Fichier PDF

Partagez, hébergez et archivez facilement vos documents au format PDF

Partager un fichier Mes fichiers Boite à outils PDF Recherche Aide Contact



201700001841 Costecalde T .pdf



Nom original: 201700001841 Costecalde T.pdf
Titre: Microsoft Word - _system_appendPDF_cover-forpdf.htm
Auteur:

Ce document au format PDF 1.5 a été généré par Appligent AppendPDF Pro 6.0 / activePDF DocConverter, et a été envoyé sur fichier-pdf.fr le 19/07/2018 à 10:14, depuis l'adresse IP 132.168.x.x. La présente page de téléchargement du fichier a été vue 234 fois.
Taille du document: 2 Mo (41 pages).
Confidentialité: fichier public




Télécharger le fichier (PDF)









Aperçu du document


Neuromodulation Proof

A long term BCI study with ECoG recordings in freely
moving rats

Journal:

NER-1935-08-2016.R3

Fo

Manuscript ID

Neuromodulation: Technology at the Neural Interface

Manuscript Type:

Complete List of Authors:

07-Apr-2017

rP

Date Submitted by the Author:

Basic Research

Keywords:

rR

ee

costecalde, thomas; CEA, Clinatec
Aksenova, tetiana; CEA, Clinatec
Eliseyev, Andriy; CEA, Clinatec
Torres-Martinez, napoléon; CEA, Clinatec
Mestais, corinne; CEA, Clinatec
Moro, cécile; CEA, Clinatec
Benabid, Alim-Louis; CEA, Clinatec; Universite Joseph Fourier,
Neurosurgery

iew

ev

Neuroprosthetic, Freely moving animals, Brain Computer Interface,
Asynchronous, ElectroCorticoGraphic recordings

Neuromodulation Proof

Page 1 of 40

Abstract: Background: Brain Computer Interface (BCI) studies are performed in an increasing
number of applications. Questions are raised about electrodes, data processing and effectors.
Experiments are needed to solve these issues.
Objective: To develop a simple BCI set-up to easier studies for improving the mathematical tools to
process the ECoG to control an effector.
Method: We designed a simple BCI using transcranial electrodes (17 screws, three mechanically
linked to create a common reference, 14 used as recording electrodes) to record Electro-CorticoGraphic (ECoG) neuronal activities in rodents. The data processing is based on an online self-paced
non supervised (asynchronous) BCI paradigm. N-way partial least squares algorithm together with

Fo

Continuous Wavelet Transformation of ECoG recordings detect signatures related to motor activities.
Signature detection in freely moving rats may activate external effectors during a behavioral task,

rP

which involved pushing a lever to obtain a reward.

ee

Results: After routine training, we showed that peak brain activity preceding a lever push (LP) to
obtain food reward was located mostly in the cerebellar cortex with a higher correlation coefficient,

rR

suggesting a strong postural component and also in the occipital cerebral cortex. Analysis of brain
activities provided a stable signature in the high gamma band (~180Hz) occurring within 1500 ms

ev

before the lever push approximatively around -400msec to -500 msec. Detection of the signature from

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

a single cerebellar cortical electrode triggers the effector with high efficiency (68% Offline and 30%
Online) and rare false positives per minute in sessions about 30 minutes and up to one hour (~2 online
and offline).

Conclusions: In summary, our results are original as compared to the rest of the literature, which
involves rarely rodents, a simple BCI set-up has been developed in rats, the data show for the first time
long term, up to one year, unsupervised online control of an effector.
INTRODUCTION
Disability due to traumatic spinal cord injury represents a significant socioeconomic problem,
affording little hope to patients. Brain Computer Interface (BCI) systems may activate neuroprosthetic
devices to restore neurological functions in post-traumatic spinal cord injury handicapped individuals
[1]. The brain functions of patients with spinal cord lesions continue to function normally, including

Neuromodulation Proof

Neuromodulation Proof

generation of mental images of movements, which are correlated to electrical signals that can be
recorded through surface or implanted electrodes of BCIs [2].
In previous BCI studies (Table 1), neuronal activity has been recorded using scalp [3 and 4], epidural
[5 and 6] or subdural electrodes [7-9], together with microelectrode arrays [10-16]. Scalp
electroencephalographic (EEG) recording is safe but necessitates wearing a non-ergonomic EEG
helmet, requiring daily repositioning and BCI system recalibration [17]. On the other hand,
microelectrode arrays are very sensitive tools for recording of spikes and local field potentials, with
high signal to noise ratios, but they are highly invasive [18-20]. Furthermore, permanently implanted
ECoG-based electrode grid array do not require recalibration and provide a long-term, low

Fo

maintenance BCI solution. High quality and multi-modal signal processing facilitate information
extraction [21 and 22], yielding a BCI suitable for clinical applications with optimal quality and

rP

lifestyle issues. Decoding of ECoG signals related to a task can be either synchronous (triggered by a

ee

cue delivered before the task) [5, 8, 10 and 15] or asynchronous (unsupervised, the data processing
occurring continuously along the experiment, regardless of the execution of the task) [23-25]. The

rR

asynchronous, unsupervised, algorithms are better corresponding to the detection of events occurring
randomly (cue less), as this is the case in real life.

iew

TABLE 1

ev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

This study presents a novel, easy-to-use method for ECoG recordings in freely moving rats for the
long term (up to one year), allowing evaluation of various paradigms for data processing to control an
external food dispenser. An algorithm for asynchronous ECoG based self-learning BCI was designed
that is capable of detecting a motor-event related neural signal, called a signature, registered by a
single, or more, chronically implanted epicortical electrode during performance of a binary behavioral
task.

MATERIALS AND METHODS

Neuromodulation Proof

Page 2 of 40

Page 3 of 40

EXPERIMENTAL SET UP

1. Animal preparation
Twelve OFA (Oncins France strain A Charles River laboratories, Lyon France) female rats (150-250g)
under general anesthesia (Chloral hydrate 4%, 1ml/100g i.p.) were installed on a stereotactic frame
(David Kopf® Instruments, Tujunga, CA USA). Seventeen titanium self-tapering screw electrodes
(3.6mm long, 1.1 mm in diameter, Fig. 1A, 1B and 1C), including 3 interconnected references, were
implanted in mini burr holes drilled in definite placements corresponding to skull landmarks (lambda
and bregma, bone sutures) (Figure 1D and 1E). The electrodes were inserted through miniplates

Fo

connected by a wire to a PL700 Amphenol connector (D micro-pin 15Pts, Radiospare, France). Dental
cement (Methyl metacrylate polymer METHAX, Generique International Laboratories, France) was

rP

used to isolate and embed the plates, the screws and the connector. The skin was sutured over the
cement bloc around the connector.

ee

Among the 12 animals, 10 animals survived the full year, 2 died (one during the immediate post-

rR

surgical period, the other after breaking its connector from the calvarium).
Autopsy at the end of the experiments (general anesthesia, intracardiac perfusion with saline and then

ev

formaldehyde, brain sampling) provided anatomical confirmation of the electrode placement.

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

FIGURE 1

The rats were allowed to recover from surgery for a week. They had free access to food (SDS, Essex
UK) and water; they were housed in GLP cages stored in an air conditioned closet (9ARMV2124LR,
Techniplast, France) with temperature, light cycle and humidity control (12/12 hours day night cycles;
25°C).
Ethical approval for all experimental procedures was obtained from the Animal Ethics Committee
COMETH

(Grenoble)

and

the

French

Ministry

for

Research

(protocol

number

“4bis_Clinatec_ALB_01”) and were performed in accordance with the European Communities
Council Directive of 1986 (86/609/EEC) for care of laboratory animals.

