Abstract
The prevalence of liver fibrosis and cirrhosis is increasing globally, which re-quires automatic and accurate segmentation and detection of the liver. Prepara-tion of larger labeled data remains the challenge for developing automatic, generalizable liver segmentation and staging models. This study utilized an it-erative human-in-the-loop label generation technique for developing a general-izable liver segmentation model. The nnUNet model was optimized and itera-tively refined to get the final version of the proposed liver nnUNetLiver. The liver pseudo ground truth labels were generated iteratively from large online available unlabeled multimodality abdominal scans and the CARE Liver 2026 training datasets, starting from the available limited labeled scans. Consequent-ly, a liver segmentation model was developed using a large cohort of datasets obtained by merging labeled and pseudo-labeled datasets. The model accuracy was substantially enhanced in each iteration, and this approach also reduced the manual task of generating ground truth labels one by one. The proposed fi-nal version of the model resulted in Dice coefficients over the range of 0.92 to 0.98 on testing data for the DWI, T2W, T1W, and GED4 sequences of liver data. Dice coefficients of 0.9736 and 0.9821 were obtained while testing the trained model on the same center and new center GED4 challenge testing da-tasets, respectively. These results demonstrate that the developed model per-forms exceptionally well for segmenting the liver from multimodality multi-centric MRI data.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CARE_007.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=ZCPJHuR05E
BibTex
@InProceedings{KasEwu_Iterative_MICCAISAT2026,
author = { Kassaw, Ewunate Assaye AND Sani, Peush AND Mehindiratta, Amit AND Singh, Anup},
title = { { Iterative Human-in-the-Loop Learning for Accurate Segmentation of Liver } },
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}
}
