International contest to create AI tools that help doctors to prognosticate cancer cases

24 August 2020

More than 150 experts in artificial intelligence and natural language processing are taking part in an international competition, organized to create automatic tools capable of tracing and classifying wherever cancer tumours are mentioned in huge volumes of clinical texts written in Spanish language.

The contest, which started in April and will end in September, gathers participants from 15 countries and has been organized by the Barcelona Supercomputing Center, within the frame of the Plan for the Boosting of Language Technologies (Plan TL) of the Spanish Secretariat for Digitalization and Artificial Intelligence.

The competition is part of the tasks carried out in the context of personalized medicine, in order to create technologies that will help doctors to execute a more accurate prognosis in cases of cancer, and also to select medical treatments.

Participants must create automatic tools capable of finding in clinical cases any mention to the form and characteristics of carcinogenic tumours (their histological and morphological information), and of labelling them according to the ICD-O classification (International Classification of Diseases for Oncology) of the World Health Organization.

These tools will be part of more complex artificial intelligence systems to be developed in the future, which will be able to extract as well any other information contained in clinical texts (such as treatments, patient’s progress, etc) and to perform predictions based on the analysis of massive amounts of medical writings.

The ability to extract information from clinical records in Spanish language is considered as a critical area for the development of systems of support to the decision taking in medicine, with a great industrial potential.

The CANTEMIST contest (Cancer Text Mining Shared Task) takes in at this moment more than 60 teams with 156 individual participants from 15 different countries, including 12 companies that develop their activity in this field. Those tools presented by participants will be evaluated by an international expert committee, and the result will be disclosed the last week of September, during the Iberlef workshop organized by the Spanish Association for the Processing of Natural Language (SEPLN).