Abstract

Risk stratification of intraductal papillary mucinous neoplasms (IPMNs) offers a crucial opportunity for early intervention against pancreatic cancer but typically requires invasive biopsy. Radiomics and deep-learning representations, multisequence MRI (T1W/T2W), and anatomically decomposed (head/body/tail) analysis provide complementary signals; fusing them effectively is central to ordinal prediction of IPMN dysplasia. We present cUPMI, a class-conditional Gaussian augmentation of a stacking combiner’s log-probability meta-features. In our multi-center analysis, cUPMI adds limited value to properly regularized L2 - logistic binary stacks but consistently regularizes higher-capacity tree combiners (RF ∆+0.015, XGBoost ∆+0.024 binary AUC, positive in all seeds), with its cleanest ordinal benefit for XGBoost on an 8-stream radiomics task (+0.022 QWK in all seeds). Separately, fold-locked fusion of radiomics and 2.5D CNN streams yields the strongest overall model, an RF stack reaching QWK 0.595 (95% CI [0.54, 0.64]) and binary AUC 0.839, surpassing radiomics, 2.5D ResNet, and 3D DenseNet-121 baselines. Code: https://github.com/maxanelson/cUPMI.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MLMI_073.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=kbOL4CcyYc

BibTex

@InProceedings{NelMax_Gaussian_MICCAISAT2026,
        author = { Nelson, Max A. AND Tasci, Eminenur Sen AND Wang, Zhixiang AND Zhou, Zongwei AND Aktas, Halil Ertugrul AND Bejar, Andrea M. AND Keles, Elif AND Hong, Ziliang AND Taflan, Sıtkı Safa AND Tasci, Muhammed Enes AND Miller, Frank H. AND Wallace, Michael B. AND Keswani, Rajesh N. AND Durak, Gorkem AND Bagci, Ulas},
        title = { { Gaussian Meta-Space Augmentation for Stacking Ensembles in Multimodal IPMN Risk Stratification } },
        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}
}


back to top