2026-04-14T02:04:16Zhttps://recercat.cat/oai/requestoai:recercat.cat:10230/546122025-12-19T20:29:51Zcom_2072_6col_2072_452954
RECERCAT
author
Romero Hernández, Joel
2022-10-26T15:33:43Z2022-10-26T15:33:43Z2022
Tutors: Óscar Cámara Rey, Ricard Solé VicenteTreball de fi de grau en BiomèdicaComplex diseases like cancer are one of the most serious sources of suffering and death.
Their multifactorial nature and multiscale effects compromise the patient’s health and
life, so there is a need for new tools for clinicians to navigate through the intricate search
space associated with their treatment. In this context, the present project focuses on
the creation of a computational platform to optimise the treatment of complex diseases
using AI and simulations. Specifically, the system is applied to study the combination
of chemotherapy, radiotherapy and immunotherapy in 4 real patients: 3 cases of non-
small cell lung cancer (2 adenocarcinomas and 1 squamous cell carcinoma), and 1 case of
small cell lung cancer. The prototype takes as input biomedical images directly from the
hospital equipment and converts them into a 3D tissue-labelled point cloud that approx-
imates the state of the disease at the beginning and the end of the treatment. Then, it
takes quasi-natural language instructions to generate customisable dynamic models of the
patient’s disease and treatment, combining methods like cellular automata and diffusion-
reaction. These models can be visualised and used to run controlled simulations with the
prototype’s graphic interface. Furthermore, the system can automatically parameterise
them to replicate the behaviour of the disease and treatment using a genetic algorithm.
Finally, the platform can also take instructions to generate customisable Deep Reinforce-
ment Learning agents that interact with the patient-specific simulations to search for
policies that improve their outcomes, so that this knowledge can be used to help future
patients.
©Tots els drets reservats info:eu-repo/semantics/openAccess
Complex diseases
Computational physiology,
Deep learning
Reinforcement learning in healthcare
Lung cancer
VProject: creating an AI and simulation system for the study of complex diseases and applying it to lung cancer
info:eu-repo/semantics/bachelorThesis