Machine Learning Models on Prognostic Outcome Prediction for Cancer Images with Multiple Modalities

Abstract

Machine learning algorithms have been applied to predict different prognostic outcomes for many different diseases by directly using medical images. However, the higher resolution in various types of medical imaging modalities and new imaging feature extraction framework brings new challenges for predicting prognostic outcomes. Compared to traditional radiology practice, which is only based on visual interpretation and simple quantitative measurements, medical imaging features can dig deeper within medical images and potentially provide further objective support for clinical decisions. In this dissertation, we cover three projects with applying or designing machine learning models on predicting prognostic outcomes using various types of medical images.