Previous research has suggested that for breast and other solid tumors, the interaction between the tumor and its microenvironment is connected to outcome and response to treatment. For example, in breast cancer, Beck et al. 3 has shown that the stromal compartment of the tumor contains more prognostic information than the epithelial component.
The additional information contained in the histopathology image features of the stromal compartment was also shown to contain independent prognostic information next to the known clinical, molecular and pathological risk factors. In this work package we will develop a digital Mammaprint that combines automated assessment of the tumor geometry in combination with clinical, pathological and molecular factors (mammaprint©) to optimize the risk assessment of disease recurrence and treatment response in patients with early stage breast cancer.
We will first train this on histopathology slides from patients treated in our hospital with a Mammaprint result and clinical outcome. Once we have these models, we will approach large randomized international clinical trials (Mindact n = 6,693) for which histopathology slides, clinical, pathological and molecular data are available to evaluate / validate our models. This is of utmost clinical importance to detect women who may be spared adjuvant chemotherapy.
