A machine learning ensemble approach for 5- and 10-year breast cancer invasive disease event classification

• DOI: 10.1371/journal.pone.0274691
• Publication Date: 2022-09-19T17:29:54Z
• Authors: Raffaella Massafra; Maria Colomba Comes; Samantha Bove; Vittorio Didonna; Sergio Diotaiuti; Francesco Giotta; Agnese Latorre; Daniele La Forgia; Annalisa Nardone; Domenico Pomarico; Cosmo Maurizio Ressa; Alessandro Rizzo; Pasquale Tamborra; Alfredo Zito; Vito Lorusso; Annarita Fanizzi
• Keywords: Adjuvant Therapy

Designing targeted treatments for breast cancer patients after primary tumor removal is necessary to prevent the occurrence of invasive disease events (IDEs), such as recurrence, metastasis, contralateral and second tumors, over time. However, due molecular heterogeneity this disease, predicting outcome efficacy adjuvant therapy challenging. A novel ensemble machine learning classification approach was developed address task producing prognostic predictions IDEs at both 5- 10-years. The method based on concept voting among multiple models give a final prediction each individual patient. Promising results were achieved cohort 529 patients, whose data, related cancer, provided by Istituto Tumori "Giovanni Paolo II" in Bari, Italy. Our proposal greatly improves performances returned baseline original model, i.e., without voting, finally reaching median AUC value 77.1% 76.3% IDE 5-and 10-years, respectively. Finally, proposed allows promote more intelligible decisions then greater acceptability clinical practice since it returns an explanation patient through procedure.