![[2506.06092] LinGuinE: Longitudinal Guidance Estimation for Volumetric Tumour Segmentation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2506.06092] LinGuinE: Longitudinal Guidance Estimation for Volumetric Tumour Segmentation
Summary
LinGuinE introduces a novel framework for longitudinal volumetric tumor segmentation, enhancing tracking and mask generation across multiple scans without requiring longitudinal training.
Why It Matters
This research addresses a significant gap in tumor segmentation methods, which typically lack the ability to track lesions over time. By providing a framework that integrates image registration and guided segmentation, LinGuinE improves radiotherapy planning and response assessment, potentially leading to better patient outcomes in oncology.
Key Takeaways
- LinGuinE enables lesion-level tracking and volumetric mask generation from a single radiologist prompt.
- The framework is agnostic to temporal direction and does not require training on longitudinal data.
- State-of-the-art performance is achieved across multiple datasets with minimal degradation over time.
- Ablation studies reveal the importance of autoregression and pathology-specific fine-tuning.
- The code and benchmarking data are publicly available, promoting further research in this area.
Electrical Engineering and Systems Science > Image and Video Processing arXiv:2506.06092 (eess) [Submitted on 6 Jun 2025 (v1), last revised 26 Feb 2026 (this version, v2)] Title:LinGuinE: Longitudinal Guidance Estimation for Volumetric Tumour Segmentation Authors:Nadine Garibli, Mayank Patwari, Bence Csiba, Yi Wei, Kostantinos Sidiropoulos View a PDF of the paper titled LinGuinE: Longitudinal Guidance Estimation for Volumetric Tumour Segmentation, by Nadine Garibli and 4 other authors View PDF HTML (experimental) Abstract:Longitudinal volumetric tumour segmentation is critical for radiotherapy planning and response assessment, yet this problem is underexplored and most methods produce single-timepoint semantic masks, lack lesion correspondence, and offer limited radiologist control. We introduce LinGuinE (Longitudinal Guidance Estimation), a PyTorch framework that combines image registration and guided segmentation to deliver lesion-level tracking and volumetric masks across all scans in a longitudinal study from a single radiologist prompt. LinGuinE is temporally direction agnostic, requires no training on longitudinal data, and allows any registration and semi-automatic segmentation algorithm to be repurposed for the task. We evaluate various combinations of registration and segmentation algorithms within the framework. LinGuinE achieves state-of-the-art segmentation and tracking performance across four datasets with a total of 456 longitudinal studies. Tumour segmentati...