Dissertation
Machine Learning in Confocal Laser Microscopy and Spectroscopy
Washington State University
Doctor of Philosophy (PhD), Washington State University
2022
DOI:
https://doi.org/10.7273/000005535
Abstract
Confocal laser scanning microscopy (CLSM) is a preferred method for obtaining optical images with submicron resolution. Replacing the pinhole and detector of a CLSM with a digital camera (CCD or CMOS) has the potential to simplify the design and reduce cost. However, the relatively slow speed of a typical camera results in long scans. To address this issue, in the present investigation a microlens array (MLA) was used to split the laser beam into 48 beamlets that are focused onto the sample. In essence, 48 pinhole-detector measurements were performed in parallel. Images obtained from the 48 laser spots were stitched together into a final image.Photoluminescence (PL) spectroscopy is a non-destructive optical method that is widely used to characterize semiconductors. In the PL process, a substance absorbs photons and emits light with longer wavelengths. This paper discusses a method for identifying substances from their PL spectra using machine learning, a technique that is efficient in making classifications. Neural networks were constructed by taking simulated PL spectra as the input and the identity of the substance as the output. Six different semiconductors were chosen as categories: gallium oxide (Ga2O3), zinc oxide (ZnO), gallium nitride (GaN), cadmium sulfide (CdS), tungsten disulfide (WS2) and cesium lead bromide (CsPbBr3). The developed algorithm has a high accuracy (>90%) for assigning a substance to one of these six categories from its PL spectrum.
With an XY stage, a CLSM can scan a large area on a sample. Adjusting the height of the objective is necessary which made the laser beam could focus on the sample surface. However, if the surface of the sample is not flat, the laser spot will go in and out of focus, causing bad scanning results. Deep learning especially convolutional neural networks is an efficient way to treat images. It shows its success in the field of object detection, image classification, face recognition, etc. The deep learning techniques were used to design a model that predicts the out-of-focus distance with the image of laser spot. The model can develop to a system that could automatically focusing the CLSM in real time.
Metrics
3 File views/ downloads
25 Record Views
Details
- Title
- Machine Learning in Confocal Laser Microscopy and Spectroscopy
- Creators
- Yinchuan Yu
- Contributors
- Matthew M.D. McCluskey (Advisor)Mark M.G. Kuzyk (Committee Member)Yi Y. Gu (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Department of Physics and Astronomy
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Publisher
- Washington State University
- Number of pages
- 151
- Identifiers
- 99901051427001842
- Language
- English
- Resource Type
- Dissertation