Research Areas
Our lab operates at the intersection of neuroscience, engineering, and artificial intelligence. We focus on three core pillars to advance our understanding and treatment of the human brain.
Neural Computation for Brain Imaging
Developing advanced computational frameworks to integrate multimodal brain imaging data for high-resolution mapping.
- fMRI Source Imaging
- EEG/MEG Inverse Problems
- Real-time Decoding
FIGURE 1: ESI INTERPOLATING FMRI DATA FOR HIGH-RESOLUTION MAPPING
Our research develops novel machine learning frameworks to integrate the high temporal resolution of EEG with the high spatial resolution of fMRI. By solving the electromagnetic inverse problem with deep generative models, we can image brain activity with unprecedented spatiotemporal precision.
This technology serves as the foundation for our non-invasive Brain-Computer Interfaces (BCI), allowing us to decode user intent for controlling external devices or communicating directly from brain activity.
Focused Ultrasound Neuromodulation
Pioneering non-invasive, targeted modulation of deep brain circuits using acoustic energy for therapeutic applications.
- Transcranial Focused Ultrasound (tFUS)
- Closed-loop Control
- Plasticity Induction
FIGURE 2: CLOSED-LOOP FOCUSED ULTRASOUND STIMULATION SYSTEM
Focused ultrasound (FUS) offers a revolutionary way to modulate brain activity without surgery. Unlike TMS or tDCS, FUS can reach deep brain structures with millimeter precision.
We are developing "smart" neuromodulation systems that listen to the brain's activity via EEG and deliver precisely timed ultrasound pulses to induce plasticity or disrupt pathological rhythms. This closed-loop approach holds promise for treating conditions like epilepsy, depression, and chronic pain.
AI for Brain-Computer Interface
Leveraging artificial intelligence to decode user intent from brain activity for controlling external devices.
- Liquid Neural Networks
- Dynamical Systems Theory
- Biophysical Modeling
FIGURE 3: LIQUID NEURAL NETWORK FOR MOTOR CORTEX MAPPING
How does the brain compute? We use computational models, from biophysically realistic neuron models to abstract recurrent neural networks, to simulate brain function.
A key focus is on "Liquid Neural Networks" and other dynamic architectures that can process continuous time-series data efficiently. These models not only help us understand biological intelligence but also inspire new, more efficient AI algorithms.
Interested in our work?
Explore our publications to see the detailed results of these research lines.