Abstract
Advanced medical imaging is often limited in low-resource clinical settings because high-cost scanners, specialized acquisition protocols, and expert imaging infrastructure are not uniformly available. In contrast, ultrasound is comparatively inexpensive, portable, and widely used, but provides lower soft-tissue contrast than magnetic resonance imaging (MRI). Cross-modality synthesis from ultrasound to MRI therefore offers a potential route for improving MRI-like visualization when MRI access is constrained. In this work, we investigate this problem in prostate imaging and propose a PCA-FiLM extension of a supervised Pix2Pix model for prostate US→MRI synthesis. Patient-level radiomic features are extracted from MRI volumes, reduced to a compact 16 dimensional representation using principal component analysis, and injected into the generator decoder through Feature-wise Linear Modulation. This target-derived conditioning is used as a proof of concept to investigate whether compact quantitative priors can guide cross-modality synthesis, rather than as a deployment-ready inference strategy. Experiments are conducted on the public PROSTATE-MRI-US-BIOPSY dataset using patient-level five-fold evaluation. Compared with the baseline Pix2Pix model, PCA-FiLM improves structural similarity while preserving downstream radiomics-based UCLA risk classification performance. These findings suggest that compact radiomic conditioning can provide useful patient-level context for cross-modality prostate image synthesis, while further validation with clinically available conditioning signals, volumetric models, and clinical reader assessment remains necessary.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MIRASOL_021.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/MIRASOL_021_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=UD5c1rBAKW
BibTex
@InProceedings{NasMoh_Compact_MICCAISAT2026,
author = { Nasser AbdelRaouf, Mohamed AND Lamey, Christina Maher AND Mohamed Abdelhalim, Hagar AND Selim, Sahar},
title = { { Compact Radiomics-Guided FiLM Conditioning for Ultrasound-to-MRI Translation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17264},
month = {pending},
page = {pending}
}