Machine learning analysis of biological cell laser spectral imaging

• Main Author: Sun, Jinglei
• Other Authors: Y. C. Chen
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/151956
• Institution: Nanyang Technological University

Cells are the most basic part of the human body. When cells proliferate abnormally or grow larger, it may indicate the occurrence of cancer. Therefore, the size of cells plays an important role in the characteristics of cells. With the existing technology, by fixing the cell in a resonant cavity with a certain length, single cells can be excited in order to create the laser pattern which contains more variable information compared to fluorescence dye. The narrow and long laser pattern can be separated into different laser modes according to frequency. Different laser modes have complex and rich information, and of course they also contain content related to cell size. However, it is impossible to achieve artificial recognition because of the indistinguishable features. Therefore, machine learning is used to determine the relationship between laser mode and cell size. In machine learning, neural networks, especially convolution neural networks are often used in the processing of complex information such as picture features, because they have strong fitting and generalization capabilities, and show excellent capabilities in such processing. The results show that after optimization, the model can successfully use the processed cell laser pattern pictures for regression fitting, and the final prediction absolute error is within 1 micrometer