Neuromodulation Proof

Neuromodulation Proof

2. Set Up and Data Acquisition
Experiments were performed on freely moving animals placed in a behavioral cage (Abett II starter
kit, Campden instruments, Lafayette Instruments Co., Leicestershire, UK). Sessions was about 30
minutes and up to one hour were repeated at least three times a week.
The electrodes were connected to a multichannel connector and swivel (T13EEG slip ring assembly,
Air Précision, Le Plessis Robinson, France) at the ceiling of the cage, allowing the rat to move freely.
Data (ECoGs) were continuously recorded on a Micromed® system (Micromed SD32, Micromed
Italy) compatible with Matlab®-based treatment. The signal was recorded with a sampling rate of

Fo

1024Hz on [0.1Hz; 500Hz] bandwidth and magnified 1000 fold for electrophysiological recordings.
To check the stability and quality of ECoG recordings, these brain signals were analyzed to obtain

rP

Power Spectrum Density and measure of the amplitude in µV RMS (Root Mean Square). Additional

ee

tests like Visual/Somatosensory Evoked Potentials (VEP and SEP) have been performed using a
Micromed system (Micromed Italy). For VEP, visual stimuli ((Micromed Flash 10S, 1Hz, 230mJ, 100

rR

trials) were displayed in front of the rodent. For SEP, Electrical stimulation was applied through
bipolar electrodes implanted in the hind paw to stimulate the sciatic nerve (Energy light, 1Hz, 5 volts
and 200 microseconds pulse width, 300 trials).

iew

ev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

3. Behavioral Experiments

After electrode implantation, the animals were trained to push on a lever to receive a reward (figure
2A), ECoGs were recorded as well as the markers for each lever push (LP). The training was aimed at
inducing a repetitive simple motor behavior, such as pushing on a lever to obtain a food pellet from a
dispenser. This was easily obtained, regardless of the manner the lever push was obtained, as this was
not a study of the physiology of the motricity, but only of the brain cortex signature involved in
inducing the lever push. This study included several consecutives steps during one year of experiments
and a sketch illustrating them is available in the figure 2E.

FIGURE 2

Neuromodulation Proof

Page 4 of 40

Page 5 of 40

3.1. Training
The paradigm is a simple reward oriented task: The rat is freely moving in the ABETT behavioral cage
(Figure 2B). A wall-mounted lever activates the food dispenser (Dustless precision pellets, BioServ,
New-Jersey, USA). The food dispenser delivers a reward pellet for each LP. The average training time
was about 30 minutes per session and up to one hour, 3 days a week, for one year. The aim is to define
the specific neural activity patterns, called “signature,” corresponding to the behavioral task (LP
intention).

Fo

3.2. Online uncontrolled experiment
After successful determination of the ECoG parameters, or signature, predicting the LPs, the lever was

rP

disconnected from the food dispenser, which therefore could deliver a reward only when activated by

ee

the detection (true or false) of the signature (Figure 2C).

rR

3.3. Online controlled experiments

The signal from the lever was then used to validate the true positive detections of the ECoG signatures

ev

and to make the false positive detection unable to initiate the food delivery the food dispenser is

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

triggered by the online detected ECoG signatures only when they are validated by LP (Figure 2D).

SIGNAL PROCESSING

4. Signature identification algorithm
To detect specifically the intention to press the lever, the time period preceding the LP was selectively
investigated. Continuous Wavelet Transform (CWT) of ECoG was carried out to perform timefrequency analyses (Figure 3A and 3B Left). The Meyer wavelet was chosen as the mother wavelet
taking into account its computational efficiency [27 and 28]. Following this, the time-shifted crosscorrelation (CC) of LP signals and ECoG in time frequency domains (Figure 3B Right) for the 14
electrodes were calculated. The electrode providing the highest correlation with LP behavioral task

Neuromodulation Proof

Neuromodulation Proof

(#15, over the cerebellar cortex) was identified to construct the single-electrode predictor. Only ECoG
changes related to motor preparation and decision were considered in 1.5 seconds epochs preceding
the LP. The model was calibrated for this electrode by the INPLS method. The detailed description of
the INPLS algorithm is provided in mathematical paper published by our signal processing team [28].
The signature for the occurrence of LP is derived from the calibration procedure, creating a set of
efficient combinations of ECoG features in the frequency and temporal domains recorded on the single
electrode (Figure 3A and 3B) that exhibited the highest values of the CC (electrode 15, Figure 1E).
This signature is applied to all recordings of the same animal for verification.

FIGURE 3

rP

Fo

5. Online real time signal processing

ee

The signature recognition is applied immediately to the signal in real time during continuous
monitoring of neuronal activity at real-time BCI experimental sessions. When a signature is detected, a

rR

pulse is generated that can be used to trigger the food dispenser. During the online controlled
experiments, when the animal was trained, and the signature well established, the trigger was

ev

disconnected from the food dispenser: therefore the dispenser was triggered only by the detection of

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

the signature. The disconnected lever, which was now used only as a marker of the intention, still
recorded the pushing on the lever and proved that the pellet delivery was not occurring randomly.

6. BCI performance evaluation
During the experiments, the number of LP is N, and the number of decisions made by the algorithm
(2/second) is 2D (= D duration of recording in seconds x 2 decisions/second). When the signature has
been defined, the recordings of previous experiments were played back and again screened and the
software searched for the occurrence of the signature. We used these detected signatures and the LP
occurring in the initial experiment to count TP, FP, False Negative (FN), and True Negative (TN) On
the basis of these values and of what has been done previously [29 and 30], one can derive indexes
such as the true positive rate, TPR = TP/N = TP/(TP+FN); the true negative rate TNR = TN/(2D-N);

Neuromodulation Proof

Page 6 of 40

Page 7 of 40

and the number of FP per minute (FP/min). TPR represents the percentage of LP detected by the
algorithm.

7. Statistical significance and expected random detection
To analyze the statistical significance of the TPR, the results were compared to an expected random
detection. This was estimated using the model of repeated Bernoulli trials. Both LP and detections
were considered random events with probabilities which were estimated for each recording from
previous real experimental data. The expected TPR for random detection and the corresponding
confidence intervals (CI=0.95) were calculated from this distribution for each recording.

ee

rP

RESULTS

Fo

The number N of LP increased with time, manifesting in about 1 push per minute, which is the
essential starting point for these training behavioral experiments.

rR

1. Electrode Location and Time Course of Recorded ECoGs

ev

We were able to record from neural activity using our set up from each of the rats for approximately

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

one year. The patterns were consistent, and the platform was very stable.

1.1. Functional localization of the electrodes (obtained by VEP and SEP) and characteristics of
the ECoG recordings (amplitude and frequency):
The evolution of the ECoG signals amplitude during eight months illustrates the stability of the brain
signals along time (Figure 4A). The average Power Spectrum Density (PSD) of ECoG signals
computed over the same period was also stable (Figure 4B), these data are shown for 4 rats.
Occipital cortex electrodes (6 and 13) exhibit maximal amplitude VEP, with a first positive deflection
P1 (latency of 31 msec) and a negative potential N1 (latency of 57 msec) (Figure 4C). Electrodes 6
and 13 also exhibit maximal amplitude SEP (data not shown).

Neuromodulation Proof

Neuromodulation Proof

The colocalization of SEP and VEP, as seen on electrodes 6 and 13, is coherent with literature data
[31]. The figures 4D and 4E represent the evolution of the latencies of VEP (N1 and P1, see figure 4C)
during 8 months and confirms that the stability and maintained quality of the ECoG signals recorded
during our experiments.

FIGURE 4

1.2. Anatomical localization of task-related electrodes

FIGURE 5

rP

Fo

The correlation was maximal (**) on the most posterior electrodes 8 (left middle cerebellar, Repeated

ee

measures one-way ANOVA p≤0.01) and particularly in 15 (right middle cerebellar, Repeated
measures one-way ANOVA p≤0.01) (see Figure 5 and 1E). The correlation values of electrodes 6 and

rR

13 (as well as electrode 11) are strong but not reaching the level of significance. (Figure 5).
Table 2 offers more detailed results gathered from electrode 15. For each animal examined, this

ev

electrode always had the maximal correlation, averaging 0,3 ± 0,08 from 376 experiments in 10

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

animals. There was little variation between rats, with some having a correlation as high as ~0.4 (rats 1
and 9) and others having a correlation as low as ~0.2 (rats 4 and 6). Nevertheless, electrode 15 had the
highest correlation in all cases.

