SpaHE-Infil: A spatial heterogeneity framework for decoding TME infiltration from H&E-stained slides.

Area of Science:

• Computational pathology
• Tumor immunology
• Precision medicine

Background:

• Spatial distribution of immune cells in the tumor microenvironment (TME) is crucial for immunotherapy response.
• Current methods for TME analysis are limited by sequencing dependence and spatial resolution.

Purpose of the Study:

• To develop a multimodal computational framework, SpaHE-Infil, for accurate spatial characterization of TME immune cells using standard H&E images.
• To validate the framework's ability to predict immunotherapy response and patient survival.

Main Methods:

• Integrated spatial transcriptomics and whole-slide H&E images to train a Random Forest model.
• Extracted morphological, textural, and density features to identify 12 core TME cell types in situ.
• Employed dynamic calibration to correct immune cell proportion biases in clinical samples.

Main Results:

• SpaHE-Infil accurately predicted spatial immune cell distribution across multiple cancer types, validated against multiplex immunohistochemistry (mIHC) and deconvolution algorithms.
• Stratification of TME immune infiltration using H&E images predicted enhanced immunotherapy response.
• TME immune infiltration predicted prolonged recurrence-free survival in clinical cohorts.

Conclusions:

• SpaHE-Infil offers a clinically applicable tool for spatial TME characterization from H&E images.
• The framework supports advancements in tumor immunology research and precision immunotherapy.
• Spatial TME profiling using H&E images can guide clinical decision-making for cancer treatment.