Integrative digital diagnostics for therapy optimization in early breast cancer

Purpose and aimsThis project aims to develop tools for prediction of response to neoadjuvant (pre-operative) therapy (NAT) and prognostication of post-surgery risk of recurrence in breast cancer.

To this end, input from radiology, digital pathology, genomics and informative clinical variables will be integrated using a machine learning (ML)-based multi-modal fusion strategy.

Project organisation, time plan and scientific methodsThree academic clinical trials and one population-based cohort of NAT (N=2500) will be used to train single-source predictive model priors that will be ensembled into integrative multi-omics predictive models.

These will be validated externally in independent cohorts of ~3000 patients.The project will be divided into work packages (WP), corresponding to each of the data modalities.

WP1 data and material collection (year 1-4); WP2-3 transcriptomics and genomics in tissue and blood (y 1-3); WP4 radiomics using mammography and magnetic resonance imaging (y 1-3); WP5 pathomics (y 1-3); WP6 model integration (y 3-4); WP7 external validation (y 4-5).

ImportanceThe project will contribute with novel ML methodology for clinical medicine and a precision oncology solution for optimizing NAT selection and risk stratification that will lead to less over- and under treatment, sparing patients from unnecessary toxicities and reducing financial burden to healthcare systems, and ultimately improving prognosis for patients with breast cancer.