LesionLocator: Zero-Shot Universal Tumor Segmentation and Tracking in 3D Whole-Body Imaging

Maximilian Rokuss, Yannick Kirchhoff, Seval Akbal, Balint Kovacs, Saikat Roy, Constantin Ulrich, Tassilo Wald, Lukas T. Rotkopf, Heinz-Peter Schlemmer, Klaus Maier-Hein · 2025

Read LesionLocator: Zero-Shot Universal Tumor Segmentation and Tracking in 3D Whole-Body Imaging by Maximilian Rokuss, Yannick Kirchhoff, Seval Akbal, Balint Kovacs, Saikat Roy, Constantin Ulrich, Tassilo Wald, Lukas T. Rotkopf, Heinz-Peter Schlemmer, Klaus Maier-Hein (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-20985.txt.

Description

Rokuss and a team including Maier-Hein present LesionLocator in twenty twenty-five, a model for tumor segmentation and tracking. Clinicians need to find lesions in whole body scans and follow them over time. Prior promptable models performed poorly on this task. Longitudinal data for training are also scarce.

The model trains on more than twenty-three thousand annotated scans, plus synthetic longitudinal data. It accepts dense spatial prompts and outputs lesion masks in three dimensions. Tracking extends across time, making it a four dimensional end to end system. It beat all existing promptable models by near ten Dice points on segmentation. Tracking results set a new state of the art for retrieval and segmentation. The synthetic data set and model are released openly.

Catalog record

CREATORMaximilian Rokuss, Yannick Kirchhoff, Seval Akbal, Balint Kovacs, Saikat Roy, Constantin Ulrich, Tassilo Wald, Lukas T. Rotkopf, Heinz-Peter Schlemmer, Klaus Maier-Hein
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-20985

Browse the full library · About the Tower of Babel