TABLE 2

1.3. Time Course and Average Frequency of Task-related ECoG Activity

The topography of the correlation varied along time on all electrodes. (Figure 6). Again, electrodes 8
and 15 showed the maximal correlation, being approximately 333ms before the LP in this example.
There was a milder activation in the electrodes over the occipital region (electrodes 6 and 13).

Neuromodulation Proof

Page 8 of 40

Page 9 of 40

The table 4 presents all values of frequency and time for the best correlation (E15) for 10 animals in
244 experiments.

FIGURE 6

Electrode 15 also had the maximal wavelet correlation, averaging 572 ± 207ms before the LP, in a
frequency band of 177 ± 36 Hz (Table 3).

TABLE 3

Fo

As before, there were some variations between animals, with one animal (rat 2) showing a much

rP

shorter time between correlation and LP (115ms) than others (e.g., rat 7: 929ms). There was much less

ee

variation between animals when considering the frequency band of interest (range from 147-244Hz).
In summary, our training experiments and the correlation analyses for each animal allowed us to

rR

identify a region of peak brain activity that is associated closely with the intention to press the lever to
obtain pellets. This region of peak activity was found consistently to be at E15.

2. Functional Significance of Recorded ECoGs

iew

ev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

2.1: Defining the ECoG signature: Calibration phase
Analysis of brain activities provided a stable signature in the high gamma band (~180Hz) occurring
within 1500 ms before the lever push (Figure 7).

FIGURE 7

The signature contains the temporal-frequency pattern corresponding to the intention to push the lever.
The INPLS generates a predictive regression model and the set of projectors (factors). To this purpose
the most informative electrode, i.e. the one with the highest correlation with LP behavioral task across

Neuromodulation Proof

Neuromodulation Proof

all the electrodes was identified. Then the predictive model was generated by the INPLS approach (the
number of factors was chosen equal to six).
The first three of factors are shown in figure 8 below. The relative weights of all six factors in the final
decomposition are 0.567, 0.228, 0.107, 0.051, 0.047, 0.020 (figure 8 below).

FIGURE 8

The principal frequency band of interest of these signatures was between 50Hz to 300Hz, and all
signatures occurred approximately 400ms preceding the LP, corresponding to time zero.

Fo

Notwithstanding small variations between animals, for each individual animal, an optimal signature
was identifiable and remarkably stable over time, as illustrated in figure 9. The TPR remains stable in

rP

offline analysis of 5 months of experiments and also during online controlled experiments during 3

ee

months. The FP/minutes are also represented in this figure and the same conclusion can be drawn.

FIGURE 9

ev

rR

2.2. Offline validation of the ECoG signature on training experiments

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Between all signatures obtained by playing back the previous experiments, the optimal signature for
each of the 10 animals is defined by the optimal TPR and the minimal FP per minute. (Table 4).

TABLE 4

For the 10 animals and 206 experiments the average percentage of the true positive rate (TPR) was
67.82 ± 4.13 with 2.17 ± 0.52 false positives/min and the average rate of the push/minute rate was
3.61, (4.54 (rat 5) for the best performing rat and 2.56 (rat 3) for the worst performing one. (Table 4).
In the figures 9A and 9B, the evolution along time of the TPR and FP/minute for these offline analysis
is shown. The performance remained stable along these 5 months.

Neuromodulation Proof

Page 10 of 40

Page 11 of 40

2.3. Comparison with random detection
This series of experiments tested whether the results of detection were not, in fact, totally random.
This random detection analysis was undertaken in 6 rats for offline analysis (n=105 experiments;
Table 5). The random detection cases were ~25 times lower than the real ones.

TABLE 5

3. Online Experiments

Fo

3.1: Online uncontrolled experiments
In this condition (Table 6), the number of motor events (LPs) quickly decreased, resulting in a high

rP

TPR (36.17±8.53%) even for a small number of LPs.

TABLE 6

rR

ee

For the 10 animals and 69 experiments the average percentage of the true positive rate (TPR) was

ev

36.17 ± 8.53 with 3.21±0.44 false positives/min and the average rate of the push/minute rate was
decreased strongly to 0.57±0.47 pushes/min. (Table 6).

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

3.2. Online controlled experiments
The 10 rats completed at least 13 experimental sessions. The animals were obliged to validate the
detection by pushing the lever (disconnected from the dispenser), so the number of pushes as a
consequence increases and the FPs decrease. Then the average values are TPR = 36.67±9.34% with
1.77±0.47 FP/min (Table 7).

TABLE 7

Neuromodulation Proof

Neuromodulation Proof

In the figures 9C and 9 D, the evolution along time of the TPR and FP/minute during these online
controlled experiments is shown. The performance remained stable along these 3 months.

4. Histological observations
In the figure 9, we have compared the slices of a brain with the rat atlas to check that electrodes have
recorded the areas of the brain. The electrodes 6 and 13 are on the posterior side of the brain, just
anterior to Lambda (in black in figure 9). The most posterior electrodes, 8 and 15 (our electrode of
interest for this study), are implanted posteriorly to the lambda: on the cerebellum (in orange in figure
9). The histological data, by staining the Nissl bodies (rough endoplasmic reticulum) confirmed that

Fo

the cellular components, did not show degeneration or necrosis.

DISCUSSION

ee

rP

BCI are being developed in several studies to provide solutions for handicapped persons, based on the
possibility to pilot effectors using recording of ECoG brain activity. Decoding of the relationships

rR

between brain activity and motor activity needs the development of mathematical paradigms, based on
animal experiments. The current paper reports the design of a rather inexpensive, easy to perform, set-

ev

up in rodent to develop and evaluate the paradigms elaborated to achieve the most efficient detection

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

of signatures. The study has provided three major findings. First, we developed a strategy to record
neural activity continuously at the level of several cortical electrodes for a long period of time, up to
12 months, in freely moving rats. Second, we developed a self-paced, asynchronous, unsupervised
BCI application to detect a signature activity (predominantly in the high gamma band), specific for
each rat and stable for as long as one year, that can activate an external effector by online detections.
Third, the peak neural activity producing the task-related signature preceding a motor task was
predominantly localized to the cerebellar cortex (which is coherent with postural component of the
task, standing on the hindlimbs to reach the lever), and to a lesser extent to occipital area and also the
motor area (which is coherent with the profile of activity during the task of raising and reaching the
body, and then pushing with the forelimbs the lever).

Neuromodulation Proof

Page 12 of 40

Page 13 of 40

I Experimental design for ECoG recording

A. Experimental set up
The present study showed that ECoG recordings provide high quality data for analysis with specific
algorithm software. This is one of the major findings of this study compared to previous studies
presenting different BCI systems based on recordings of ECoGs over several days but not a year, as in
the study by Blakely [32] that related a control of 7 days of implantation or EEG signals with acute
experiments each day, but not long-term [33-37].
This first step on rodents gave us a lot of information for future BCI experiments: Brain ECoGs can be

Fo

used up to one year, they are stable, usable for data processing during one year, the performances of
this BCI system in offline analysis were high (67,82% of True Positive Rate with a small 2.17 false

rP

positive/minute) and our results show encouraging data for the online use of this BCI (control of the

ee

effector by mental task). The accuracy of the algorithm appeared to be rather stable along the period of
the whole experimentation as shown by our results presented on the stability of the performance (TPR

rR

and FP/minute) during 5 months of offline analysis and 3 months of online controlled experiments.
Recent BCI publications on rats have shown studies of a similar binary task (push a lever to obtain a

ev

reward) but using corticospinal tract activity [38] or with detection from multi-unit intra cortical

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

signals [23]. The Hammad team [23], using a 4 x 4 intracortical tungsten micro-wire electrode array,
found a good detection of the motor task with an accuracy of 75 ± 6% when wavelet denoising was
applied (n=4, t=4weeks), thus illustrating the feasibility of a spinal cord computer interface (SCCI)
brain computer interface for generation of command signals in paralyzed individuals. Their results,
based on penetrating microelectrodes arrays, showed a correlation of 0.67 between measured and
predicted forces in the vertical direction (n=6, t=3weeks) [38]. Another study [24] presented a BCI
controlled by non-motor area of rat brain. Their results demonstrated that it is possible for rats
tocontrol a 1D machine with an encoding-based BCI system using prefrontal cortex activity to obtain a
reward (n=16, recording time=600min). These good results highlight the interest of our experiments
using similar motor binary task, but with detection stability up to one year. The limitation of this type

Neuromodulation Proof

Neuromodulation Proof

of experiment to determine the cortical area where the signature occurs, is that the task used is binary
simple behavior aimed at triggering the delivery of the rewarding pellet.
Most reported SBCIs detect less than 50% TP with 10-20 FP/minute [25, 39-42]. Fatourechi et al. [35]
presented an EEG-based BCI with 56% TP and 0.7 FP/minute. However, these offline results of short
recording periods (2-5 minutes) may translate to long-term clinical applications that require
continuous processing of neural activity.

