Neural networks for healing of burns
Supervisor TU Delft: Fred Vermolen
start of the project: February 2020
In June 2020 the
has been given.
The Master project has been finished in November 2020
by the completion of the
and a final
has been given.
For working address etc. we refer to our
Summary of the master project:
The MSc project deals with neural network applied to machine learning in the context of improving health care for burn injuries. Patients with large burn injuries are often faced with scars that can be hypertrophic or contracting. Hypertrophy poses the patient with aesthetic problems, whereas contractures represent wound contraction that is so large that it results into disability of the patient. In this case, the disability is caused by the scar on the skin, and not by the musculature or neural system of the patient. Mathematical models are used and investigated and aim at providing more efficient health care. Using these models, we want to simulate various scenarios that represent several methodologies of treatment and we aim at providing the practitioners with a tool to optimize treatment. In order to find the optimum, one has to perform many simulations on the basis of the wound scenario and patient-specific data. All these data contain uncertainties due to poor available of measurements and due to the genetic composition of the patient. This is a major reason why one has to perform many (Monte Carlo) simulations.
In order to provide the doctors with a fast tool so that simulation results for various scenarios are easy to obtain, we investigate the applicability of neural networks. The neural network will provide very speedy access to output parameters such as the likelihood that a contraction will exceed a certain severeness (threshold). During this project first various simple mathematical models will be investigated, which will be followed by more complicated models. During this work, one plays with the topology and number of nodes in the neural network.
Back to the
Master students page of Kees Vuik