Abstract
Volumetric pancreatic cyst MRI segmentation remains challenging because cysts are typically small, sparsely distributed, exhibit ambiguous boundaries, and often appear in only a limited number of slices. Existing 2D segmentation networks effectively capture fine-grained in-plane semantics but ignore inter-slice anatomical context, whereas 3D networks capture volumetric information but may be less effective for segmenting small and sparsely distributed lesion regions. To address these limitations, we propose SliceLift, a sparse slice-guided hybrid 2D-3D framework for volumetric pancreatic cyst MRI segmentation. Instead of densely encoding every slice, SliceLift extracts multi-scale representations from a sparse set of uniformly sampled slices and lifts them into the corresponding 3D feature space through a slice-to-volume feature lifting strategy. The lifted features are fused with volumetric encoder representations and decoded by a full-scale 3D decoder, enabling effective integration of fine-grained slice-level semantics and global anatomical context. Experiments on the multi-center Cyst-X dataset demonstrate the effectiveness of the proposed framework. SliceLift achieves a Dice score of 0.5116, substantially outperforming the best 2D (0.4226) and 3D (0.4190) baselines. These results demonstrate that selectively integrating sparse slice-level semantics with volumetric feature learning provides an effective solution for pancreatic cyst MRI segmentation. Our code is available at https://github.com/ntqduy/SliceLift.git
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MLMI_082.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/MLMI_082_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=xkxfdWhr2g
BibTex
@InProceedings{NguTra_SliceLift_MICCAISAT2026,
author = { Nguyen, Tran Quoc Duy AND Vu, Vi AND Nguyen, Thanh-Huy AND Aktas, Halil Ertugrul AND Keles, Elif AND Tasci, Eminenur Sen AND Tasci, Muhammed Enes AND Durak, Gorkem AND Miller, Frank H. AND Wallace, Michael B. AND Bagci, Ulas},
title = { { SliceLift: Slice-to-Volume Feature Lifting for Volumetric Pancreatic Cyst MRI Segmentation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17272},
month = {pending},
page = {pending}
}
