Single particle tracking (SPT) has proven to be a powerful technique in studying the dynamics in complicated systems such as live cells and tissue. Prominent examples include the resolving of individual steps for molecular motors and the studying of molecular diffusion on cell membrane. Recently, single particle tracking has been extended to the three dimensional (3D) space by using non-conventional, z-sensitive point spread functions (PSFs). For example, by using astigmatism-based imaging, the 3D spatial information of the probe can be recovered in one camera exposure. However, the use of special PSFs results in complicated image patterns. In addition, the noise and the background interfere with the image pattern especially when the signal-to-noise ratio is low. These make data analysis for 3D SPT challenging if not impossible. Here, using the astigmatic diamond PSFs as an example, we developed a new algorithm based on deep neural networks (DNNs) for 3D SPT. We show that this algorithm can effectively recognize complicated image patterns. Especially, this algorithm shows resistance to high noise and high interfering background. It becomes possible to achieve 3D SPT in highly congested cellular environment. The underlying mechanism is that DNNs can extract the most significant features from the background in both the training and the sample sets. Although only the diamond PSF and 3D tracking is tested, the method can be generalized and applied in a variety of pattern recognition-based problems. We envision DNN can greatly improve the capability in 3D SPT, not merely improving the time and spatial resolutions but also making impractical problems possible.

Developing three-dimensional single particle tracking in complicated environment using deep learning neural networks

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.