Utilizing deep learning to identify granuloma features associated with host responses to Mycobacterium tuberculosis
Year: 2023
Type: dissertation
Abstract: Tuberculosis is a disease caused by Mycobacterium tuberculosis (M. tb). M. tb infection causes various immune responses, one of which is the formation of lung granulomas. A granuloma is an isolated area of infection, where different immune cells and cytokines can interact with the pathogen. \n \nThe aim of this thesis is to find correlations between different granuloma features and host responses associated with M. tb infection. To achieve this, correlations in three scenarios are investigated: between the bacterial quantity expressed as colony-forming units (CFU) and the bacterial quantity enumerated by an artificial intelligence (AI) model, between granuloma image features and survival, and lastly between cell-level features and specific immune responses. \n \nAn AI model was used to identify granuloma features from whole slide images of the lungs of M. tb-infected Diversity Outbred mice. The granuloma image features are necrosis, bacterial mat, cells, the bacterial quantity expressed as CFU, and the bacterial quantity enumerated by the AI model. The cells class is a segmented area that contains a predominance of immune cells, and bacterial mat is a segmented area with a dense bacterial overgrowth. The cell-level features are macrophages, lymphocytes, neutrophils, plasma cells, the bacterial quantity expressed as CFU, and the bacterial quantity enumerated by the AI model. In addition, previous research data collected from these mice was utilized which includes the concentrations of the following cytokines: tumor necrosis factor-alpha (TNF-α), chemokine ligand 1 (CXCL-1), and interferon-gamma (IFN-γ). All of the data was analyzed using R programming software. \n \nThe following correlations were observed: The cells area correlates positively with survival, while necrosis, bacterial mat, and both bacterial quantities have negative correlations with survival in varying degrees. Both bacterial quantities correlate positively with each other as well as with TNF-α and CXCL-1. Macrophages, lymphocytes, and plasma cells all correlate negatively with TNF-α and CXCL-1. Neutrophils had virtually no correlations with any of the cytokines. None of the cell-level features correlate with IFN-γ.
Source: Aaltodoc (Aalto University)
Author: Rekinen, Saana
Language: English
Subfield: Infectious Diseases
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W6997201940