Computational model for phantom thermal perception 

Brief

Study how thermal perception adapts to dynamic temperature changes and model phantom thermal sensations using computational approaches.

Description

Thermal perception is strongly context-dependent: the same physical stimulus can be perceived as warm or cool depending on the prior thermal state of the skin. This reflects adaptive mechanisms in thermosensory processing.

At the TNE Lab, we developed a wearable thermal feedback device that enables individuals with upper-limb amputation to perceive temperature referred to their missing limb — a phenomenon we termed phantom thermal sensations. In previous experiments, temperature was continuously modulated while participants reported the intensity and quality of their phantom percepts.

This project aims to

1. replicate the dynamic thermal modulation protocol in a group of non-amputated participants,
2. develop a computational model of thermal perception dynamics,
3. compare perceptual responses and model parameters between amputated and non-amputated participants.

The modeling component will develop a hierarchical Bayesian framework that learns subject-specific dynamical models (e.g., ARX, state-space models, or Gaussian processes) describing the relationship between thermal stimuli and perception, while sharing information across participants. Inference will be performed using Hamiltonian Monte Carlo in a probabilistic programming environment such as PyMC.

The project will help clarify whether phantom thermal sensations follow canonical principles of thermosensory adaptation and provide insights into optimizing thermal encoding strategies for future neuroprosthetic systems.

Required skills

Programming in Python, basic signal processing or modeling, and an interest in computational neuroscience or sensory perception.

Type of Project

Semester project or Master project

Organization

40% experimentation, 60% modeling

Expected start and end date

Flexible

Contact person