Are Transformers More Robust Than CNNs? abs: arxiv.org/abs/2111.05464

In our (newly accepted! 🥳) #NeurIPS2022 paper, we develop a multi-scale and locality-sensitive self-supervised representation learning recipe for image-to-image architectures The trick? Regularization and denoising. The result? Better neuroimage segmentation w/ lower bias (1/5)

Michel Thiebaut de S and I wrote a #review on the last decade of #brain #connectivity #research for a special issue in Science Magazine : 'circuits create networks by stringing together many brain regions to orchestrate brain symphony conducted by finely attuned connections.' A 🧵

Early Access Paper Just Posted: Concurrent ischemic lesion age estimation and segmentation of CT brain using a Transformer-based network. Read the paper: ieeexplore.ieee.org/document/10155…

白学姐好棒！！

祝贺中国青年学者主办了一场鼓舞人心的神经影像人工智能研讨会并享用了丰盛的晚宴！ (Congratulations to the Chinese Young Scholars for hosting an inspiring Neuroimaging #AI Symposium and a great dinner!) #OHBM2023 Chao-Gan Yan Li Shen Qiu Anqi GeRuiyang-蓋瑞洋 Yu-Ping Wang

The 2023 #FutureSciencePrize Laureate Lecture was successfully held yesterday. #CUHK alumnus Dr Kaiming He (2011 PhD of Information Engineering) delivered a lecture and shared insights on "Deep Residual Learning and How It Shaped the #AI Landscape" with over 160 participants.

🧠 Neuroimaging has transformed our grasp of brain function! 📢 Ultrasound localization microscopy (ULM) is taking brain imaging to a whole new level. It can capture cerebrovascular flow down to micrometer scales, providing incredible insights. 🌟 #openbci #neuro #research

Educational Day for #OHBM2024 🥳🥳

HL-HGAT [1] explained by Jinghuan Huang (Nat Univ Singapore [2]): it uses Hodge-Laplacian filters for graph CNN processing of graphs arising in neuroimaging. The multisimplicial interaction layer will integrate edge graphs and node graphs. After each pooling layer the # of edges