LeCor improves interactive 3D lung tumour outlining with fewer corrections
LeCor: Learning to Be Corrected by Meta-Learned Test-Time Training for Interactive 3D Lung-Tumour Segmentation
Computer Vision and Pattern RecognitionMachine Learning
Summary
Drawing the borders of lung tumours on CT scans can take a long time for doctors during radiotherapy planning. The researchers show that while existing AI models help by letting doctors make corrections that influence the whole scan session, their accuracy stops improving after a few corrections. They propose LeCor, a new method that treats each correction as a way to quickly train the model during the session, improving performance on untouched slices. LeCor achieves higher accuracy faster, meaning fewer corrections are needed to get a good tumour outline.
What this means in practice
- •For radiotherapy planning teams: Use LeCor to reduce clinician time spent correcting tumour outlines on CT by improving accuracy with fewer interactive corrections.
- •For medical imaging software developers: Integrate LeCor’s meta-learned test-time training approach to enhance segmentation tools that improve interactively from user inputs.
Authors
Yi Luo, Yike Guo, Wenxuan Li, Zongwei Zhou, Rui Zhang, Kai Ding
Abstract
Delineating lung tumours on computed tomography (CT) takes a considerable share of the time spent on radiotherapy planning, and a contour proposed by a model can be refined interactively by the clinician. Promptable foundation models such as SAM 3 support this workflow by writing each correction into a session memory that conditions the remaining slices, while the model weights stay fixed. On 690 test cases from five public CT cohorts, fine-tuning SAM 3 on lung tumours raises the Dice obtained from a single point prompt from 0.298 to 0.757, and seven rounds of corrections raise it further to 0.765, but under memory conditioning alone the accuracy on slices the annotator has not touched stops improving after six rounds. We therefore treat each correction as a training signal and propose LeCor, which performs test-time training on a small set of case adapters that are reset for every case and meta-learned such that a single gradient step driven by a click improves the slices that were not clicked. On the 133 test cases that span at least eight slices, LeCor raises the Dice reached after seven correction rounds from 0.787 with the fine-tuned model to 0.827, reduces the number of cases that never reach a Dice of 0.80 from 47 to 27, and reaches in three correction rounds the accuracy that the fine-tuned model attains in seven.