English
 
User Manual Privacy Policy Disclaimer Contact us
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT
  Detecting fat artifacts in diffusion MRI with deep learning

Podranski, K., Zhang, R., Thierbach, K., Thieleking, R., Witte, A. V., Villringer, A., et al. (2019). Detecting fat artifacts in diffusion MRI with deep learning. Poster presented at 9th IMPRS NeuroCom Summer School, Leipzig, Germany.

Item is

Basic

show hide
Item Permalink: http://hdl.handle.net/21.11116/0000-0004-C6C1-2 Version Permalink: http://hdl.handle.net/21.11116/0000-0004-C6C2-1
Genre: Poster

Files

show Files

Locators

show

Creators

show
hide
 Creators:
Podranski, Kornelius1, Author              
Zhang, Rui2, Author              
Thierbach, Konstantin1, Author              
Thieleking, Ronja2, Author              
Witte, A. Veronica2, Author              
Villringer, Arno2, Author              
Weiskopf, Nikolaus1, Author              
Scherf, Nico1, Author              
Affiliations:
1Department Neurophysics (Weiskopf), MPI for Human Cognitive and Brain Sciences, Max Planck Society, ou_2205649              
2Department Neurology, MPI for Human Cognitive and Brain Sciences, Max Planck Society, ou_634549              

Content

show
hide
Free keywords: -
 Abstract: Sometimes magnetic resonance imaging (MRI) data is distorted by severe artifacts, due to non optimal sequence settings. Analyses of those images are going to produce false results and will confound affected studies. For that reason it is important to detect and remove datasets containing artifacts. Manual quality control of all images in large cohorts is not feasible; automatic methods to detect relevant artifacts are needed. We trained a fully convolutional feed forward network to detect artifacts caused by imperfect fat suppression in MRI diffusion measurements (DWI) of the human head (128x128x72@1.7mm³/voxel, 60 diffusion directions, b=1000 s/mm²). The deep neural network was pre-trained for two epochs on an auto-encoding task using one million unlabeled slices from 2000 different DWI datasets. After pre-training the encoder was separated and extended with one additional layer. The prediction task was trained on 1200 labeled slices from 100 different DWI datasets with binary cross-entropy objective. The optimal training state was selected posterior based on the best accuracy. Inference on 3303 before unseen slices achieved an accuracy of 0.82 (true positives: 179, true negatives: 2537, false positives: 26, false negatives: 561). Exploration of misclassified images often showed positive response to very bright spots and negative response if the contrast between artifact and surrounding voxels was low. Considering, that only few labels were available the achieved results are very promising. Predicting several slices for each dataset and applying majority voting might be used to further reduce false predictions.

Details

show
hide
Language(s):
 Dates: 2019-06-18
 Publication Status: Not specified
 Pages: -
 Publishing info: -
 Table of Contents: -
 Rev. Type: -
 Identifiers: -
 Degree: -

Event

show
hide
Title: 9th IMPRS NeuroCom Summer School
Place of Event: Leipzig, Germany
Start-/End Date: 2019-06-16 - 2019-06-19

Legal Case

show

Project information

show

Source

show