT_RBCS and POLIMI_PHYS is actually focusing on further improvements of the technique.
Recordings of neuronal activity (individual neurons, evoked potentials, local field potentials) are generally performed on cerebral regions identified according to anatomical criteria, the main ones being the configuration of the cerebral sulci (for the cortex) and the stereotaxic position (for the subcortical nuclei). On the other hand, the recent introduction of functional magnetic resonance makes it now possible to localize on single individuals the cerebral regions that become active during particular motor, perceptual or cognitive tasks.
In neurosurgery patients, during the electrophysiological characterization of the brain tissue surrounding a cancer, the association of fMRI (pre-operative) with recording of nerve activity (intraoperative) opens up immense research possibilities as it makes it possible to record neurons exactly where cerebral activation is known to be present. This provides on the one hand precious information to the surgeon about the functional borders of the tumor, on the other hand, the large number of case histories and the ubiquitarian nature of the gliomes (the type of tumors more often requiring such a functional mapping) allow numerous recordings to be made in motor and premotor regions every year.
This sub-project is done in collaboration with the Neurosurgery Department of the Udine Hospital and the Section of Human Physiology of the University of Ferrara.
Study and implementation of in-vivo techniques for bidirectional communication with the nervous system
The main idea of a Brain Machine Interface consists in extracting neural signals directly from the brain and use them to control external devices. In the framwork of building neuralprostheses this technique could be useful to better understand how the brain processes the sensory information coming from the environment and uses it to build motor commands. To reach this result a crucial point is to develop a BMI real-time system to create a bidirectional communication channel with the nervous system. In our lab we are implementing in-vivo techniques using multielectrode microwires arrays chronically implanted in the cortex of awake rodents. These techniques permit to record the neural activity from the cortex while the animal is behaving and simoultaneously to deliver Intracortical Micro Stimulation (ICMS) patterns providing an artificial input in a closed-loop system.
a) Gap crossing with ICMS : the goal of this experiment is to use the ICMS of the barrel cortex to encode information about the environment usually collected by the whiskers. ICMS of the barrel cortex (S1) will be used to substitute the natural sensory input to communicate information about the gap width directly to the brain.
b) Artificial perception via electrical stimulation: a two-choices paradigm is used to explore the electrical parameters of ICMS (i.e. perception threshold, frequency, duration) to increase the information quality that can be provided directly to the brain using this technique.
This subproject sets out to study, using non-invasive fMRI techniques, the anatomical-functional organisation of the cortical systems involved in the coding of action and movement in healthy volunteers. Although some aspects have been extensively investigated using electrophysiological and brain imaging (MEG, PET, fMRI) techniques to demonstrate the involvement of various parietal and frontal areas in the organisation and representation of motor acts, the anatomical-functional connectivity pattern of the parietal-frontal circuits in humans is not yet properly defined. We are therefore aiming to:
- Perform detailed morphometric studies of the structure of the human motor cortex and premotor cortex by quantitatively analysing cortical grey matter from high spatial resolution MR images.
- Perform a high temporal resolution study of the rest connectivity pattern of parietal-frontal cortical areas.
- Perform high-resolution fMRI studies on the spatial-temporal activation pattern and on the connectivity pattern of parietal-frontal cortical areas in relation to the execution, observation and ideation of motor acts.
Studies carried out in collaboration with the Section of Human Physiology of the University of Modena and Reggio Emilia
A problem arising from the use of traditional microelectrodes concerns the stability of the recording over time. In fact, most current off-the-shelf systems allow quality signals to be recorded for a maximum of six months/one year. After this period, the number of neurons recorded by the individual electrode decays rapidly, as does the percentage of electrodes from which it is possible to obtain significant information. The reason for the signal downgrade is to be found in the reaction of the nerve tissue surrounding the electrode. The reactive gliosis to the foreign body produces a kind of cap around the electrode and, consequently, the neurons from which the signal is recorded move away from the recording point.
There are at least three ways of solving this problem. The first involves coating the tip of the electrode with substances that modulate the tissue reaction. The second approach involves minimising the tissue reaction by minimising the size of the electrode. The third sets out to record the so-called local field potentials instead of the activity of individual neurons. These signals are the result of the integration of the activities of various neurons near the electrode. Though they do not possess the descriptive properties of the single action potential, they are much less affected by changes in electrode impedance.