Brightfield spheroid segmentation dataset for Box 1 and Figure 2: BT-20 TNBC spheroids analysed with SpheroidSizer, OrganoSeg2, and Ilastik

Abstract

This dataset contains brightfield images of BT-20 triple-negative breast cancer (TNBC) multicellular spheroids used in Box 1 and Figure 2 of the review article "The Evolution of Automated Spheroid Image Analysis from Rule-Based Methods to Deep Learning." Two imaging conditions are represented: an untreated spheroid with clearly defined boundaries, and a drug-treated spheroid surrounded by dead cell debris. Segmentation outputs from three analytical tools are included: SpheroidSizer outlines, OrganoSeg2 quality-control images and segmentation masks, and Ilastik probability maps and cyan overlay images. The ImageJ macro used to process Ilastik probability maps into segmentation overlays is also provided. These files are intended to illustrate characteristic segmentation behaviour across rule-based and classical machine learning analytical generations under clean and challenging post-treatment imaging conditions.