Efficient Neural Network Approximation of Robust PCA for Automated Analysis of Calcium Imaging Data

Calcium imaging is an essential tool to study the activity of neuronal populations. However, high level background fluorescence in images hinders accurate identification neurons and extraction activities. While robust principal component analysis (RPCA) a promising method that can decompose foreground such images, its computational complexity memory requirement are prohibitively process large-scale calcium data. Here, we propose BEAR, simple bilinear neural network for efficient approximation RPCA which achieves order magnitude speed improvement with GPU acceleration compared conventional algorithms. In addition, show BEAR perform foreground-background separation data as large tens gigabytes. We also demonstrate two BEARs be cascaded simultaneous non-negative matrix factorization automated spatial temporal footprints from The source code used paper available at https://github.com/NICALab/BEAR.