Segmentation Validation Framework
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Many segmentation algorithms have been published in literature. Accurate assessment of these algorithms is needed in order to gain acceptance in the clinical practice or compare different algorithms. However in medical imaging segmentation field, there is usually a lack of gold standard method. In many studies, the manual segmentation from expert raters is regarded as the gold standard. Even though these manual denotation methods suffer from high inter-rater and intra-rater variability.
In this thesis, a relatively complete segmentation validation framework was developed, which uses the “Simultaneous Truth and Performance Level Estimation” (STAPLE) method to produce the ground truth from a set of manual expert segmentations, and performs a set of quantitative validation metrics measurement to assess the automated segmentation versus the ground truth given by STAPLE. It is also designed to be easy-to-use and cover most common image formats and popular operating systems. A number of tests using synthetic data have proved the accuracy of the proposed framework. In addition, some examples of how this framework can be used to evaluate the performance of different segmentation algorithm are also presented.
Place, publisher, year, edition, pages
2013. , 53 p.
Segmentation, Validation framework, Simultaneous Truth and Performance Level Estimation (STAPLE), Ground truth, ITK, Similarity metrics, Distance metrics
Medical Image Processing Medical Engineering
IdentifiersURN: urn:nbn:se:liu:diva-94231ISRN: LiTH-IMT/MASTER-EX--13/026--SEOAI: oai:DiVA.org:liu-94231DiVA: diva2:631210
Subject / course
Master's Program Biomedical Engineering
2013-06-12, Wrannesalen at CMIV, Linköpings universitet/US,SE-581 85, Linköping, 10:00 (English)
Wang, Chunliang, MD,PhD
Smedby, Örjan, Professor, Head of Division