B. Anatomical localization of peak activity to cerebellar cortex
Perhaps our most striking finding is localization of peak brain activity within the cerebellar cortex, a

Fo

region concerned predominantly with execution of motor tasks. The electrodes located in
retrolambdatic skull areas provided the highest correlation coefficient (see Results). This was

rP

surprising, because one would have expected to find highest correlation in the motor and sensorimotor

ee

areas of the cerebral cortex, as in humans [43]. There was some activity recorded in the occipital areas
and also a little in post-central regions cortical areas of rats, but this was not as high as the cerebellar

rR

recordings and not reaching the level of significance.

On closer examination of the gross structure of the rat brain, however, this apparent discrepancy is not

ev

so surprising. The highly folded cerebellar cortex of rats is probably more capable of global control of

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

movement than its much flatter cerebral cortex. In humans, the cerebellum, although highly folded as
well, is relatively small compared to its much larger and highly folded cerebral cortex. Furthermore,
motricity in rats is more globally postural because the outreaching movement of the forepaw is
accompanied by an overall rising of the body toward the lever. Hence, every voluntary movement in
rats, unlike in humans, is initiated with some postural stabilization. The postural nature of the
motricity involved in this task may explain why ECoG activity was very high in the cerebellum.

II. Brain Activity to Drive Effectors

Behavioral experiments

Neuromodulation Proof

Page 14 of 40

Page 15 of 40

The offline calibration experiments evaluate the performance of the algorithm to detect the signature,
which in these experiments is driven by high gamma band activity. Our self-paced BCI performs
continuous monitoring of neuronal activity. The high TPR (67.82±4.13%) and the small number of FP
(2.17±0.52/min) are important results in view of using signatures in real-time experiments, and
especially in asynchronous experiments (rare in current BCI studies).
The two next steps of our online BCI paradigm estimated the capability to control an external effector
by direct decoding of rodent brain activities.

Online uncontrolled experiments: Evidence for pure neural (mental) control of the pellet
delivery

Fo

The online uncontrolled experiments evaluated the algorithm performance in real time. Here, the lever

rP

was disconnected, then only the signature recognition could deliver pellets, regardless of LPs.

ee

Actually, compared to the calibration phase, the number of pushes (LP) decreases rapidly and
significantly, to 0.57 /min. However, the number of pellets provided by signature detection is not

rR

significantly different than in offline experiments, suggesting that detection of the signature has not
been changed, by the disconnection of the lever. The animal quickly learns that the reward is delivered

ev

when she desires a pellet, making it useless to push the lever, and she ultimately stops pushing on the

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

lever as the experiment progresses, making the calculated (TPR=TP/N) value of TPR (37%) irrelevant,
as the denominator tends to zero. TP or a FP equally deliver a pellet, and cannot be distinguished
because the lever is disconnected. It is a reasonable inference, then, that a proportion of the FPs
actually represents the intent of the animal to press the lever, although it makes no LP attempt, and
may correspond to a valid signature capture, activating the dispenser only by brain control. Therefore,
there is a proportion of false FP (FFP) that should be counted along with TP to calculate the true TPR.
It is impossible to determine the number of FFP in rodents.
Online controlled experiments
As described, the number of pellets delivered during online uncontrolled activation is not dependent
on the number of LPs. Therefore, online controlled experiments were designed to consider the
detected signatures signals enabled to activate the dispenser only if they are validated by a coinciding

Neuromodulation Proof

Neuromodulation Proof

LP within a window of 500 msec. The animals quickly learned this obligation and were then forced to
push on the lever to transform their ECoG activities into a triggering signal to deliver a pellet,
increasing the number of LPs at 3.24±0.7, a level not significantly different from during the offline
phase. TPR decreased to 36.67±9.34 lower than in the offline experiments. The more stringent
constraint of the paradigm (to get a reward, the rat is required to push on the lever AND to have a
brain signature detected) does not count a real intent not associated to a corresponding LP. It therefore
cannot trigger a pellet delivery as observed during the online uncontrolled paradigm. Then the number
of TP decreases.

CONCLUSIONS

Fo

This study has shown that epidural electrodes provide recordings of brain activity over extended

rP

periods and that the signals are compatible with multi-dimensional wavelet software analysis [28],

ee

providing detection of motor task related signatures, stable over time for up to one year, and usable
without repeated calibration in similar periods. The detection of these signatures allows online

rR

detection of intention of movements from cortical ECoG continuous recording in freely moving
animals. This unsupervised ECoG analysis allows piloting effectors and makes possible the design of

ev

an online BCI system to treat motor deficits in tetraplegic patients, particularly victims of cervical

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

spinal injuries. The experiments reported in this paper demonstrate the feasibility of long-term
asynchronous BCI command of effectors in freely moving rodents performing a behavioral task. The
efficiency and robustness of the INPLS detection algorithm, which provides self-paced detection of
motor-related ECoG feature, shed light on the cortical mechanisms and structures involved in
initiation and control of motor activity. The solid performance of this detection algorithm now meets
the criteria necessary to apply this system in a real-world setting for humans with neurological injury.
High true positive rate, coupled with a low false positive rate, and long-term stability without the need
for system recalibration will allow us to extend the application of this algorithm to a multi-axis
exoskeleton. To reach this goal, complementary experimental steps in rodents and primates have been
currently achieved in our laboratory, while in parallel we have designed, realized and tested a wireless
implantable epidural 64 contact recorder [44] for human implantation. Similarly, we have developed

Neuromodulation Proof

Page 16 of 40

Page 17 of 40

and tested a human exoskeleton aimed at providing to tetraplegic patient four-limb capability to
recover some mobility.

ACKNOWLEDGEMENTS
This project received financial support through grants from the French National Research
Agency (ANR-Carnot Institute), Fondation Motrice, Fondation Nanosciences (Chair of
Excellence Professorship to TA) and Fondation de l’Avenir. Région Rhone-Alpes was the
major contributor to the building of the Clinatec Institute, Fondation Philanthropique Edmond
J Safra provided a generous unrestricted educational grant to the Clinatec Institute. John
Mitrofanis (University of Sydney, Australia) edited the English writing.

REFERENCES

rP

Fo

1. Yuan H., He B.. Brain-computer interfaces using sensorimotor rhythms: current state and
future

perspectives.

IEEE

10.1109/TBME.2014.2312397.

ee

Trans

Biomed

Eng. 2014

May;

61(5):1425-35.

doi:

rR

2. Jeannerod M.. Mental imagery in the motor context. Neuropsychologia. 1995 33: 1419-32.
Review.

ev

3. Wolpaw, J.R., McFarland, D.J., Neat, G.W., and Forneris, C.A.. An EEG-based braincomputer interface for cursor control. Electroencephalogr. Clin. Neurophysiol. 1991. 78, 252-

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

259.

