Evaluation and Training Metrics for Medical Image Segmentation
thesisposted on 01.08.2020, 00:00 by Maddalena Andreoli Andreoni
This thesis will define and develop twenty metrics that are computed over tumor segmentation in medical MRI images and that aim to describe such segmentation in an objective and quantitative way. It will proceed to compute each metric over the Brain Tumor Image Segmentation Benchmark dataset to show their descriptiveness over real data. Subsequently it will attempt classification of tumor grade label using the defined metrics.