Machine Learning & Statistics in Radiology: Benchmark Experiments and Their Statistical Evaluation for Radiomics
MCML Authors
Theresa Stüber
Dr.
Abstract
Theresa Stüber
Dr.
Abstract
This dissertation investigates the reliability and reproducibility of machine learning models in radiomics, focusing on applications in cancer diagnosis and prognosis. It introduces robust statistical evaluation methods to account for performance variability and dependencies in benchmark studies. Through applications in prostate, colorectal, and lung cancer, it highlights the challenges of model validation and replicability, emphasizing the need for rigorous evaluation before clinical deployment. (Shortened.)
BibTeXKey: Stu26a