4. Kübler, A., Kotchoubey, B., Hinterberger, T., Ghanayim, N., Perelmouter, J., Schauer, M.,
et al.. The thought translation device: a neurophysiological approach to communication in
total motor paralysis. Exp. Brain Res. 1999. 124, 223-232.
5. Rouse, A.G., and Moran, D.W.. Neural adaptation of epidural electrocorticographic
(EECoG) signals during closed-loop brain computer interface (BCI) tasks. Conf. Proc. IEEE
Eng. Med. Biol. Soc. 2009. 5514-5517.
6. Torres Valderrama, A., Oostenveld, R., Vansteensel, M.J., Huiskamp, G.M., and Ramsey,
N.F.. Gain of the human dura in vivo and its effects on invasive brain signal feature detection.
J. Neurosc. Methods 2010. 187, 270-279.
7. Mehring, C., Nawrot, M.P., de Oliveira, S.C., Vaadia, E., Schulze-Bonhage, A., Aertsen,
A., et al.. Comparing information about arm movement direction in single channels of local

Neuromodulation Proof

Neuromodulation Proof

and epicortical field potentials from monkey and human motor cortex. J. Physiol. Paris 2004.
98, 498-506.
8. Leuthardt, E.C., Schalk, G., Wolpaw, J.R., Ojemann, J.G., and Moran, D.W.. A braincomputer interface using electrocorticographic signals in humans. J. Neural. Eng. 2004. 1, 6371.
9. Schalk, G., Kubánek, J., Miller, K.J., Anderson, N.R., Leuthardt, E.C., Ojemann, J.G., et
al.. Decoding two-dimensional movement trajectories using electrocorticographic signals in
humans. J. Neural. Eng. 2007. 4, 264-275.
10. Donoghue, J.P., Sanes, J.N., Hatsopoulos, N.G., and Gaal, G.. Neural discharge and local
field potential oscillations in primate motor cortex during voluntary movements. J.
Neurophysio. 1998. 79, 159–173.

Fo

11. Donoghue, J.P. Connecting cortex to machines: recent advances in brain interfaces.
Nature Neuroscience Supplement 2002. 5, 1085-1088.

rP

12. Nicolelis, MA.. Actions from thoughts. Nature 2001. 409, 403-7.
13. Nicolelis, MA.. Brain-machine interfaces to restore motor function and probe neural

ee

circuits. Nat. Rev. Neurosci. 2003. 4, 417-422.

14. Nicolelis, M.A., and Lebedev, M.A.. Principles of neural ensemble physiology underlying

rR

the operation of brain-machine interfaces. Nat. Rev. Neurosci 2009. 10, 530-540.
15. Hochberg, L.R., Serruya, M.D., Friehs, G.M., Mukand, J.A., Saleh, M. et al.. Neuronal

ev

ensemble control of prosthetic devices by a human with tetraplegia. Nature 2006. 442, 164174.

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

16. Velliste, M., Perel, S., Spalding, M.C., Whitford, A.S., and Schwartz, A.B.. Cortical
control of a prosthetic arm for self-feeding. Nature 2008. 453(7198), 1098-1101.
17. McFarland, D.J., Wolpaw, J.R.. Sensorimotor rhythm-based brain-computer interface
(BCI): feature selection by regression improves performance. IEEE Trans. Neural Syst.
Rehabil. Eng. 2005. Sep;13(3):372-9.
18. Williams, J.C., Rennaker, R.L., Kipke, D.R.. Long-term neural recording characteristics
of wire microelectrode arrays implanted in cerebral cortex. Brain Res. Protoc. 1999
Dec;4(3):303-13.
19. Schwartz, A.B., Cui, X.T., Weber, D.J., Moran, D.W.. Brain-controlled interfaces:
movement restoration with neural prosthetics. Neuron 2006. 5;52(1):205-20.
20. Margalit E., Weiland, J.D., Clatterbuck, R.E., Fujii, G.Y., Maia, M., Tameesh, M., et al..
Visual and electrical evoked response recorded from subdural electrodes implanted above the

Neuromodulation Proof

Page 18 of 40

Page 19 of 40

visual cortex in normal dogs under two methods of anesthesia. J. Neurosci. Methods 2003.
Mar 15;123(2):129-37.
21. Wang W., Collinger J.L., Degenhart A.D., Tyler-Kabara E.C., Schwartz A.B., Moran
D.W., Weber D.J., Wodlinger B., Vinjamuri R.K., Ashmore R.C., Kelly J.W., Boninger M.L..
An electrocorticographic brain interface in

an individual with tetraplegia.

PLoS

One. 2013;8(2):e55344. doi: 10.1371/journal.pone.0055344. Epub 2013 Feb 6.
22. Nakanishi Y., Yanagisawa T., Shin D., Kambara H., Yoshimura N., Tanaka M., Fukuma
R., Kishima H., Hirata M., Koike Y.. Mapping ECoG channel contributions to trajectory and
muscle activity prediction in human sensorimotor cortex. Sci Rep. 2017 Mar 31;7:45486. doi:
10.1038/srep45486.
23. Hammad S.H.H., Farina D., Kamavuako E.N., and Jensen W.. Identification of a self-

Fo

paced hitting task in freely moving rats based on adaptive spike detection from multi-unit M1
cortical signals. Front. Neuroeng. 2013. Nov 15;6:11. doi: 10.3389/fneng.2013.00011.
eCollection 2013.

rP

24. Lang, Y., Du, P., Shin, H.C.. Encoding-based brain-computer interface controlled by non-

ee

motor area of rat brain. Sci. China Life Sci. 2011. Sep;54(9):841-53. doi: 10.1007/s11427011-4214-6. Epub 2011 Sep 16.

rR

25. Townsend, G., Graimann, B., and Pfurtscheller, G.. Continuous EEG classification during
motor imagery-simulation of an asynchronous BCI. IEEE Trans. Neural Syst. Rehabil. Eng.

ev

2004. 12, 258-65.

26. Swanson, L.W.. Brain Maps: Structure of the Rat Brain (with CDROM) Spiral Bound,

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

(Amsterdam, Netherlands : Elsevier Sciences and Technology Editor), 2003.
27. Sherwood J. and Derakhshani R.. 2009. Proc. Int. Joint Conf. on Neural Networks. pp
2508–15.

28. Eliseyev A., Moro C., Costecalde T., Torres N., Gharbi S., Mestais C.et al.. Iterative Nway partial least squares for a binary self-paced brain-computer interface in freely moving
animals. J. Neural. Eng. 2011. Aug; 8(4):046012. Epub 2011 Jun 10.
29.

Rijsbergen,

C.J.

1979.

Information

retrieval.

Available

at

htttp://www.dcs.gla.ac.uk/~iain/keith/.
30. Huggins J.E., Levine S.P., BeMent S.L., Kushwaha R.K., Schuh L.A., Passaro E.A. et al..
Detection of event-related potentials for development of a direct brain interface. J. Clin.
Neurophysiol. 1999. Sep;16(5):448-55.
31. Vasconcelos, N., Pantoja, J., Belchior, H., Caixeta, F.V., Faber, J., Freire, M.A., et al..
Cross-modal responses in the primary visual cortex encode complex objects and correlate

Neuromodulation Proof

Neuromodulation Proof

with tactile discrimination. Proc. Natl. Acad. Sci. USA. 2011. Sep 13;108(37):15408-13. doi:
10.1073/pnas.1102780108. Epub 2011 Aug 29. Erratum in: Proc Natl Acad Sci U S A. 2012
Jan 17;109(3):995.
32. Blakely, T., Miller, K.J., Zanos, S.P., Rao, R.P. and Ojemann, J.G.. Robust, long-term
control of an electrocorticographic brain-computer interface with fixed parameters.
Neurosurg. Focus. 2009. Jul;27(1):E13. doi: 10.3171/2009.4.FOCUS0977.
33. Piccione, F., Giorgi, F., Tonin, P., Priftis, K., Giove, S., Silvoni, et al.. P300-based brain
computer interface: reliability and performance in healthy and paralysed participants. Clin.
Neurophysiol. 2006. Mar;117(3):531-7. Epub 2006 Feb 2.
34. Mak, J.N., McFarland, D.J., Vaughan, T.M., McCane, L.M., Tsui, P.Z., Zeitlin, D.J., et
al.. EEG correlates of P300-based brain-computer interface (BCI) performance in people with

Fo

amyotrophic lateral sclerosis. J Neural Eng. 2012. Apr;9(2):026014. doi: 10.1088/17412560/9/2/026014. Epub 2012 Feb 21.

rP

35. Fatourechi, M., Ward R.K., and Birch, G.E.. A self-paced brain-computer interface system
with a low false positive rate. J. Neural. Eng. 2008a. 5, 9-23.

ee

36. Fatourechi, M., Ward, R.K., and Birch, G.E.. Evaluating the performance of a self-paced
BCI with a new movement and using a more engaging environment. Conf. Proc. IEEE Eng.
Med. Biol. Soc. 2008b. 650-653.

rR

37. Fatourechi, M., Ward, R.K., and Birch, G.E.. Performance of a self-paced brain computer

ev

interface on data contaminated with eye-movement artifacts and on data recorded in a
subsequent session. Comput. Intell. Neurosci. 2008c. 749204.

