Han Zheng
Hong and Van Hooser Laboratories
Department of Computer Science and Biology
Brandeis University
The Integrated Framework of Calcium Imaging Analysis and Extraction
Poster Abstract
The Integrated Framework of Calcium Imaging Analysis and Extraction consists of three major components: drifting correction, denoising, and matrix factorization. Denoising is designed using Fourier analysis and proximal gradient-based optimal algorithm to perform soft bandwidth and pixel-wise intensity estimation for the temporal information. The spatial noise and background are filtered using morphology. The matrix factorization is simultaneously updated with the estimation of pixel-wise intensity estimation, with the aim to select easy-pick cells first by exploiting the framework of curriculum learning. We also offer a few fine-tuned denoising strategies including RNN-based cell/ noise detection to enhance to pixel-wise intensity estimation of the calcium imaging data.
Personal Statement
This summer, through the support of the M.R. Bauer Foundation Fellowship, I had the opportunity to work in an interdisciplinary pursuit aiming to solve a long-standing problem in cell imaging. The interdisciplinary pursuit goes not only in my research but also exerts many effects in my work environment as I had chance to talk with various people from different domains including math, biology, computer science, and physics. Many of the ideas that turned out to be successful in addressing problems of my research were in fact inspired by brainstorming across many different disciplines. For instance, the idea of performing denoising in the Fourier domain came from my discussion with people in physics. The frequent exchange of ideas across disciplines provided me a rich ground on which I could formulate my idea more precisely and thoroughly and thereby contribute better to my research goal. I also learned the proper way to formulate problems and ask questions from empirical observations, and from this, construct my framework. The program also offers me a great opportunity to train my programming skills.