Cerebral aneurysm image segmentation based on multi-modal convolutional neural network. (September 2021)

• Record Type: Journal Article
• Authors: Meng, Chengjie; Yang, Debiao; Chen, Dan
• Is Part Of: Computer methods and programs in biomedicine. Volume 208(2021)
• Publication Date: 2021-09
• Subjects: Cerebral aneurysm -- Convolutional neural network -- Computed tomography angiography -- Image segmentation -- Algorithm; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2021.106285
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: Cerebral aneurysm and adjacent vessels can be segmented by convolutional neural network efficiently. The segmentation ability were different in various combination of extraction methods. Abstract: Background and Objective: Accurate segmentation of cerebral aneurysms in computed tomography angiography (CTA) can provide an essential reference for diagnosis and treatment. This study aimed to evaluate a more helpful image segmentation method for cerebral aneurysms. Methods: Firstly, the original CTA images were filtered by Gaussian and Laplace, and both the processed image and original image constitute multi-modal images as input. Then, through multiple parallel convolution neural networks to multi-modal image segmentation. Eventually, all of the segmentation results were fused by linear regression to extract cerebral aneurysm and adjacent vessels. Results: The cerebral aneurysm and adjacent vessels were extracted correctly. When the threshold value is about 0.95, the overall performance of the segmentation effect is the best. The dice, accuracy, and recall rate were different in various combinations of the three extraction methods. Conclusion: Multi-modal convolutional neural network can improve the segmentation accuracy by multi-modal processing of the original brain CTA image.