February 5th - STITCH Workshop
Computational Intelligence per la
Bioinformatica e la Diagnostica Clinica
EXPLAINABLE AI FOR COMPLEX SYSTEMS MODELLING
Explainable Artificial Intelligence (XAI) deals with the problem of designing machine learning systems yielding predictive models whose decision rules can be easily understood by field experts, possibly in natural language. The ability to synthetize white (or grey) models is a fundamental issue in both medical diagnosis problems and system biology applications, where there is a strong request to have interpretable decision making systems, on one hand, and the possibility of using machine learning techniques to perform knowledge discovery, on the other hand. Both medicine and biology deal with complex systems, and consequently it is of utmost importance to develop explainable modelling systems able to face complexity, finding regularities in data and to discover at the same time the best information granulation (semantic) level where these regularities take place. This talk introduces some promising approaches for complex systems modelling, where the interpretability requisite is an essential issue.
GRAPH-BASED PATTERN RECOGNITION PROBLEMS IN BIOINFORMATICS
Many biological systems are conveniently represented by networks of interacting entities, like protein contact networks, metabolic pathways, protein-protein interactions networks. In the last years, machine learning and pattern recognition emerged as breakthrough disciplines in order to model complex systems such as biological ones and in order to extract knowledge from data. This talk introduces some common strategies for dealing with graph-based pattern recognition and presents recent techniques able to model the structure-vs-function and the structure-vs-solubility relationships in protein contact networks and techniques in order to mine metabolic networks.
ACCELERATING MAGNETIC RESONANCE IMAGING WITH DEEP LEARNING
The acceleration of magnetic resonance imaging (MRI) analysis has become an attractive research topic in the last years due to its potential advantages, such as higher availability of MRI scanners and lower healthcare costs for patients. Recently, the acceleration of MRI has been addressed by using deep learning methods, which have shown significant capabilities in reconstructing undersampled MR images. This talk introduces the most popular deep learning methods of MR image reconstruction and presents some novel deep neural network models that are able to exploit multiple information from an MRI analysis, thus enhancing the reconstruction quality of the image with an acceleration factor for the MRI scanning up to 8 times. Specific examples will be shown in the context of multiple sclerosis MRI.
NATURAL LANGUAGE PROCESSING FOR HEALTHCARE AND MEDICAL DIAGNOSIS
Speaker: ENRICO DE SANTIS
The past decade has seen an explosion in the amount of digital information stored in electronic health records (EHRs). While primarily designed for archiving patient information and performing administrative healthcare tasks like billing, many researchers have found secondary use of these records for various clinical informatics applications. Over the same period, the Artificial Intelligence community has seen a widespread advance in many areas, such as Natural Language Processing, allowing to elaborate a huge amount of written text for a plethora of interesting applications. This talk will cover the main topics regarding the adopted techniques for text processing and how to construct a Machine Learning pipeline for analyzing and synthesizing models build upon EHR data.
mercoledì 5 febbraio 2020, ore 15.00 - 17.00 Aula A - “Livio Capocaccia” - Dipartimento di Medicina traslazionale e di precisione - III Clinica Medica in
viale del Policlinico 155, Roma
Per partecipare occorre registrarsi inviando una mail a firstname.lastname@example.org