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

38. Guo, Y., Foulds, R.A., Adamovich, S.V., Sahin, M.. Encoding of forelimb forces by
corticospinal tract activity in the rat. Front. Neurosci. 2014. May 1;8:62. doi:
10.3389/fnins.2014.00062. eCollection 2014.

39. Mason, S.G., and Birch, G.E.. A brain-controlled switch for asynchronous control
applications. IEEE Trans. Biomed. Eng. 2000. 47, 1297–1307.
40. Yom-Tov, E., and Inbar, G.F.. Detection of movement-related potentials from the electroencephalogram for possible use in a brain-computer interface. Medical and Biological
Engineering and Computing 2003. 41(1), 85-93.
41. Bashashati, A., Mason, S., Ward, R.K., and Birch, G.E.. An improved asynchronous brain
interface: making use of the temporal history of the LF-ASD feature vectors. J. Neural. Eng.
2006. 3, 87–94.
42. Bashashati, A., Ward, R.K., and Birch, G.E.. Towards development of a 3-state self-paced
brain computer interface. J. Computational Intelligence & Neuroscience 2007. 84386.

Neuromodulation Proof

Page 20 of 40

Page 21 of 40

43. Wallace, M.T., Ramachandran, R., and Stein, B.E.. A revised view of sensory cortical
parcellation. PNAS 2004. 101, 2167–2172.
44. Mestais CS, Charvet G, Sauter-Starace F, Foerster M, Ratel D, Benabid AL..
WIMAGINE: wireless 64-channel ECoG recording implant for long term clinical
applications. IEEE Trans Neural Syst Rehabil Eng. 2015. Jan;23(1):10-21. doi:
10.1109/TNSRE.2014.2333541. Epub 2014 Jun 30.

iew

ev

rR

ee

rP

Fo

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

Reviewer: 1
Comments to the Author
The manuscript continues to lack necessary methodological description as regards the instrumentation and
physiological assessment of the model.
We do not really understand what Referee 1 really wants. We have added in the previous revision many
descriptions which seemed necessary to the manuscript, how far should we go? This paper is mostly aimed
at the results and the usefulness of the rat model for BCI studies, the methodological details are here to
show how we did it and in our opinion were sufficient for this paper.
Reviewer: 2
The authors thank reviewer 2 for precise and useful suggestions. We have tried to answer properly his
remarks and included adequate and detailed changes in the text.
Comments to the Author
The authors in this version have addressed the reviewers comments to some extent and the manuscript has
become easier to follow and understand the overall work although it lacks a lot of details. The authors have
justified the lack of details (in their cover letter) as their aim of this manuscript was to summarize their
previous studies. So far I have the following comments:

rP

Fo

L9,p1: please add a reference.
We have added a reference for BCI Neuroprosthesis in the text: [1] and in the bibliography:
“Yuan H, He B. Brain-computer interfaces using sensorimotor rhythms: current state and future
perspectives. IEEE Trans Biomed Eng. 2014 May; 61(5):1425-35. doi: 10.1109/TBME.2014.2312397.”

ee

L28 - 44, p1: please add a reference.
We have added two references for ECoG BCI advantages in the text: [21 AND 22] and in the bibliography:
“21. Wang W., Collinger J.L., Degenhart A.D., Tyler-Kabara E.C., Schwartz A.B., Moran D.W., Weber
D.J., Wodlinger B., Vinjamuri R.K., Ashmore R.C., Kelly J.W., Boninger M.L.. An electrocorticographic
brain interface in an individual with tetraplegia. PLoS One. 2013;8(2):e55344. doi:
10.1371/journal.pone.0055344. Epub 2013 Feb 6.”
“22. Nakanishi Y., Yanagisawa T., Shin D., Kambara H., Yoshimura N., Tanaka M., Fukuma R., Kishima
H., Hirata M., Koike Y.. Mapping ECoG channel contributions to trajectory and muscle activity prediction
in human sensorimotor cortex. Sci Rep. 2017 Mar 31;7:45486. doi: 10.1038/srep45486.”

iew

ev

rR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

We have added some references already in this manuscript for synchronous [5, 8, 10 and 15] and
asynchronous [23-25].
L46, p3: It would be better if you add the current amplitude used in the stimulation.
The current amplitude used in the stimulation during SEP experiments is not the same in all experiments,
the value is determined (in Ampere) by the minimal current needed to achieve the threshold of muscular
activity in the limb stimulated (upper limb for median nerve stimulation or lower limb for sciatic nerve
stimulation) that’s why no value could be indicated in the manuscript.
L15-17, p5: Please check the reference at (Sherwood and Derakhshani 2009 and ref 21).
We have modified the notification of these references in the text [27 and 28] and in the bibliography
paragraph:
“27. Sherwood J. and Derakhshani R.. 2009. Proc. Int. Joint Conf. on Neural Networks. pp 2508–15.”
“28. Eliseyev A., Moro C., Costecalde T., Torres N., Gharbi S., Mestais C.et al.. Iterative N-way partial
least squares for a binary self-paced brain-computer interface in freely moving animals. J. Neural. Eng.
2011. Aug; 8(4):046012. Epub 2011 Jun 10.”
L52, p7 “positive deflection P1 (latency of 31.2 msec) and a negative potential N1 (latency of 56.6 msec)”,
I believe these latency values are total average value aren’t they? If yes please clarify by mentioning the

Neuromodulation Proof

Page 22 of 40

Page 23 of 40

averaging way (e.g. over rats over days etc.).
These values for P1 and N1 are not the average of several experiments or several rats, it’s only the values
obtained by analysis of the example of the topoplot on the figure 4C on the Right panel. The value has
been rounded and the modification has been done in the manuscript IN THE PARAGRAPH 1.1.
Functional localization of the electrodes (obtained by VEP and SEP) and characteristics of the ECoG
recordings (amplitude and frequency):
“positive deflection P1 (latency of 31 msec) and a negative potential N1 (latency of 57 msec) (Figure 4C)”.
In the figure 4C, the Left panel has been modified to show the same VEP than the topoplot to have the
same values of P1 and N1, we have done the modification of the P1 value in the legend (previous figure 4C
Left panel: P1=34ms and with the new figure 4C Left panel, P1=31ms).
The modification in the legend has been done: “(C) Responses to flash light stimulation (1Hz, 107 flashes),
on the left: example of Visual Evoked Potential (P1: 31ms; N1: 57ms P2: 64ms), on the right: temporospatial distributions of their amplitude (Visual EP Amplitude Topoplot);”
L32-36, p8 “left middle cerebellar, Repeated measures one-way ANOVA p ≤0.01) and particularly in 15
(right middle cerebellar, Repeated measures one-way ANOVA p≤0.01)” what were the correlation values
compared by the ANOVA test? I mean were the ANOVA tests applied for comparing the cross-correlation
between these electrodes and all other electrodes or just the electrodes in the vicinity? Please clarify it.
The ANOVA test was applied on the correlations values for each electrodes in comparison with all the
others electrodes and only two electrodes (8 and 15) reached the level of significance with all the others not
only in the vicinity.

