Background
Freezing of Gait (FoG), a debilitating motor symptom in Parkinson’s disease (PD), manifests as sudden and unpredictable difficulty initiating or maintaining forward movement. These episodes significantly impact daily activities, increase fall risk, and diminish the quality of life for people with Parkinson’s disease. The MAPP project addresses these challenges by investigating the underlying mechanisms of FoG through objective biomarkers derived from the integration of different neurophysiological and biomechanical information.
This specific thesis is part of a large-scale international collaboration between EPFL (Prof. Silvestro Micera, Prof. Friedhelm Hummel), CHUV (Prof. Solaiman Shokur), UFABC (Prof. Daniel Boari Coelho), and the AACD clinic in São Paulo. The goal is to collect comprehensive multimodal data encompassing electroencephalography (EEG), electromyography (EMG), and inertial measurement unit (IMU) data from patients to develop real-time prediction models.
Objectives
The primary objective of this project is two-fold: conducting extensive multimodal data acquisition on PD patients (with and without FoG) and developing a real-time FoG decoding pipeline. You will be uniquely positioned to test your algorithmic pipeline in real-time on the very patients you are recording. This immersive project requires spending a few days training in Switzerland to master the setup, followed by a relocation to São Paulo, Brazil, to work directly with the clinical team at AACD.
Tasks
- System Training (Switzerland): Spend the initial phase (~2 weeks) of your thesis at EPFL learning the multimodal acquisition framework, hardware synchronization, and experimental protocols.
- Clinical Data Acquisition (Brazil): Immerse yourself with the clinical team at the AACD clinic in São Paulo. You will actively participate in recording healthy controls, PD patients without FoG, and PD patients with FoG. This includes guiding patients through normal walking and specific FoG-inducing protocols (e.g., 180° turns, figure-eight trajectories, and narrow passages).
- Decoding Pipeline Development: Concurrently with data collection, develop and optimize a machine learning/deep learning pipeline to decode FoG events based on one or more of the acquired modalities (EEG, EMG, or IMU).
- Real-Time Clinical Testing: Integrate your decoding algorithm into the portable signal acquisition platform. Evaluate the decoding speed and accuracy of FoG episodes in a real-time scenario during patient recording sessions.
Required Skills
- Strong programming skills in Python (experience with real-time processing frameworks is a plus).
- Solid theoretical and practical background in Machine Learning and Deep Learning frameworks (e.g., PyTorch, TensorFlow).
- Experience or strong interest in biosignal processing (EEG, EMG, IMU) and clinical human subject research.
- Willingness to travel and relocate to São Paulo, Brazil for the core duration of the thesis.
- Fluent English; proficiency in Portuguese is a strong asset for interacting with patients, though the local team will be present to assist.
- Highly motivated, adaptable, and capable of working in a dynamic clinical environment.
Supervisors
- Swiss Side: Dr. Vincent Mendez, Dr. Daniel Leal (under the direction of Prof. Silvestro Micera, Prof. Friedhelm Hummel, Prof. Solaiman Shokur)
- Brazilian Side: Prof. Daniel Boari Coelho
How to Apply
(Following standard TNE guidelines)
Applications should be sent directly to the supervisors listed above. Your application should include:
- A brief motivation letter tailored to this specific project (generic letters will not be considered).
- A CV.
- An academic transcript.