Abstract
Accurate assessment of regional lung ventilation is crucial for the diagnosis and management of obstructive lung diseases, such as chronic obstructive pulmonary disease (COPD) and asthma-COPD overlap syndrome (ACOS). While various functional imaging modalities, including xenon-enhanced dualenergy CT (DECT), have been utilized for this purpose, their widespread clinical adoption is often limited by high costs, radiation exposure, or restricted accessibility. Although deep learning-based image-to-image translation (I2IT) has emerged as a promising alternative to synthesize ventilation maps from standard paired CTs, a systematic comparative evaluation of current generative models regarding their clinical utility remains largely unexplored. In this study, we provide a comprehensive comparative analysis of generative paradigms—specifically GAN-based and Diffusion-based models—for CT-based ventilation map synthesis. Using a clinical dataset of 189 patients, we evaluate the synthesized maps using both conventional image fidelity metrics and clinical metrics essential for diagnostic reliability, including defect localization accuracy and physiological biomarkers. Our results demonstrate that while GAN-based models maintain stable global anatomical structures, they tend to over-smooth local functional defects, leading to lower sensitivity. In contrast, Diffusion-based models exhibit superior texture reconstruction and higher correlation with clinical ventilation defect percentage indices, albeit with occasional structural hallucinations. This study delineates the strengths and limitations of current I2IT frameworks, providing a guide for the clinical deployment of AI-based functional lung imaging.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SASHIMI_024.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=UACzY8DOgP
BibTex
@InProceedings{ChoCha_Synthesizing_MICCAISAT2026,
author = { Choi, Changyong AND Park, Pyeongjin AND Lee, Sang Min AND Kim, Namkug},
title = { { Synthesizing CT-based Functional Lung Ventilation Map: A Comparative Study of Generative Models for Image-to-Image Translation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17258},
month = {pending},
page = {pending}
}
