DTU Studieprojekt - Bayesian Optimization with Preference-Order Constraints for Personalized Stimulus Paradigm

Bayesian Optimization with Preference-Order Constraints for Personalized Stimulus Paradigm

Udbyder
Vejleder
Sted
København og omegn
PurposeDevelop a prototype for live optimization of brain stimulation using Bayesian Optimization and EEG in the OptoCeutics BRAINLIGHT framework.
DescriptionGamma Stimulation using Sensory Stimulation (GENUS) has shown potential as a treatment option for neurodegenerative diseases. OptoCeutics has developed a class I medical device approved under the medical device regulation (MDR 2017), called EVY LIGHT. EVY LIGHT uses light-based 40 Hz GENUS with patented invisible spectral flicker technology. Studies have shown a high degree of individual variability in the peak gamma frequency, and our hypothesis is that a fixed 40 Hz stimulation frequency is not optimal across all users. Thus, we wish to develop an individually optimised stimulation paradigm, with personal settings for stimulation frequency and colour combinations. We propose to achieve this using a closed-loop system with stimulation during recording of scalp potentials with electroencephalography (EEG) where the stimulus response is measured from the EEG in a semi real-time manner. In an iterative manner with 1-minute stimulation sessions, the response landscape is explored by evaluating the response and updating the stimulus settings. The response will be quantified using an extensive set of features derived from the time-, frequency- & time-frequency domains, information theory, and graph theory. As a previous study has shown a clear correlation in response (evaluated by signal-to-noise ratio at the peak frequency) and discomfort, we impose a constraint on the discomfort. The best personal setting is then attempted obtained using Bayesian optimization with preference order constraints, in which the discomfort constraint is preferred. The project will be run as a prototype biomedical product development effort in which the students are encouraged to follow the MDR 2017 guidelines [https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32017R0745] I samarbejde med
OptoCeutics ApS

Forudsætninger
Interests in machine learning and biomedical product development

Emneord

• Bioteknologi og biokemi
• Medicin og medicoteknik
• Billedanalyse
• Fysiologisk modellering
• Høreapparater
• Hospitalsindretning
• Kunstige organer
• Medicinske apparater og systemer
• Nanomedicin
• Radioaktive lægemidler og isotoper
• Sundhed og sygdomme

• machinelearning
• Neurostimulation
• optimization
• ProductDevelopment

Navn
Kristoffer Hougaard Madsen

Stilling
Lektor

Mail
khma@dtu.dk

Vejleder-info
Kandidatuddannelsen i Medicin og Teknologi
Vejleder
Kristoffer Hougaard Madsen

ECTS-point
30 - 35

Type
Kandidatspeciale

Skal have taget
22435

Skriv i din ansøgning, at du fandt jobbet på ofir.dk