An international competition evaluates artificial intelligence tools to help in the diagnosis of cancer cases

22 October 2020

The CANTEMIST competition has been organized by BSC within the framework of the TL Plan by the SEDIA of the Ministry of Economic Affairs and Digital Transformation.

The CANTEMIST international competition (Cancer Text Mining Shared Task) has shown the capacity of the new generation of artificial intelligence (AI) tools to extract and catalog clinical information to help doctors carry out more precise diagnostics of cancer cases. The competition, which started in April and finished recently, has been organized by the Barcelona Supercomputing Center (BSC) within the framework of the plan to promote Language Technologies (plan TL, in Spanish) of the State Secretariat for Digitalization and Artificial Intelligence (SEDIA, in its Spanish acronym), of the Ministry of Economic Affairs and Digital Transformation.

During this time the participant teams have created automatic tools able to localize references in clinical cases to the form and the characteristics of tumors, and using IA techniques they have been cataloged according to the International Classification of Oncological illnesses (CIE-o in its Spanish acronym). These systems overcome the barrier of previous systems, since they can use written information of the clinical cases, and not only the one previously codified.

The results of this task and the systematic and harmonized access to information on the morphology of tumors are key to prioritizing therapeutic indications in cancer patients, facilitating the compliance of treatments according to clinical practice guidelines, as well as monitoring and evolution of cancer patients, including both common and rare tumors.

Martin Krallinger, coordinator of the Natural Language Processing group in BSC and main promoter of the competition, is very satisfied with the results and says that “for the first time we have been able to implement and compare systems that allow to accelerate and perform the digitalization process in key oncological aspects, as well as pathological anatomy of tumors. This allows to improve the knowledge, management and analysis of oncological patients’ information with the perspective of IA and data mining use within the context of a personalized treatment of cancer – precision medicine”.

The systems generated during this competition could facilitate the automatization of tumor biobanks, as well as clinical reports of tumor pathology. According to Alfonso Valencia, ICREA Professor, and director of the Life Sciences department of the BSC, “these type of competitions not only make specific technological fields advance rapidly with the incorporation of training and evaluation data, but they also represent a guarantee for the professionals and users about the proven quality of IA systems”.

25 teams from 16 countries have participated in CANTEMIST, including academic groups and companies such as Bosch Center for Artificial Intelligence, Siemens Healthineers and Vicomtech Foundation. The tools that have been shown have been evaluated by a committee of international experts, and the result of the competition has been presented during the Iberlef congress, organized by the Spanish Society for the Processing of Natural Language. The results of the competition and the description of the different systems can be checked at .


Caption: The members of the Vicomtech team, awarded first and second place in different categories of the Cantemist competition.