LINKS: Learning-Based Multi-source IntegratioN FrameworK for Segmentation of Infant Brain Images

Abstract

Segmentation of infant brain MR images is challenging due to insufficient image quality, severe partial volume effect, and ongoing maturation and myelination processes. In the first year of life, the image contrast between white and gray matters of the infant brain undergoes dramatic changes. In particular, the image contrast is inverted around 6-8 months of age, and the white and gray matter tissues are isointense in both T1-and T2-weighted MR images and thus exhibit the extremely low tissue contrast, which poses significant challenges for automated segmentation. Most previous studies used multi-atlas label fusion strategy, which has the limitation of equally treating the different available image modalities and is often computationally expensive. To cope with these limitations, in this paper, we propose a novel learning-based multi-source integration framework for segmentation of infant brain images. Specifically, we employ the random forest technique to effectively integrate features from multi-source images together for tissue segmentation. Here, the multi-source images include initially only the multi-modality (T1, T2 and FA) images and later also the iteratively estimated and refined tissue probability maps of gray matter, white matter, and cerebrospinal fluid. Experimental results on 119 infants show that the proposed method achieves better performance than other state-of-the-art automated segmentation methods. Further validation was performed on the MICCAI grand challenge and the proposed method was ranked top among all competing methods. Moreover, to alleviate the possible Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. anatomical errors, our method can also be combined with an anatomically-constrained multi-atlas labeling approach for further improving the segmentation accuracy. NIH Public Access

Automated segmentation of dental CBCT image with prior-guided sequential random forests

“…It should be noted that similar work has been presented in Ref. 26, in which a sequence of random forest classifiers was used for infant brain segmentation. However, the spatial prior, which is important for segmentation, 23,24 was ignored in Ref.…”

“…Recently, random forest [22][23][24][25][26][27] has attracted rapidly growing interest. It is a nonparametric method that builds an ensemble model of decision trees from random subsets of features and the bagged samples of the training data and has achieved state-of-the-art performance in many imageprocessing problems.…”

“…However, the spatial prior, which is important for segmentation, 23,24 was ignored in Ref. 26. In this paper, we propose to integrate spatial prior into classification-based segmentation, which shows much better performance than using image appearances alone.…”

The developing human connectome project: A minimal processing pipeline for neonatal cortical surface reconstruction

“…This includes the development of specialised tools for tissue segmentation that address the difficulties in resolving tissue boundaries blurred through the presence of low resolution and partial volume. A variety of techniques have been proposed for tissue segmentation of the neonatal brain in recent years: unsupervised techniques (Gui et al, 2012), atlas fusion techniques (Weisenfeld and Warfield, 2009;Gousias et al, 2013;Kim et al, 2016), parametric techniques (Prastawa et al, 2005;Song et al, 2007;Xue et al, 2007;Shi et al, 2010;Cardoso et al, 2013;Makropoulos et al, 2012;Wang et al, 2012;Wu and Avants, 2012;Beare et al, 2016;Liu et al, 2016), classification techniques (Anbeek et al, 2008;Srhoj-Egekher et al, 2012;Chiţȃ et al, 2013;Wang et al, 2015;Sanroma et al, 2016;Moeskops et al, 2016) and deformable models (Wang et al, 2011;Dai et al, 2013;Wang et al, 2013Wang et al, , 2014. A review of neonatal segmentation methods can be found in Devi et al (2015); Makropoulos et al (2017).…”

Nonlocal atlas‐guided multi‐channel forest learning for human brain labeling

“…[1][2][3][4][5] Thus, automatic labeling of brain MR images becomes a notable topic in the field of medical image analysis. Due to the burden of manual brain labeling, it is imperative to develop an automatic and reliable brain labeling method.…”