rP

Fo

L44 on p8 as has been stated the electrode 15 had always the highest correlation in all cases, so was there
any electrode that consistently had the lowest correlation? And what was the total average of the
correlation over the different electrodes over these experiments and the different rats? Have you investigate
the relation between the correlation value and the duration of the experiment?
Yes, some electrodes had lower correlations (E2 and E9 have the lowest) but no significance level
appeared for lowest correlations. It is not possible to directly make a link between the correlation value and
the efficiency.
Concerning the average of the correlation over the different electrodes, the figure 5 presented the histogram
of the mean correlations values on the 14 electrodes in 376 experiments on 10 rats. The results for each
animal separately has not been presented because the shape of this histogram is similar for each (electrodes
8 and 15 with the best correlation and reaching the level of significance) and the inter animals variation is
represented by the Standard Deviation in the Histogram of the figure 5.
We have calculated the average of the correlation for all electrodes and all experiments for all rats and the
value obtained is 0,18 as compared to the global correlation on E15 which is 0,3.
We have not investigated the relation between the duration of the experiments and the correlation values.
We can just say that the best correlations observed are not systematically obtained on the longest
experiments.

iew

ev

rR

ee

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

L13, P10 “To this purpose the most informative electrode, i.e. the one with the highest correlation with LP
behavioral task across all the electrodes was identified” what about the detection? Have you used the
ECoG signal from all electrodes or just from the most informative electrodes for the offline and online
detection?
The best electrode, with the best correlation (Electrode 15) has been used to calibrate the model of
prediction and we used only this electrode for the off-line analysis and for the on-line detection of the
signature.
Similar to the above comment, what are number of the electrodes used for calculating these presented data
in Table 3 and 4 i.e all electrodes or just the best?
We used only Electrode 15 for off-line analysis and for on-line detection. In the table 3 we shown the
values of frequency and time corresponding to the best correlation on the electrode 15 and in the table 4 the
off-line detection of the signature has also been done on the Electrode 15, as for the results summarized in
the table 5, 6 and 7.

Neuromodulation Proof

Neuromodulation Proof

L11-17,p14 “The Hammad team [32], using a 4 x 4 intracortical tungsten micro-wire electrode array, found
a good detection of the motor task with an accuracy of 75 ± 6% when wavelet denoising was applied (n=4,
t=4weeks), thus illustrating the feasibility of a spinal cord computer interface (SCCI) for generation of
command signals in paralyzed individuals”
This sentence can be comprehended as the authors of [32] have also done spinal cord BCI which
contradicts its preceding sentence where only the authors of [31] have performed corticospinal BCI as
mentioned ”Recent BCI publications on rats have shown studies of a similar binary task (push a lever to
obtain a reward) but using corticospinal tract activity [31] or with detection from multi-unit intra cortical
signals [32].”
We agree with this comment and modify the sentence in the text: “thus illustrating the feasibility of a brain
computer interface for generation of command signals in paralyzed individuals”

iew

ev

rR

ee

rP

Fo

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 24 of 40

Page 25 of 40

rP

Fo
Figure 1: Anatomical Localization of the Recording Cortical Electrodes. On the left: (A) Skull distribution; (B)
intracranial penetration of titanium screws; (C) cortical fingerprints of the electrodes; (D) anatomical
distribution of the electrodes viewed on atlas sections [26] coronal (above) and sagittal views (below). λ:
Lambda Suture; (E) from back to front, electrodes are situated over the cerebellum (retro-lambdatic: 8, 15),
occipital (visual area: 6, 13), postcentral (4, 11), precentral (3, 10), and prefrontal (2, 9) cortices. 4
additional electrodes are temporal, left and right: anterior (5, 12), and posterior (7, 14). The 3 reference
electrodes are numbered (1). On the right: Comparison of the brain slices (cresyl violet staining) with atlas
(sagittal and coronal slices) and the implantation map. (F) slice of the brain with footprints of electrodes 6
and 13 (in black); (G) slice of the cerebellum with electrodes 8 and 15 (in orange).

rR

ee

177x87mm (150 x 150 DPI)

iew

ev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

rR

ee

rP

Fo
Figure 2: Experimental set-up and Paradigms of this study. (A) Picture of a rat pressing the lever in the
behavioral box; (B) Experimental set-up and Paradigms for training experiments. The rat presses a lever
producing a pulse activating the reward dispenser, to deliver a pellet of food. The ECoG and the square
pulse of the lever are recorded simultaneously. The off-line ECoG data processing provides the correlation
coefficient (CC) for all electrodes; (C) Experimental set-up and Paradigms for on-line uncontrolled
experiments. The ECoG and the square pulse of the lever are recorded simultaneously. The food dispenser is
triggered by the on-line detected ECoG predictors only; (D) Experimental set-up and Paradigms for on-line
controlled experiments. The ECoG and the square pulse of the lever are recorded simultaneously. The food
dispenser is triggered by the on-line detected ECoG predictors only when they are validated by LP within a
time interval of 500ms. (E) Sketch of the consecutive steps of the global paradigm and the duration of each
of them.

iew

ev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

255x183mm (150 x 150 DPI)

Neuromodulation Proof

Page 26 of 40

Page 27 of 40

iew

ev

rR

ee

rP

Fo

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Figure 3: Off-Line calibration of the algorithm. (A) continuous Wavelet Transform (time-frequency analysis)
of subsequent blocks of ECoG data; (B) principal projections to low dimensional space (latent variables)
which are applied to the Wavelet Transformed ECoG.
140x196mm (150 x 150 DPI)

Neuromodulation Proof

Neuromodulation Proof

rP

Fo
Figure 4: ECoG signals analysis. (A) Amplitude of the ECoGs over a period of eight months for eight animals
and mean amplitude (dotted line indicates the trend curve of this mean). The mean amplitude shows a small
decrease during this long period and the trend curve (dotted line) illustrates a nearly stability (see equation:
y= -0.2271x + 66.86); (B) Mean Power Spectrum Density for four animals. The dotted lines corresponds to
the standard deviation and the black line to the average during eight months. There is no important
variation observed on the PSD; (C) Responses to flash light stimulation (1Hz, 107 flashes), on the left:
example of Visual Evoked Potential (P1: 31ms; N1: 57ms P2: 64ms), on the right: temporo-spatial
distributions of their amplitude (Visual EP Amplitude Topoplot); (D) and (E) the time-course of VEP Latencies
of peaks P1 and N1 respectively, selected along time at different moments.

ev

rR

ee

332x190mm (150 x 150 DPI)

iew

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 28 of 40

Page 29 of 40

rP

Fo
Figure 5: Histograms of the mean correlations values obtained by analysis of ECoG signals with algorithms
on the 14 electrodes (n=10 rats, 376 experiments). ** : p≤0.01

ee

180x103mm (96 x 96 DPI)

iew

ev

rR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

Figure 6: Example of spatial distribution of the correlation values along time in one experiment. The value of
correlation is indicated by a color code (red=maximal values). Time 0 corresponds to a LP. This figure is only
qualitative, it is one example of evolution along the time of the correlation over the 14 electrodes.

Fo

134x48mm (150 x 150 DPI)

iew

ev

rR

ee

rP

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 30 of 40

Page 31 of 40

Fo

Figure 7: Factors of signatures identified in 10 animals: Mean (blue line) and standard deviation (dotted
line). The ECoG signatures are highly comparable between animals: among the latent variables, Frequency
(Right graph) is the most important factor, in the high gamma band, Time (Left graph) is the second most
influential factor.

ee

rP

190x87mm (96 x 96 DPI)

iew

ev

rR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

ee

rP

Fo
Figure 8. Results of calibration for the best electrode: (A) frequency and temporal projections of the first
three factors; (B) weights of the factors in the final decomposition.

rR

208x129mm (96 x 96 DPI)

iew

ev

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 32 of 40

Page 33 of 40

Fo

Figure 9: (A) and (B) represent the time course of TPR and FP/minute respectively obtained for the offline
analysis (5 months); (C) and (D) represent the time course of TPR and FP/minute during the online
controlled experiments, respectively (3 months).

rP

330x146mm (150 x 150 DPI)

iew

ev

rR

ee

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

Author

Reference

Signal recorded

Model

Paradigm

Effector

Efficiency

Wolpaw, J.R.

