Public thesis defense Sylvain Favresse - ICTEAM
sst |
Application-Specific Design of Ultra-Low-Power Low-Noise Vagus Nerve Front-End Circuits for the Detection of Epileptic Seizures
Thursday September 10th, 2026 - 4pm - Auditorium BARB94 - Place Sainte-Barbe 1, 1348 Louvain-la-Neuve
Vagus nerve (VN) stimulation is a possible treatment for drug-resistant epilepsy. Despite its widespread use, it has a moderate efficacy that can be improved by delivering extra stimulation during epileptic seizures. To do so, seizure onsets must be detected in real time by the stimulator implant via biomarkers. One minimally invasive solution is the use as seizure markers of the VN electroneurogram (VENG) signals that feature an amplitude around 20 µV. In this context, the design of integrated front-end circuits for VENG signals acquisition faces three main technical challenges: (i) minimizing intrinsic noise in the circuits, (ii) limiting power consumption to 10 µW, and (iii) rejecting interference from nearby muscles.
This thesis tackles these issues by proposing new analysis methodologies and high-performance analog circuits. First, we worked at the level of the input low-noise amplifier by proposing a new semi-analytical methodology that estimates the noise and power performance without requiring any numerical optimization. Using this method, we designed and validated an on-chip innovative amplifier topology that rejects low-frequency interference. Second, we demonstrated a complete ultra-low-power analog front-end that includes amplifiers, filters, and an analog-to-digital converter. The proposed circuits additionally reject in-band interference thanks to a tripolar architecture with low gain mismatch. Finally, we conducted a system-level simulation-based analysis of the trade-off between front-end power consumption and accuracy of the seizure detection processing. Using real VENG recordings, we show that the seizure detection algorithm tolerates a noise level up to 3 µV, allowing the reduction of the front-end power consumption down to 2.1 µW.
Jury members
Prof. Denis Flandre (UCLouvain), Supervisor
Prof. David Bol (UCLouvain), Supervisor
Prof. Benoît Macq (UCLouvain), Chairperson
Dr. Rémi Dekimpe (UCLouvain), Secretary
Prof. Antoine Nonclercq (ULB)
Dr. Nick Van Helleputte (IMEC, Belgique)