Combining multiple spatial statistics enhances the description of immune cell localisation within tumours

Area of Science:

• Computational pathology
• Digital pathology
• Histological image analysis

Background:

• Digital pathology allows large-scale computational analysis of histological images, including immune cell identification in tumors.
• Precise spatial coordinates of immune cells are extracted but challenging to interpret.
• Immune cell localization within tumors is linked to patient prognosis and functional status.

Purpose of the Study:

• To develop novel descriptors for immune cell spatial distributions in tumors.
• To assess the utility of spatial statistics in analyzing immune cell patterns.
• To improve the interpretation of spatial coordinates from immunohistochemistry slides.

Main Methods:

• Application of three spatial statistics to CD68+ macrophage locations in human head and neck tumors.
• Analysis of spatial statistics in relation to pathologist-based semi-quantitative groupings.
• Generation of a synthetic dataset to validate the methodology.
• Development of a maximum likelihood approach combining multiple spatial statistics.

Main Results:

• Pathologist-grouped images exhibited similar spatial statistics.
• Combining multiple spatial statistics with a maximum likelihood approach outperformed single statistics in predicting human classifications.
• The methodology allowed for error estimation in classifications.

Conclusions:

• A combined spatial statistics approach offers improved prediction of immune cell distribution classifications in tumors.
• This methodology is adaptable for other histological analyses and point pattern studies.
• Enhanced interpretation of spatial point patterns can provide valuable biological and clinical insights.