Extension from histology to three-dimensional medical imaging

Extend the SlideCRF framework from histological whole-slide image analysis to three-dimensional medical images whose volumes exhibit analogous spatial connectivity.

Background

The proposed framework models spatial and biological relationships among patches in two-dimensional histological whole-slide images. The authors note that three-dimensional medical image volumes have comparable spatial connectivity, suggesting that the framework may be transferable beyond histology. Developing and assessing such an extension remains an explicitly identified open direction.

References

Several directions remain open. A natural next step is to validate the proposed annotation protocols directly with pathologists, in order to confirm their clinical realism. Efficiency and accuracy could also be improved by exploiting patches at multiple resolutions, following the pyramidal structure of WSIs. Beyond histology, our framework could be extended to 3D medical images, where volumes exhibit analogous spatial connectivity.

— Whole-Slide Image Analysis under Realistic Few-Shot Annotation Protocols  (2608.30420 - Godelaine et al., 31 Aug 2026) in Section Conclusion, subsection “Future Work”