Fighting infant mortality thanks to artificial intelligence with tools capable of assisting doctors and promptly identifying risk factors in preterm infants. This is the reason the Italian team created PISA (Preterm Infants Survival Assessment), an application to evaluate the survival of preterm infants, now freely available to the international scientific and medical community.
The study which led to the creation of PISA has been published in ‘Scientific Reports’, a journal from the Nature group, and was carried out by researchers from the Department of Computer Science at the University of Pisa coordinated by Professor Alessio Micheli with Davide Bacciu, PhD and the team of neonatologists led by Dr. Luigi Gagliardi from Versilia Hospital (north-west Tuscany health authority).
In order to understand what PISA is and how it works, it is possible to access the website http://pisascore.itc.unipi.it/single-sample-mode/, fill in the required data – for example birth weight, sex or birth mode – and click to receive an answer. However, this simplicity of use belies sophisticated technology based on ‘Machine Learning’, automatic learning, or rather the idea that computers can learn to carry out specific tasks without being programmed to do so, thanks to how they use the data in their possession.
“In order to create PISA,” explains Alessio Micheli, “we took into consideration the data, obviously anonymous, of over 29,000 preterm Italian infants and we used this information to create ‘Machine Learning’ models to obtain a more accurate prediction algorithm than those currently in use at international level, which are, instead, based on classical statistical models.”
Professor Alessio Micheli and Davide Bacciu, Phd
The study which led to the creation of PISA was, therefore, the first on a worldwide basis to gather such an enormous quantity of data; in particular, the researchers used information from the Italian Neonatal Network, a project which includes 89 hospitals all over Italy with Versilia Hospital as one of the coordinating centres.
“Each year around 4,500 infants are born very preterm, before the 30th week of gestation or weighing less than 1,500g, and while they represent less than 1% of births, they account for more than half the rate of infant mortality in Italy and the developed countries,” emphasizes Luigi Gagliardi. “ PISA, therefore, represents an important tool both in the care of individual patients, and to increase understanding over the causes of mortality, in order to identify more effective therapies, and ultimately to improve the prognosis of this fragile population.”
The creation of PISA, also funded by the University of Pisa thanks to PRA (the University of Pisa Research Project) ‘Metodologie informatiche avanzate per l’analisi di dati biomedici’, is part of the research activity carried out by the Computational Intelligence & Machine Learning group, CIML-Unipi, and includes PhD student Marco Podda who is the co-author of the work.