Do's and don'ts of tumor segmentation with 3D slicer: A practical guide for radiologists, by radiologists
Kalina Chupetlovska, Kevin Groot Lipman, Zuhir Bodalal, Francesco Marcello Aricò, Laurens Topff, Monique Maas, Stefano Trebeschi · European Journal of Radiology Artificial Intelligence · 2025
Background Accurate and consistent medical image segmentation is essential for both clinical practice and research, particularly with the growing role of artificial intelligence. However, while technical aspects are well described in the literature, radiologists often lack practical, hands-on guidance tailored to their role. Methods This educational article delivers practical, experience-based recommendations, using real-world examples to illustrate common pitfalls and effective strategies for achieving accurate and consistent medical image segmentations. Results Key steps before initiating a segmentation project are outlined, including defining goals, establishing protocols, and collaborating with technical experts. We demonstrate how imaging parameters such as slice thickness, phase, sequence, and windowing influence segmentation quality and reproducibility. Common pitfalls such as over- and under-segmentation, floating or unfilled pixels, and misuse of software tools are addressed with practical solutions. The limitations of segmentation as the ‘ground truth' in AI research are addressed, with emphasis on inter-rater variability as well as differences in experience and software familiarity. Conclusions By adopting these best practices, radiologists can improve the accuracy and consistency of their segmentations, minimize errors, and contribute to more reliable and reproducible outcomes in both clinical and research settings.