Spatial Phenotyping of Tumour Immune Microenvironment in the Non-small Cell Lung Cancer Using Artificial Intelligence
Loading...
Embargo End Date
ICR Authors
Authors
Zhang, H
Document Type
Thesis or Dissertation
Date
2024-02-26
Date Accepted
Abstract
Lung cancer is the second most common cancer type worldwide, with a five-year survival rate of about 20%. The aggressiveness of the disease and the inefficiency of treatment are partially attributed to the intra-tumour heterogeneity (ITH), which is reflected by the diverse cell interactions in the tumour microenvironment (TME), complicated arrangement of extracellular matrix, and the presence of heterogeneous tumour cell subpopulations. Deconvolving ITH from histo-pathological evaluations plays a critical role in risk stratification and guiding treatment strategies.
In this PhD thesis, I aim to use spatial statistics powered by artificial intelligence (AI) to decipher the ITH, specifically the heterogeneous tumour immune microenvironment in non-small cell lung cancer and to identify the prognostic value of image-derived features.
I first investigate the heterogeneous composition and interactions of immune cell types at different tissue compartments. Specifically, I identified differences in B and T cell interplays between intratumoral and peritumoral immune cell hotspots beyond tertiary lymphoid structures in lung squamous cell carcinoma, which might be associated with an immunosuppressive TME at the tumour-immune interface. I further developed spatial analysis methods to systematically characterise local TME surrounding individual tumour islands in lung squamous cell carcinoma, which revealed distinct phenotypes of tumour islands associated with patient outcomes, metastasis, and immune evasion. Lastly, I developed a self-supervised deep learning pipeline SANDI to enable cost-efficient yet accurate identification of diverse immune cell phenotypes on multiplex images. The SANDI pipeline facilitates the analysis of ITH and the identification of biomarkers in noisy patient samples.
To conclude, by developing AI and spatial statistical analysis pipeline tailored for digital pathology, I identified the composition and interplay of immune cells underpinning their spatial organisation pattern, which signifies the role of the tumour immune microenvironment in tumour progression and therapeutic response.
Citation
2024
DOI
Source Title
Publisher
Institute of Cancer Research (University Of London)
ISSN
eISSN
Collections
Research Team
Comp Path & Int Genomics