Abstract
Accurate liver fibrosis staging is central to prognosis and treatment, but biopsy is invasive and routine clinical MRI is heterogeneous across vendors, sequences and missing modalities. We developed a deployable segmentation-to-staging cascade for the MICCAI CARE-Liver challenge that stages fibrosis directly from multi-parametric ab-dominal MRI while remaining self-contained at inference. Using 460 training and 60 validation cases, with dense liver masks for only 30 GED4 volumes, LiSeg fine-tunes TotalSegmentator-MRI within nnU-Net to segment the liver from GED4, and the mask defines a liver-centred crop for LiFS. For each of three retained sequences (GED4, DWI_800, T2) we train a single-channel, multi-task 3-D ResNet-10 ensemble initialised from MedicalNet/Med3D and fuse per-modality probabilities by missing-modality-robust late fusion. LiSeg achieved cross-validation Dice 0.9701. In nested five-fold out-of-fold evalua-tion, weighted LiFS fusion reached mean endpoint AUC 0.7639 and accuracy 0.7413. Official validation was AUC 0.7022/0.7007 for the S1- and S4-versus-rest endpoints. Ablations identified DWI_800 as the strongest single sequence, showed that additional dynamic contrast phases did not improve fusion, and found negligible benefit from hand-crafted morphology features. A compact three-sequence, trans-fer-learned cascade thus provides robust fibrosis staging on heteroge-neous multi-vendor MRI while explicitly handling scarce masks and missing modalities.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CARE_001.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=0yFC2C7qlH
BibTex
@InProceedings{LowJor_ATransferLearned_MICCAISAT2026,
author = { Low, Jordan Jun Yi AND Li, Lei AND Lyu, Yilin},
title = { { A Transfer-Learned Segmentation-to-Staging Cascade for Robust Liver Fibrosis Staging on Real-World Multi-Vendor MRI } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17273},
month = {pending},
page = {pending}
}