Abstract
Pretreatment prediction of pathological complete response (pCR) after neoadjuvant chemotherapy supports individualized breast cancer (BC) treatment, but remains challenging due to heterogeneous tumor biology and variable MRI response patterns. We propose a multiencoder gated-fusion framework that integrates clinical variables, tumorlevel, and breast enhancement features from pretreatment dynamic contrastenhanced MRI. Each feature modality is processed by an independent encoder to learn modality-specific representations, after which a lightweight gating module performs patient-adaptive fusion by weighting each modality embedding prior to classification. The proposed model achieved the best internal performance, with a balanced accuracy of 0.80 ± 0.04 (mean ± std) and an AUC of 0.79 ± 0.06, outperforming baseline models. On external validation, it retained a balanced accuracy of 0.71 ± 0.08 and AUC of 0.76 ± 0.06. Modality attribution and SHAP analysis suggested complementary contributions from clinical and MRI-derived phenotypes, supporting model interpretability.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/Deep_Brea3th_015.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=muAIrmDlcy
BibTex
@InProceedings{KhaAdn_Adaptive_MICCAISAT2026,
author = { Khalid, Adnan AND Mursil, Muhammad AND Meriaudeau, Fabrice AND Lalande, Alain AND Puig, Domenec AND Rashwan, Hatem A.},
title = { { Adaptive Multi-Encoder Gated Fusion for Interpretable Prediction of Pathological Complete Response in Breast Cancer } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17256},
month = {pending},
page = {pending}
}