← Back

MSc student vacancy on Brain tissue segmentation with deep learning


Brain tissue segmentation with deep learning

From T1 MR images, invariant to white matter lesions

Automated algorithms for the segmentation of gray matter (GM), white matter (WM) and cerebrospinal-spinal fluid (CSF) are performing sub-optimally in patients with white matter hyperintensities (WMH). WMH occur commonly in MRI scans of elderly people and subjects with neurodegenerative disease such as Alzheimer, Parkinson or multiple sclerosis. A method that has proven to enhance segmentation accuracy is so called ‘lesion filling’ in which white matter lesions are inpainted with normal appearing white matter (e.g. Chard 2010), before the actual segmentation algorithm is run. However, this solution is suboptimal and requires to have a WMH segmentation. The goal of the current project is to develop a deep neural network based algorithm that is capable to segment brain MRI’s from patients with WMHs, into GM, WM and CSF. We will use SienaX segmentation results of inpainted brain scans as gold standard and then train a deep neural network on this gold standard and the non-inpainted, original, MRI scans. A large collection of brain scans (both healthy controls and MS subjects) is available for this project. This should result in a fast method to segment GM, WM and CSF in original MRI scans from people with white matter hyperintensities, without requiring any pre-processing (such as brain-extraction, bias field correction, or inpainting). Tissuesegmentation.png

Requirements

  • Students with a background in computer science, biomedical engineering, artificial intelligence, physics, or a related area in the final stage of master level studies are invited to apply
  • Affinity with programming is required, interest and experience with machine learning and deep learning preferred.

Information

  • Project duration: 6-12 months
  • Supervision: Martijn Steenwijk (Amsterdam UMC), Bram Platel (Radboudumc & Amsterdam UMC) and Rashindra Manniesing
  • Location: Radboudumc, Nijmegen & VUmc, Amsterdam
  • For more information please contact Bram Platel