Jonathan Arvidsson1, Fredrik Johansson1, Andrew Mehnert1, 2, Darryl McClymont3, Dominic Kennedy4
1Signals and Systems, Chalmers University of Technology, Gothenburg, Sweden; 2MedTech West, Gothenburg, Sweden; 3ITEE, The University of Queensland, Brisbane, Queensland, Australia; 4Queensland X-Ray, Greenslopes, Queensland, Australia
A novel method for automatically segmenting 3D lesions in dynamic contrast-enhanced breast MRI data is proposed. It is based on assigning a suspiciousness score to each voxel using features extracted from its time series, and then computing the spatial co-occurrence of this score in a 3D neighborhood about the voxel. In this way both the spatial and temporal variation in contrast enhancement are characterized. An empirical evaluation of the efficacy of this technique versus a competing method based on multispectral co-occurrence is also presented.