Using spatial proteomics to enhance cell type assignments in histology images

Area of Science:

• Computational pathology
• Biomedical imaging
• Machine learning in medicine

Background:

• Hematoxylin and eosin (H&E) staining is a cornerstone of clinical histopathology, offering morphological detail but lacking molecular specificity.
• Multiplexed spatial proteomics imaging provides rich molecular and cellular information but is not widely accessible in clinical settings.

Purpose of the Study:

• To develop a machine learning framework for enhanced cell type annotation on H&E-only histopathology images.
• To integrate insights from spatial proteomics with traditional H&E staining for improved diagnostic capabilities.

Main Methods:

• Leveraging histopathology foundation models trained on paired H&E and spatial proteomic imaging data.
• Developing and evaluating a machine learning approach for cell type annotation using H&E images.
• Utilizing kidney datasets for training and validation of the proposed framework.

Main Results:

• The developed machine learning framework significantly outperforms models trained directly on imaging data for cell type annotation.
• The approach successfully enables enhanced cell type identification on H&E-stained slides.
• Demonstrated utility of the framework in analyzing biological differences in kidney diseases.

Conclusions:

• The proposed machine learning framework effectively bridges the gap between morphological and molecular information in histopathology.
• This method offers a pathway to leverage advanced spatial proteomics insights within the constraints of standard clinical H&E imaging.
• The framework holds potential for advancing the study and diagnosis of various diseases, including kidney diseases.