Evaluating glioma growth predictions as a forward ranking problem

• DOI: 10.48550/arxiv.2103.11651
• Publication Date: 2021-01-01
• Authors: Karin A. van Gard...; Sebastian R. van ...; Maarten M.J. Wijn...; Fatih Incekara; Georgios Kapsas; Renske Gahrmann; Ahmad Alafandi; Marion Smits; Stefan Klein
• Keywords: Predictive power; Performance metric

The problem of tumor growth prediction is challenging, but promising results have been achieved with both model-driven and statistical methods. In this work, we present a framework for the evaluation predictions that focuses on spatial infiltration patterns, specifically evaluating future growth. We propose to frame as ranking rather than segmentation problem. Using average precision metric, can evaluate segmentations while using full spatiotemporal prediction. Furthermore, by separating model goodness-of-fit from predictive performance, show in some cases, better fit parameters does not guarantee power.