College of Science & Engineering
Photoactivation localization microscopy (PALM) is a well-established super-resolution method that was recognized with the Nobel Prize in Chemistry a few years ago. Unfortunately, the original implementation of PALM is restricted to a single imaging plane and yields 2D super-resolution images, which significantly limits the its biological applications. To overcome this challenge a specially designed phase plate inserted in the Fourier imaging plane of the microscope modulates the point spread function (PSF) of the instrument and adds axial super-resolution information, a technique known as PSF engineering. These researchers use a double helix (DH) PSF, which splits each object into two lobes with the height of the object encoded by the rotation angle of the lobes with respect to a chosen direction. The relationship between the lobe position and the axial height of an object are established in calibration experiments. The analysis of the lobes is performed by Gaussian fitting of each lobe to determine its position, which leads to systematic biases as the shape of the lobes changes as well with height. This project is aimed at removing these biases by implementing and optimizing a machine learning algorithm that recognizes the shape and angular orientation of the lobes as a function of height and applies these templates to experimental data to accurately determine the 3D location of biomolecules with a resolution of better than 50 nm.