Reproducibility assessment of a published cell shape transition analysis workflow

Image quantification is central to modern biomedical research. However, the reproducibility of image-based studies remains a persistent challenge due to limited access to raw data and incomplete reporting of analysis workflows. To assess these issues, the Global Bioimage Analyst Society (GloBIAS) launched a coordinated effort in 2024 to examine published bioimage analysis workflows. Here, we evaluated the reproducibility of a published workflow for analysing cell shape transitions in Dictyostelium discoideum. While the reported kymographs and biological conclusions could be qualitatively reproduced, a critical upstream step - cell segmentation - was insufficiently described. Reproducing the segmentation step based on the available information required extensive parameter tuning and failed to consistently capture accurate cell contours, thereby compromising downstream analysis. To evaluate whether the downstream analytical components could be reproduced when segmentation quality was improved, we adopted an alternative segmentation strategy using ilastik, a machine learning-based image analysis tool. Using cell-boundary coordinates extracted from ilastik-based masks as alternative inputs for downstream analysis, we reconstructed kymographs that partially matched the published results. Our findings demonstrate that accessible software and raw data alone are insufficient to ensure reproducibility in image-based research. Complete reporting of image-processing workflows, including segmentation strategies and parameter settings, is essential for the reliable reproduction of published quantitative analyses.