Παρακαλώ χρησιμοποιήστε αυτό το αναγνωριστικό για να παραπέμψετε ή να δημιουργήσετε σύνδεσμο προς αυτό το τεκμήριο:
https://hdl.handle.net/123456789/1621| Τύπος: | Ανακοίνωση σε συνέδριο |
| Τίτλος: | Observation to execution cross-decoding of hand grips from ventral premotor cortex spiking activity |
| Συγγραφέας: | [EL] Τζαμαλή, Ελευθερία[EN] Tzamali, Eleftheria [EL] Παπαδουράκης, Βασίλειος[EN] Papadourakis, Vasileios [EL] Ράος, Βασίλειος[EN] Raos, Vassilis |
| Ημερομηνία: | 25/02/2021 |
| Περίληψη: | Our goal was to (i) explore whether hand grips can be decoded from spiking activity of the ventral premotor cortex (PMv) and, most critically, (ii) investigate the reliability and efficiency of cross-decoding i.e. whether a model trained with the action observation-elicited activity can effectively and consistently decode hand configurations using activity during action execution. Mirror neuron (MirN) activity was recorded from the ventral premotor cortex (n=122) of macaque monkeys who either reached and grasped objects using distinct grips (n=4), or observed the same movements executed by the experimenter. Support vector machines (SVM) were used for decoding. Using the entire PMv population, the algorithm correctly predicts all grips, both in observation and execution conditions. When k-subset of neurons is randomly selected, 50 (30) neurons are sufficient to achieve 95% mean accuracy in observation (execution) across 100 runs. The size of the population required to reach high accuracy levels was also estimated using a greedy-selection procedure: the unit with the best performance was initially selected and at each subsequent step a unit was added to those of the previous step so that the performance of the resulting population was the highest. Following this procedure five and two neurons were sufficient to achieve 100% accuracy during observation and execution, respectively. The mean accuracy of cross-decoding using either the entire population or randomly selected ensembles was around 25%, close to random. However, the application of the greedy-selection procedure resulted in a cross-decoding accuracy of 97% using only 6 neurons. This is the first study to report high cross-decoding accuracies of hand configurations from spiking cortical activity. This finding renders PMv, and MirNs in particular, promising for the development of accurate decoders that can control distinctive hand shapes in applications where activity and behavior cannot be co-recorded as in the case of paralysis. |
| Γλώσσα: | Αγγλικά |
| Τόπος δημοσίευσης: | Virtual Conference |
| Σελίδες: | 1 |
| Θεματική κατηγορία: | [EL] Νευροεπιστήμες[EN] Neurosciences |
| Κάτοχος πνευματικών δικαιωμάτων: | © The Author(s) 2021 |
| Ηλεκτρονική διεύθυνση του τεκμηρίου στον εκδότη: | https://www.cosyne.org/s/Cosyne2021_program_book.pdf |
| Ηλεκτρονική διεύθυνση περιοδικού: | https://www.cosyne.org/past-conferences |
| Όνομα εκδήλωσης: | Computational and Systems Neuroscience (COSYNE 2021) |
| Τοποθεσία εκδήλωσης: | USA |
| Ημ/νία έναρξης εκδήλωσης: | 23/02/2021 |
| Ημ/νία λήξης εκδήλωσης: | 26/02/2021 |
| Εμφανίζεται στις συλλογές: | Ερευνητικές ομάδες |
Αρχεία σε αυτό το τεκμήριο:
Το πλήρες κείμενο αυτού του τεκμηρίου δεν διατίθεται προς το παρόν από το αποθετήριο