PORTALE DELLA DIDATTICA

Ricerca CERCA
  KEYWORD

GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA

Machine and Deep Learning for Disease Mapping

keywords ARTIFICIAL INTELLIGENCE, DEEP LEARNING, HEALTH, HEALTHCARE, MACHINE LEARNING, NEURAL NETWORKS

Reference persons DANIELE JAHIER PAGLIARI

External reference persons Paola Berchialla (UNITO)

Research Groups DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA, ELECTRONIC DESIGN AUTOMATION - EDA, GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA

Thesis type RESEARCH

Description One of the common threads that unites the problems of the COVID-19 pandemic in the era of Big Data is timeliness. The aim of the TrustAlert project is to create an integrated platform that allows the analysis of real-time data streams from (i) structured health databases such as hospital discharge records, outpatient and emergency room visits, and (ii) unstructured data such as news and social media. The platform aims to provide early warnings and prediction tools to local healthcare services to anticipate medical needs in emergency settings, like pandemics.

The main objective of this thesis will be focused on the first part. The aim will be to map the health status and patterns of morbidity and vulnerability in a specific population using artificial intelligence techniques applied to administrative health data, collected from hospital discharge and medication prescriptions. This objective is essential to understand the current health needs of the population and provide an informational basis for resource allocation and priority setting in case of an emergency. Ideally, priority setting should adapt to specific local needs, which must also be met in case of an emergency.

More in detail, we will extract data from administrative databases (HADs) about diagnoses recorded during hospitalizations, medications prescriptions, and outpatient data to describe at a local level the vulnerability and the comorbidity status of a population. Then, we will extend a Natural Language Processing algorithm with a spatial component, to use textual codes extracted from HADs (such as diagnoses and medical prescriptions’ codes) to learn co-existing medical conditions. Furthermore, we will provide a pre-trained Deep Learning model on HADs which will be a tool that can be used for further research.

Required skills Machine and Deep Learning Algorithms, Python Programming. Basics of R programming are useful, although not required.

Notes Thesis in collaboration with the Department of Clinical and Biological Sciences, University of Torino (UNITO).


Deadline 08/03/2022      PROPONI LA TUA CANDIDATURA




© Politecnico di Torino
Corso Duca degli Abruzzi, 24 - 10129 Torino, ITALY
Contatti