Machine learning approach to assess brain metastatic burden in preclinical models.

Abstract

Brain metastases (BrM) occur when malignant cells spread from a primary tumor located in other parts of the body to the brain. BrM is a deadly complication for cancer patients and severely lacks effective therapies. Due to the limited access to patient samples, preclinical models remain a very valuable tool for studying metastasis development, progression, and response to therapy. Thus, reliable methods to assess metastatic burden in these models are crucial. Here we describe step by step a new semi-automatic machine-learning approach to quantify metastatic burden on mouse whole-brain stereomicroscope images while preserving tissue integrity. This protocol uses the open-source and user-friendly image analysis software QuPath. The method is fast, reproducible, unbiased, and gives access to data points not always accessible with other existing strategies.