Aim:

In recent years, the cerebellum has been proposed to play a dual role in motor control and adaptation. From the one hand. it may work as an inverse-dynamics model of the muscolo-skeletal system by providing corrective feedback to the motor commands generated by the motor cortex. From the other hand, the cerebellum may be considered as a forward model, which predicts the sensory consequences of a motor command. 

The aim of the thesis is to further investigate the dual nature of the cerebellum, by replicating complex sensorimotor tasks in a virtual robotic environment.

Project phases:

• Literature review
• Transfer of the control system to the virtual environment
• Simulation and data analysis
• Discussion of the results and comparison with the literature

Requirements:

• Basic knowledge of Python 

Aim:

Neural decoding is a critical step in BCI technologies. Different machine learning algorithms have been used to guide neural prosthetic limbs but results are far to get close at the natural body performance. In the last years, the availability of multi-electrode array system equipped with more and more recording channels is requiring a big data approach necessary. Together with the rising in computing power, the artificial neural networks (ANN) are a promising tool to address neural decoding problem. We propose to take advantage of modern ANN implementations to decode motor intentions from neural data recorded from behaving monkeys. Exploring different ANN architectures, the aim is to increase decoder robustness and reliability to be effectively implemented in clinical neuroprosthesis.

Project phases:

• Literature research
• Implementation of the ANN architecture
• Data analysis
• Interpretation and discusion of the results

This thesis foresees a development part at Scuola Superiore Sant’Anna in Pisa.

Aim:

During tasks and actions of humans and animals, neuronal populations communicate using a complex interplay of transient oscillatory rhythms. Neural signals generated by the activity of neuronal populations also display this type of transient behavior. This makes non-stationary spectral analysis of neural signals using wavelet functions a useful tool to investigate neural activity.

Wavelet–based techniques are suitable for this type of signals since they possess one key property called multiresolution. Multiresolution allows to accurately determines low frequency components and simultaneously localize time rapidly transient events.

Here, we propose to analyze extracellular signals recorded from multi-electrode array implanted in the posterior parietal area of monkeys during reach-to-grasp and reach-to-target task. Specifically, we will consider the low frequency part (< 300 Hz) of the extracellular field, called Local Field Potential (LFP). LFPs will be analyzed using wavelet-based techniques, such as Continuous Wavelet Transform and Wavelet Coherence to determine which type of scales (frequencies) characterizes the different tasks both at single and multi-electrode levels. This is a key step in the decoding of neuronal population dynamics for brain-machine-interfaces and biomedical applications

Project phases:

– Literature research
– Data analysis
– Interpretation and discusion of the results

This thesis foresees a development part at Scuola Superiore Sant’Anna in Pisa.