Random Forests with Selected Features for Stroke Lesion Segmentation

Matthias Wilms, Heinz Handels, Oskar Maier


Ischemic stroke is the third leading cause of death in industrialized countries [8].
Due to its excellent soft tissue contrast, magnetic resonance imaging (MRI) has
become to be the modality of choice for clinical evaluation of ischemic stroke
lesions [4]. As ischemic stroke lesions usually change over time and secondary
and remote changes may occur, it is therefore necessary characterizing the tissue
changes with different acquisition parameters to produce images of the same
physical space in distinctive spectral signatures [1].
In clinical practice, Diffusion weighted images (DWI), T1-weighted (T1W),
T2-weighted (T2W) and fluid attenuated inversion recovery (FLAIR) images are
often acquired to monitor progression of strokes [7]. In acute stage, hyperintense
signal observed on DWI provides important information about the anatomical
location and extent of the infarcted territory. In more chronic phase, T2W and
FLAIR images are normally used to delineate the final lesion volume. Chronic ischemic
lesions appear as hyperintense regions in FLAIR with some heterogeneity
within the lesion volume due to ongoing gliosis and demyelination [6].
Early and accurate diagnosis of brain lesion by multi-spectral magnetic resonance
images is the key for implementing successful therapy and treatment
planning [5]. However, the diagnosis is a very challenging task and can only be
performed by professional neuro-radiologists. Lesion segmentation can improve
this situation and help radiologists diagnose and make treatment plan. However,
due to the variety of the possible shapes, locations, and intensity inhomogeneity,
accurate segmentation is still a challenging task [2]. Manual segmentation can beperformed by trained radiologists, but it is a tedious and time consuming task
and is non-reproducible [4].
In this paper, we propose a automatic ischemic stroke lesion segmentation
algorithm in multi-spectral images (DWI, T1-w, T2-w, and FLAIR) using bias
correction embedded FCM and three phase level set method. The rest of this
paper is organized as follows.
Original languageEnglish
Number of pages6
Publication statusPublished - 2015
Event18th International Conference on Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015
- Munich, Germany
Duration: 05.10.201509.10.2015


Conference18th International Conference on Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015
Internet address


Dive into the research topics of 'Random Forests with Selected Features for Stroke Lesion Segmentation'. Together they form a unique fingerprint.

Cite this