Advanced search
Start date
Betweenand

Deep learning and intermediate representations for pediatric image analysis

Grant number: 20/06744-5
Support type:Scholarships in Brazil - Post-Doctorate
Effective date (Start): September 01, 2020
Effective date (End): August 31, 2022
Field of knowledge:Physical Sciences and Mathematics - Computer Science - Computer Systems
Principal Investigator:Roberto Marcondes Cesar Junior
Grantee:Hugo Neves de Oliveira
Home Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
Associated research grant:15/22308-2 - Intermediate representations in Computational Science for knowledge discovery, AP.TEM

Abstract

Medical imaging requires the development of methods to improve the accuracy of the results of image analysis. Advances in medical image analysis provide such tools, but there is still an important gap in relation to pediatric brain imaging, although there is an increasing medical demand. This project aims to contribute to fill this gap, focusing on brain Magnetic Resonance (MR) of babies, newborns and premature babies, which raise specific questions due to the particular contrast of gray/white matter related to the physiological myelination process, to the evolution very fast, but not continuously observed, of brain structures and possible pathologies, as well as high intra and inter-subject variability. One of these issues is that the data is typically noisy, ambiguous, scarce and sparse over time. In turn, specialized medical knowledge is available, but it is prone to change and evolution. From this point of view, the project addresses one of the cutting edge issues in data analysis, that is, how to extract and understand significant patterns where data is scarce, but specialized knowledge, continuously enriched, is available. We propose to develop structural representations of knowledge and image information in the form of graphs and hypergraphs, which will be explored to guide the understanding of space-time images (segmentation, recognition, quantification, comparison over time, description of the image content and evolution). Such techniques will be complemented by deep learning approaches for processing 2D or 3D images. The objective of the project is to develop computational methods to support the diagnosis, pathology analysis and monitoring of patients. (AU)