Multiple Signal Classification Algorithm: A computational microscopy tool for fluorescence microscopy

Abstract

Fluorescence Microscopy is still the most preferred tool for studying the cell's inner structures. One of the reasons for this preference is its great selectivity, which allows it to label specific types of structures and visualize them with high contrast. However, its resolution is conventionally limited due to the diffraction of light, which makes the study of cells at scales below 200 nm challenging. To go beyond this limitation, different techniques are required, such as Super-resolution microscopy.

The objective of this thesis is to explore and further develop one of these techniques: the Multiple Signal Classification Algorithm (MUSICAL), a computational method that enables the reconstruction of super-resolution images from a temporal stack of diffraction-limited fluorescent images. MUSICAL exploits the fluctuations in light intensity produced by the fluorescent emitters in the sample to improve resolution and contrast. It is based on the separation of a signal and noise component, followed by their subsequent recombination.

This thesis includes five publications that contribute to different aspects of MUSICAL. Paper I enhances the usability of the method, bridging the gap between the algorithm and the end-user through an easy-to-use ImageJ plugin. This plugin allows the customization of various parameters for quick testing and facilitates the processing of multi-channel data and long stacks of images. Papers II and III contribute to expanding the technical knowledge of the algorithm by extending its definition and providing further insights into its working mechanism. In these papers, soft MUSICAL and contrast enhancement (CE) are presented for the first time. These methods impose fewer constraints on the separation between signal and noise and rely less on the user's input.

Finally, the thesis concludes with Papers IV and V, which provide case studies of practical illumination mechanisms for microscopy that can potentially complement MUSICAL to achieve a resolution gain up to three times the conventional limit. These two systems correspond to chip-based microscopy and structured illumination, whose data was studied and then processed with MUSICAL.

Overall, this thesis contributes to the improvement of MUSICAL in three key areas and can be read as a story of the algorithm. Therefore, it also serves as a window briefly showcasing the tremendous potential behind MUSICAL, which will hopefully inspire further developments in super-resolution microscopy.