Using Artificial Intelligence to Identify Tumor Microenvironment Heterogeneity in Non-Small Cell Lung Cancers

Area of Science:

• Oncology
• Computational Pathology

Background:

• Lung cancer treatment is hindered by tumor heterogeneity, including diverse cell types within the tumor microenvironment.
• Current methods for analyzing tumor-associated cells are often complex, costly, and destructive.

Purpose of the Study:

• To evaluate the utility of HALO AI nuclear phenotyping software for characterizing cell types in lung cancer.
• To assess the correlation between cell type distribution and prognosis in human lung cancer.

Main Methods:

• Utilized standard hematoxylin and eosin (H&E) stained slides from murine and human lung cancer samples.
• Employed HALO AI nuclear phenotyping software to identify and quantify six distinct cell types.
• Validated lymphocyte and macrophage identification using CD3 and F4/80 immunohistochemistry, respectively.

Main Results:

• Macrophages were predominant in murine adenocarcinomas, while neutrophils predominated in squamous cell carcinomas.
• Human samples showed negative correlations between neutrophils and lymphocytes, and mesenchymal cells and lymphocytes.
• Increased mesenchymal cell percentages correlated with poorer prognosis in human lung cancer.

Conclusions:

• HALO AI software effectively identifies, quantifies, and compares cell type distributions on standard H&E slides.
• This simple, cost-effective AI approach can benefit lung cancer research and clinical treatment strategies.