Clique identification and propagation for multimodal brain tumor image segmentation

Sidong Liu*, Yang Song, Fan Zhang, Dagan Feng, Michael Fulham, Weidong Cai

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contributionpeer-review

1 Citation (Scopus)


Brain tumors vary considerably in size, morphology, and location across patients, thus pose great challenge in automated brain tumor segmentation methods. Inspired by the concept of clique in graph theory, we present a clique-based method for multimodal brain tumor segmentation that considers a brain tumor image as a graph and automatically segment it into different sub-structures based on the clique homogeneity. Our proposed method has three steps, neighborhood construction, clique identification, and clique propagation. We constructed the neighborhood of each pixel based on its similarities to the surrounding pixels, and then extracted all cliques with a certain size k to evaluate the correlations among different pixels. The connections among all cliques were represented as a transition matrix, and a clique propagation method was developed to group the cliques into different regions. This method is also designed to accommodate multimodal features, as multimodal neuroimaging data is widely used in mapping the tumor-induced changes in the brain. To evaluate this method, we conduct the segmentation experiments on the publicly available Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) dataset. The qualitative and quantitative results demonstrate that our proposed clique-based method achieved better performance compared to the conventional pixel-based methods.

Original languageEnglish
Title of host publicationBrain informatics and health
Subtitle of host publicationInternational Conference, BIH 2016, Proceedings
EditorsGiorgio A. Ascoli, Michael Hawrylycz, Hesham Ali, Deepak Khazanchi, Yong Shi
PublisherSpringer-VDI-Verlag GmbH & Co. KG
Number of pages10
ISBN (Electronic)9783319471037
ISBN (Print)9783319471020
Publication statusPublished - 1 Jan 2016
Externally publishedYes
EventInternational Conference on Brain Informatics and Health, BIH 2016 - Omaha, United States
Duration: 13 Oct 201616 Oct 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9919 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


ConferenceInternational Conference on Brain Informatics and Health, BIH 2016
CountryUnited States


Dive into the research topics of 'Clique identification and propagation for multimodal brain tumor image segmentation'. Together they form a unique fingerprint.

Cite this