University of British Columbia
Macrophage phenotyping using autofluorescence microscopy and machine learning
Abstract
dc:descriptionMacrophages are highly plastic immune cells which morph into different phenotypes depending on their surrounding environment. The two main paradigms for macrophage phenotyping are phenotyping by function and by origin. Phenotyping based on function separates macrophages into three polarizations: naïve (M0), pro-inflammatory (M1), and anti-inflammatory (M2). Phenotyping based on origin distinguishes macrophages from their place of origin. This can be from monocytes created in the bone marrow, or local proliferation from tissue resident macrophage populations. These phenotypes have different functions and can be associated with different metabolic states, which can be detected via immunofluorescence assays or autofluorescence. Detecting these phenotypes can be a powerful tool in the diagnosis of different diseases, as many diseases impact macrophage populations in the affected tissue. Currently macrophage phenotyping is difficult and requires use of unreliable surface markers or expensive sequencing techniques. In this thesis we use fluorescent microscopy-based images for the training and validation of different image-based algorithms for effective classification. We first imaged THP-1 derived M0/M1/M2 macrophages for brightfield and fluorescently stained mitochondrial images. We then applied this dataset to multiple state-of-the-art deep learning models, achieving 5-fold validation accuracy of 80.50% using PNASNet-5. We also explore phenotyping murine tissue resident macrophages, bone-marrow monocyte-derived macrophages, and monocytes. This was conducted by imaging for autofluorescence and applying to an original convolutional neural network (MacNet) as well as using CellProfiler feature engineering and classical classifiers. We achieve 97.72% validation accuracy using MacNet and 98.58% validation accuracy on CellProfiler features in random forest.
Degree
thesis:*- Name thesis:degree_name
- Master of Applied Science - MASc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Mechanical Engineering
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Xu, Alec
- Contributors dc:contributor
-
- Ma, Hongshen
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/82745