• Researchers develop a mathematical method to predict the behaviour of leukemia with hospital data

    March 14, 2016

    This new method, designed by a team joined by the University, allows experts to predict which patients will need chemotherapy or develop autoimmune diseases.

    Algoritmo de imputación de las variables no muestreadas.

    A multidisciplinary group of researchers, led by professors of the University of Oviedo, have developed a model consisting of mathematical algorithms that can predict how chronic lymphocytic leukemia behaves. This new methodology may imply a great support for the clinical staff, as it foresees which patients will need chemotherapy or will have autoimmune diseases.

    This study, which has just been published in the Journal of Biomedical Informatics, is the result of the joint work of the researchers of the group of Inverse Problems of the Department of Mathematics and the Center for Artificial Intelligence of the University of Oviedo, the Oncology University Institute of the Principality of Asturias (IUOPA), and the Department of hematology of the Central University Hospital of Asturias (HUCA). The work is part of the PhD Thesis written by Enrique de Andrés Galiana, entitled Design of biomedical robots and their application in translational medicine.

    The authors of the work, published in ‘Journal of Biomedical Informatics', emphasize that the predictions are based on algorithms that are cheap, easy and accessible for any medical center.

    Juan Luis Fernández Martínez, professor of the Department of Mathematics of the University of Oviedo and doctor Ana Pilar González-Rodríguez of the Central University Hospital of Asturias, both authors of the paper, chronic lymphocytic leukemia as a disease with a great clinical variability. Ana Pilar González-Rodríguez points out that "one of the greatest challenges of this pathology is the early diagnosis of its evolution so that the patients may access an early and more intense treatment. That is the origin of our collaboration with these experts in modeling".

    Researchers analyzed a data base with different clinical variables of 265 patients of the Cabueñes Hospital in Gijón, to outline this new mathematical model. The resulting algorithms were able to predict the development of autoimmune diseases with an accuracy of up to 90% and the need of chemotherapy up to 80%. The decrease in accuracy of the need of chemotherapy, despite being rather high, is due to the heterogeneity involved when deciding whether to give chemotherapy treatments or not, which does not always depend on biological criteria. The methodology that the research group used included the risk assessment through curves (Receiver Operating Characteristic), which led to the balance between false positive and false negative results. This tool, and was later included in different fields of medicine and psychology.  

    The study conducted with patients of Cabueñes Hospital revealed the relevance of different prognostic variables related to the characteristics of platelets, reticulocytes (immature red blood cells) and NK cells (Natural Killer), which are the main targets in the development of autoimmune diseases. Moreover, the study also revealed other factors that are not usually taken into account when deciding whether to use chemotherapy or not.

    Professor Fernández Martínez highlights that one of the strongest points of this methodology, due to being so simple, was outline in an Excel spreadsheet, which was published along with this paper. "The data on which predictions are based are simple, cheap and accessible for any hospital. Similar analyses could be performed with other data and other pathologies, as we have done in the past with Hodgkin´s lymphoma. The hospital big-data will become a mine of knowledge ", confirms the researcher.


    • Analysis of clinical prognostic variables for Chronic Lymphocytic Leukemia decision-making problems. Journal of Biomedical Informatics. Enrique J. de Andrés Galiana, Juan L. Fernández-Martínez , Óscar Luaces, Juan J. del Coz, Leticia Huergo-Zapico, Andrea Acebes-Huerta, Segundo González, Ana P. González-Rodríguez.