AUTHOR=Terasaki Yuki , Yokota Hajime , Tashiro Kohei , Maejima Takuma , Takeuchi Takashi , Kurosawa Ryuna , Yamauchi Shoma , Takada Akiyo , Mukai Hiroki , Ohira Kenji , Ota Joji , Horikoshi Takuro , Mori Yasukuni , Uno Takashi , Suyari Hiroki TITLE=Multidimensional Deep Learning Reduces False-Positives in the Automated Detection of Cerebral Aneurysms on Time-Of-Flight Magnetic Resonance Angiography: A Multi-Center Study JOURNAL=Frontiers in Neurology VOLUME=Volume 12 - 2021 YEAR=2022 URL=https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2021.742126 DOI=10.3389/fneur.2021.742126 ISSN=1664-2295 ABSTRACT=Current deep learning-based cerebral aneurysm detection demonstrates high sensitivity, but produces numerous false-positives, which hampers clinical application of automated detection systems for time-of-flight magnetic resonance angiography. To reduce the false-positive (FP) rate while maintaining high sensitivity, we developed a multimodal convolutional neural network (CNN) designed to unite planar and stereoscopic information about aneurysms. This retrospective study enrolled TOF-MRA images of cerebral aneurysms from three institutions from June 2006 to April 2019. In the internal test, 80% of the entire data set was used for model training and 20% for the test, while for the external tests, data from different pairs of the three institutions were used for training and testing. Three deep learning models (planar information-only [2D-CNN], stereoscopic information-only [3D-CNN], and multimodal information [MM-CNN] were trained to classify whether the voxels contained aneurysms, and they were evaluated on each test. The performance of each model was assessed using free-response operating characteristic curves. In total, 732 aneurysms (5.9. ± 2.5 mm) of 559 cases (327, 120, and 112 cases from institutes A, B, and C; 469 and 263 cases) were included in this study. In the internal test, the highest sensitivities were 80.4%, 87.4%, and 82.5%, and the FP rates were 6.1, 7.1, and 5.0 FPs/case at a fixed sensitivity of 80% for the 2D-CNN, 3D-CNN, and MM-CNN, respectively. In the external test, the highest sensitivities were 82.1%, 86.5%, and 89.1%, and 5.9, 7.4, and 4.2 FPs/cases for them, respectively. MM-CNN was a new approach to maintain sensitivity and reduce the FP rate simultaneously.