The researchers used an approach called deep-learning to train the software to learn visual patterns associated with patient survival using microscopic images of brain tumor tissue samples.