Brain tumor segmentation and characterisation In medical collaboration with: Klinikum rechts der Isar der TU München, Neuroradiologie: Prof. Dr. med. Claus Zimmer and Priv.-Doz. Dr. med. Jan Stefan Bauer Scientific Director: Prof. Dr. Bjoern Menze Contact Person(s): Esther Alberts

Keywords: IBBM: image computing

Abstract

A generative-discriminative framework is built for multi-modal brain tumor segmentation with the use of longitudinal growth models. The framework will be validated and compared with the state-of-the-art. In a later stage, image features based on tumor-specific bio-markers are explored, and, more specifically, their contribution to more accurate brain tumor segmentation is studied.

Detailed Project Description

Data acquisition and pre-processing
We use MR images of four modalities: T1, T1c, T2 and FLAIR. Images are pre-processed by means of ITK MR biasfield correction and NiftiReg non-rigid registrations.

Image segmentation techniques
We use a generative tumor segmentation algorithm based on Expectation-Maximisation, which calculates spatial tumor probability maps for every modality. These tumor probability maps are subsequently spatially and temporally regularized using a 4-dimensional Markov random field solved by graph cut.

Patient-specific tumor growth estimations
A non-parametric growth model is build on top of the graph cut segmentations and estimates tumor volumes along time.

Pictures

Figure 1: Temporal series of a growing brain tumor segmented out of a temporal FLAIR series with scan intervals of about one month.

Figure 2: Outline of an algorithm segmenting temporal brain MR scans, calculating tumor volumes and estimating tumor growth or shrinkage in between all time points.

Figure 3: Outline of an algorithm segmenting temporal brain MR scans and calculating tumor volumes with an additional quantification of volumetric uncertainties based on Gumbel perturbations.

Clinical Relevance

Oncology, Neuroradiology

Team

Contact Person(s)

M.Sc. Esther Alberts

Working Group

Prof. Bjoern Menze

M.Sc. Esther Alberts

Location

Technische Universität München Institut für Informatik / I16 Boltzmannstr. 3 85748 Garching bei München Tel.: +49 89 289-17058 Fax: +49 89 289-17059

Visit our lab at Garching.

Klinikum rechts der Isar der Technischen Universitüt München Ismaninger Str. 22 81675 München IFL Lab - Room: 01.3a-c Tel.: +49 89 4140-6457 Fax: +49 89 4140-6458

Visit our lab at Klinikum rechts der Isar.



internal project page

Please contact Esther Alberts for available student projects within this research project.