Inteligencia artificial aplicada a la radiómica en patología pulmonar

Abstract

This thesis presents the results of two studies in which diagnostic models based on radiomics, a novel quantitative and high-throughput medical imaging analysis technique powered by artificial intelligence, mainly Deep Learning, are developed to optimize imaging diagnosis of pulmonary embolism (PE) and pulmonary nodules. The first study demonstrates an intelligent radiomic model capable of making a localized diagnosis of different imaging patterns (PE, pneumonia, and healthy lung) in patients with and without COVID-19 through SPECT-CT/Q. The second study presents an integrative model that merges a Deep radiomic model with a clinical data model to predict malignancy in incidental or screening-detected pulmonary nodules. Both experimental models have shown good diagnostic performance in the group of patients analyzed, being able to address the proposed issues. The use of new technologies, such as those presented in this thesis, brings us closer to advanced imaging diagnosis that can potentially be applied to daily clinical practice for the benefit of patients and medical practice.

Date of Award	11 Oct 2024
Original language	Spanish
Supervisor	Debora Gil Resina (Director) & Antoni Rosell Gratacos (Director)