Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Imputation Strategy for Reliable Regional MRI Morphological Measurements

Imputation Strategy for Reliable Regional MRI Morphological Measurements Regional morphological analysis represents a crucial step in most neuroimaging studies. Results from brain segmentation techniques are intrinsically prone to certain degrees of variability, mainly as results of suboptimal segmentation. To reduce this inherent variability, the errors are often identified through visual inspection and then corrected (semi)manually. Identification and correction of incorrect segmentation could be very expensive for large-scale studies. While identification of the incorrect results can be done relatively fast even with manual inspection, the correction step is extremely time-consuming, as it requires training staff to perform laborious manual corrections. Here we frame the correction phase of this problem as a missing data problem. Instead of manually adjusting the segmentation outputs, our computational approach aims to derive accurate morphological measures by machine learning imputation. Data imputation techniques may be used to replace missing or incorrect region average values with carefully chosen imputed values, all of which are computed based on other available multivariate information. We examined our approach of correcting segmentation outputs on a cohort of 970 subjects, which were undergone an extensive, time-consuming, manual post-segmentation correction. A random forest imputation technique recovered the gold standard results with a significant accuracy (r = 0.93, p < 0.0001; when 30% of the segmentations were considered incorrect in a non-random fashion). The random forest technique proved to be most effective for big data studies (N > 250). http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Neuroinformatics Springer Journals

Imputation Strategy for Reliable Regional MRI Morphological Measurements

Download PDF
12 pages

Loading next page...
 
/lp/springer-journals/imputation-strategy-for-reliable-regional-mri-morphological-TCy0HxmfQo

References (62)

Publisher
Springer Journals
Copyright
Copyright © Springer Science+Business Media, LLC, part of Springer Nature 2019
Subject
Biomedicine; Neurosciences; Bioinformatics; Computational Biology/Bioinformatics; Computer Appl. in Life Sciences; Neurology
ISSN
1539-2791
eISSN
1559-0089
DOI
10.1007/s12021-019-09426-x
Publisher site
See Article on Publisher Site

Abstract

Regional morphological analysis represents a crucial step in most neuroimaging studies. Results from brain segmentation techniques are intrinsically prone to certain degrees of variability, mainly as results of suboptimal segmentation. To reduce this inherent variability, the errors are often identified through visual inspection and then corrected (semi)manually. Identification and correction of incorrect segmentation could be very expensive for large-scale studies. While identification of the incorrect results can be done relatively fast even with manual inspection, the correction step is extremely time-consuming, as it requires training staff to perform laborious manual corrections. Here we frame the correction phase of this problem as a missing data problem. Instead of manually adjusting the segmentation outputs, our computational approach aims to derive accurate morphological measures by machine learning imputation. Data imputation techniques may be used to replace missing or incorrect region average values with carefully chosen imputed values, all of which are computed based on other available multivariate information. We examined our approach of correcting segmentation outputs on a cohort of 970 subjects, which were undergone an extensive, time-consuming, manual post-segmentation correction. A random forest imputation technique recovered the gold standard results with a significant accuracy (r = 0.93, p < 0.0001; when 30% of the segmentations were considered incorrect in a non-random fashion). The random forest technique proved to be most effective for big data studies (N > 250).

Journal

Neuroinformatics – Springer Journals

Published: Jan 4, 2020

There are no references for this article.