An automatic detection method of cerebral aneurysms in time-of-flight magnetic resonance angiography images based on attention 3D U-Net

• Published in: Computer methods and programs in biomedicine Vol. 225; p. 106998
• Main Authors: Chen, Geng, Meng, Chen, Ruoyu, Di, Dongdong, Wang, Liqin, Yang, Wei, Xia, Yuxin, Li, Daoying, Geng
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2022
• Subjects: 3D U-Net; Cerebral aneurysm; Computer assisted detection; SENet; TOF-MRA; Cerebral aneurysm; CNN; ICA; CAD; CPU; MCA; FP; VA; GPU; PCA; TOF-MRA; FCN; DSA; 3D U-Net; MIP; SENet; Computer assisted detection; DICOM; RAM; ACA; BA
• ISSN: 0169-2607; 1872-7565
• DOI: 10.1016/j.cmpb.2022.106998

•Early detection of cerebral aneurysms can prevent cerebral hemorrhage.•Magnetic resonance angiography, the only non-invasive technique to detect aneurysms.•Deep learning can help doctors detect aneurysms faster and more accurately.•Existing methods have a high false positive rate and rely on large amounts of data.•The proposed method obtained high sensitivity with low false positive rate.•This article is the first to apply 3D U-Net related methods in aneurysm detection. Subarachnoid hemorrhage caused by ruptured cerebral aneurysm often leads to fatal consequences. However, if the aneurysm can be found and treated during asymptomatic periods, the probability of rupture can be greatly reduced. At present, time-of-flight magnetic resonance angiography is one of the most commonly used non-invasive screening techniques for cerebral aneurysm, and the application of deep learning technology in aneurysm detection can effectively improve the screening effect of aneurysm. Existing studies have found that three-dimensional features play an important role in aneurysm detection, but they require a large amount of training data and have problems such as a high number of FPs per case. This paper proposed a novel method for aneurysm detection. First, a fully automatic cerebral artery segmentation algorithm without training data was used to extract the volume of interest, and then the 3D U-Net was improved by the 3D SENet module to establish an aneurysm detection model. Eventually a set of fully automated, end-to-end aneurysm detection methods have been formed. A total of 231 magnetic resonance angiography image data were used in this study, among which 132 were training sets, 34 were internal test sets and 65 were external test sets. The presented method obtained 97.89±0.88% sensitivity in the five-fold cross-validation and obtained 90.8% sensitivity with 2.47 FPs/case in the detection of the external test sets. Compared with the results of our previous studies and other studies, the method in this paper achieves the best sensitivity while maintaining low number of FPs per case. This result proves the feasibility, superiority, and further improvement potential of the improved method combining 3D U-Net and channel attention in the task of aneurysm detection.