Dynamic features extraction method of resting-state BOLD-fMRI signal and its application to brain data classification between normal and glioma

Release time:2021-12-25  Hits:

• DOI number:10.1109/ICOSP.2014.7015176
• DOI number:10.1109/ICOSP.2014.7015176
• Journal:2014 12th International Conference on Signal Processing (ICSP)
• Journal:2014 12th International Conference on Signal Processing (ICSP)
• Place of Publication:UNITED STATES
• Place of Publication:UNITED STATES
• Key Words:Feature extraction, Correlation, Training, Algorithm design and analysis, Heuristic algorithms, Blood, Accuracy
• Key Words:Feature extraction, Correlation, Training, Algorithm design and analysis, Heuristic algorithms, Blood, Accuracy
• Abstract:The functional connectivity of brain is a key point of brain network analysis. The BOLD (blood oxygen level dependent) fMRI (functional magnetic resonance imaging) signal is an effective projection signal of brain function. A dynamic method in resting-state (RS) functional connectivity analysis of brains is proposed in this paper. In contrast to traditional static method, a sliding window is used to separate whole period RS-BOLD signal into variable segments in time domain to rebuild a dynamic set of RS-BOLD and enlarge the sample size. It will enable the utilization of neural network classifier or other machine learning algorithms to analyze features and patterns. By training module from features extracted from brain network of glioma patients and normal people, it states 100% accuracy in glioma diagnosis. Besides, this dynamic analysis method also extracts 124 feature connections of glioma brain network with 70% confidence coefficient. By comparison, we also exploit brain network using general graph-based static method. It fails to reveal significant alternations between glioma and normal.
• Abstract:The functional connectivity of brain is a key point of brain network analysis. The BOLD (blood oxygen level dependent) fMRI (functional magnetic resonance imaging) signal is an effective projection signal of brain function. A dynamic method in resting-state (RS) functional connectivity analysis of brains is proposed in this paper. In contrast to traditional static method, a sliding window is used to separate whole period RS-BOLD signal into variable segments in time domain to rebuild a dynamic set of RS-BOLD and enlarge the sample size. It will enable the utilization of neural network classifier or other machine learning algorithms to analyze features and patterns. By training module from features extracted from brain network of glioma patients and normal people, it states 100% accuracy in glioma diagnosis. Besides, this dynamic analysis method also extracts 124 feature connections of glioma brain network with 70% confidence coefficient. By comparison, we also exploit brain network using general graph-based static method. It fails to reveal significant alternations between glioma and normal.
• Co-author:Ziyi Wang, Weibei Dou Xue Wang, Min Li, Mingyu Zhang, Hongyan Chen, Shaowu Li, Jianping Dai,DOU Weibei
• Co-author:Ziyi Wang, Weibei Dou Xue Wang, Min Li, Mingyu Zhang, Hongyan Chen, Shaowu Li, Jianping Dai,DOU Weibei
• First Author:Wenbo Zhang,高宇
• First Author:Wenbo Zhang,高宇
• Indexed by:会议论文
• Indexed by:会议论文
• Correspondence Author:Weibei Dou,DOU Weibei
• Correspondence Author:Weibei Dou,DOU Weibei
• ISSN No.:2164-5221
• ISSN No.:2164-5221
• Translation or Not:no
• Translation or Not:no
• Date of Publication:2014-10-19
• Date of Publication:2014-10-19