# Cell Invasion Analysis of Tumor Spheroids Using 2D Image Data

## TL;DR

This paper introduces a new method to analyze how tumor spheroids invade surrounding tissue using fluorescence images.

## Contribution

A novel methodology is presented for detecting and quantifying tumor spheroid invasion using fluorescence imaging and automated mask computation.

## Key findings

- Two strategies for mask computation were developed to handle different spheroid shapes and boundary behaviors.
- The method enables analysis of images with nonconstant backgrounds, common in fluorescence imaging.
- The approach allows for automated evaluation with manual parameter adjustments for accuracy.

## Abstract

Metastatic disease is the most severe complication in
oncological
patients. The quantification of cellular invasion into the surrounding
tissue is crucial for the identification of strategies to suppress
this process. Extracellular matrix-embedded 3D cancer models, such
as spheroids and organoids, are commonly used to mimic tumor progression
under in vitro conditions. However, robust and widely
used algorithms to detect and quantify spheroid growth and invasion
into the surrounding matrix are still lacking. In this study, we use
fluorescently labeled 3D models, as fluorescence images are generally
of higher quality than bright-field images. We present a methodology
to compute the mask of the spheroid core and to detect and characterize
cells outside this mask. We have developed two strategies for mask
computation, one for compact spheroids and another for models that
lose their boundaries soon after insertion into the extracellular
matrix. In both modes, masks can be created for spheroids of various
shapes. Cells or their clusters outside the mask are recognized on
the basis of filtered local maxima. This method enables the analysis
of images with a nonconstant background, which is often found in real
fluorescence images. The evaluation is largely automated but allows
visual inspection based on the overlay of the objects detected by
the algorithm with the original fluorescence signal of the spheroid
core and the invading cells. A user-friendly manual adjustment of
the parameters for mask fitting and cell detection is implemented.