2

EEG

Human

Synchronous

1 D Cursor control

-----

Rouse, A.G.

4

Epidural electrode

Primate

Closed loop

2D Computer cursor

----

Leuthardt, E.C.

7

Subdural electrode

Human

Synchronous

Cursor control

74–100% final accuracy

Donoghue, J.P.

9

Microelectrode arrays

Primate

Synchronous

----

----

Hochberg, L.R.

14

Microelectrode arrays

Human

Synchronous

Prosthetic hand

----

Fatourechi, M.

28

EEG

Human

Synchronous

----

56% TP and 0.7 FP/minute

Guo, Y.

31

Intra cortical

Rat

Asynchronous

Lever press

Correlation 0,67 for 6
animals in offline analysis

Hammad, S.H.H.

32

Intra cortical

Rat

Synchronous

Food reward

75 ± 6% TP for 4 rats, 4
weeks

Lang, Y.

33

Fo
prefrontal cortex

Rat

Synchronous

1D machine (water

56 trials, 16 animals for

dish)

600min total

1 D Cursor control

less than 50% TP with 10-

activity

Townsend, G.

36

rP

EEG

Human

Asynchronous

20 FP/minute

ee

Table 1: Overview of some BCI studies with different approaches, paradigms and models.

iew

ev

rR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Page 34 of 40

Neuromodulation Proof

Page 35 of 40

Correlation E15
Rat 1 (n=41)

0.41±0.11

Rat 2 (n=63)

0.32±0.09

Rat 3 (n=32)

0.29±0.06

Rat 4 (n=32)

0.21±0.04

Rat 5 (n=32)

0.24±0.05

Rat 6 (n=32)

0.22±0.05

Rat 7 (n=25)

0.27±0.07

Rat 8 (n=38)

0.34±0.15

Rat 9 (n=56)

0.4±0.11

Rat 10 (n=25)

0.28±0.1

Average. SEM

0.3±0.08

rP

Fo

Table 2: Values of best correlation (E15) for 10 animals in 376 experiments. Experiments
were considered for off line analysis only if LP was superior to 30, which is necessary for a
robust and efficient signal processing.

iew

ev

rR

ee

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

Time mean

Frequency mean

(ms before lever)

(Hz)

Rat 1 (n=16)

566±307

244±61

Rat 2 (n=16)

115±125

147±39

Rat 3 (n=16)

768±327

198±89

Rat 4 (n=16)

568±242

153±76

Rat 5 (n=12)

637±271

168±80

Rat 6 (n=28)

591±252

158±78

Rat 7 (n=28)

929±497

169±73

Rat 8 (n=28)

482±411

178±97

Rat 9 (n=28)

367±204

210±92

Rat 10 (n=28)

700±456

144±3

527±207

177±36

Average. SEM

ee

rP

Fo

Table 3: Values of frequency and time for the best correlation (E15) for 10 animals in 244
experiments. Experiments were considered for off line analysis only with the best correlation
on E15.

iew

ev

rR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 36 of 40

Page 37 of 40

TPR (%)

FP/min

Push/min

Rat 1 (n=16)

67.99±11.64

2.44±0.67

3.39±1.43

Rat 2 (n=16)

68.95±4.64

2.36±0.65

3.31±1.01

Rat 3 (n=16)

65.23±5.39

2.42±0.78

2.76±1.2

Rat 4 (n=16)

64.19±4.8

2.23±0.79

2.56±1.35

Rat 5 (n=12)

77.95±12.40

0.83±0.35

4.54±1.95

Rat 6 (n=29)

64.44±19.86

1.79±0.78

2.92±1.73

Rat 7 (n=27)

69.04±7.76

2.53±0.98

4.47±1.68

Rat 8 (n=22)

63.79±5.77

2.39±0.92

4.53±2.62

Rat 9 (n=35)

67.98±8.09

2.17±0.97

3.52±2.08

Rat 10 (n=17)

68.66±9.87

2.5±1.24

4.11±2.17

Average. SEM

67.82±4.13

2.17±0.52

3.61±0.76

ee

rP

Fo

Table 4: Table represents data for 10 animals for off-line analysis of 206 experiments.
Experiments were considered for off line analysis only if LP was superior at 1 per minute.

iew

ev

rR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

Ration TPR/

TPR random (%)

TPR (%)

Off-line Rat 1 (n=16)

3.53

67.99

19.26

Off-line Rat 2 (n=16)

2.79

68.95

24.71

Off-line Rat 3 (n=16)

2.26

65.23

28.86

Off-line Rat 4 (n=16)

2.06

64.19

31.16

Off-line Rat 5 (n=12)

3.9

77.95

19.99

Off-line Rat 6 (n=29)

2.6

64.44

24.78

Average. SEM

2.86±0.72

68.06±5.05

24.66±4.31

TPR random

rP

Fo
Table 5: The data for off-line analysis of experiments compared with “random detection”. The
values of percentage of True Positive are indicated. The signature has been detected in a
randomly manner for the TPR random.

iew

ev

rR

ee

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 38 of 40

Page 39 of 40

TPR (%)

FP/min

Push/min

Rat 1 (n=2)

35.71±5.51

3.39±0.07

0.15±0.06

Rat 2 (n=2)

46.67±18.86

3.78±0.38

0.23±0.06

Rat 3 (n=2)

37.50±17.68

3.80±0.17

0.16±0.11

Rat 4 (n=2)

50.79±8.98

3.04±0.21

0.27±0.02

Rat 5 (n=4)

29.54±8.68

2.78±0.46

0.32±0.17

Rat 6 (n=2)

25.60±10.94

2.78±0.21

0.47±0.46

Rat 7 (n=15)

39.54±10.41

3.62±1.08

1.43±0.98

Rat 8 (n=15)

25.50±7.71

3.37±1.81

1.35±0.78

Rat 9 (n=15)

30.62±5.12

2.56±0.73

0.77±0.33

Rat 10 (n=10)

40.25±14.65

3.00±1.47

0.50±0.28

Average. SEM

36.17±8.53

3.21±0.44

0.57±0.47

rP

Fo
Table 6: The data for 10 animals for uncontrolled on-line experiments.

iew

ev

rR

ee

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Neuromodulation Proof

Neuromodulation Proof

TPR (%)

FP/min

Push/min

Rat 1 (n=13)

36.65±13.80

1.48±0.27

3.50±1.18

Rat 2 (n=13)

19.25±15.00

1.94±0.47

3.45±1.13

Rat 3 (n=13)

22.49±16.05

1.91±0.44

3.04±0.94

Rat 4 (n=13)

35.64±17.18

1.55±0.42

3.38±1.27

Rat 5 (n=14)

50.26±13.77

1.46±0.32

3.21±0.99

Rat 6 (n=15)

39.87±19.04

1.31±0.74

1.73±0.93

Rat 7 (n=25)

39.47±7.24

1.24±0.57

4.51±1.74

Rat 8 (n=22)

39.51±5.77

1.76±0.88

3.65±1.48

Rat 9 (n=24)

44.63±8.24

2.33±1.09

3.07±1.91

Rat 10 (n=20)

38.92±6.32

2.70±0.87

2.85±1.45

Average. SEM

36.67±9.34

1.77±0.47

3.24±0.70

rP

Fo

Table 7: The data for “direct” on-line experiments for 10 animals in 172 experiments.

iew

ev

rR

ee

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

Neuromodulation Proof

Page 40 of 40


Documents similaires


Fichier PDF new interpretation of libet s experiments
Fichier PDF 201700003722 ratel d
Fichier PDF 1025
Fichier PDF elec1 2
Fichier PDF brochure rectifier info 1
Fichier PDF prescott 2009


Sur le même sujet..