Annotation-granularity versus scalability in large-scale medical imaging datasets

Determine how to balance annotation granularity and scalability when constructing large-scale medical image datasets, particularly to support reliable lesion-level annotations alongside scalable organ-level annotation coverage.

Background

MetaStructAtlas provides organ-level functional assessment but does not include lesion-level segmentation annotations. Consequently, precise tumor delineation, sub-centimeter nodule characterization, and lesion-burden quantification are not directly addressed. The authors identify the trade-off between obtaining fine-grained annotations and maintaining the reliability and scalability required for large medical imaging datasets as an unresolved methodological problem.

References

This is a recognized challenge in large-scale medical image dataset construction, where the trade-off between annotation granularity and scalability remains an open problem .

— MetaStructAtlas: A Grounded 3D Vision-Language Dataset and Benchmark for Functional and Structural Reasoning in Whole-Body PET/CT  (2609.03690 - Zheng et al., 3 Sep 2026) in Section Discussion and Conclusion, paragraph beginning “Several limitations of the current work merit acknowledgment”