Generating Post-Acetazolamide Cerebral Blood Flow MRI for High-Risk Stroke Patients

Rydham Goyal, Camila Gonzalez, Sasha Alexander, Aja Zou, Michael E Moseley, Moss Y Zhao, Gary K Steinberg
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:3899-3910, 2026.

Abstract

Cerebrovascular reserve (CVR) quantifies the brain’s ability to augment cerebral blood flow in response to a vasodilatory stimulus. It is a key biomarker in Moyamoya disease and other steno-occlusive cerebrovascular disorders. Clinically, CVR is typically assessed by administering acetazolamide (ACZ) and acquiring post-ACZ perfusion maps, but this workflow is time-consuming, costly, and contraindicated in a subset of patients. In this work, we investigate whether deep learning can predict post-ACZ perfusion directly from baseline arterial spin labeling (ASL) MRI, enabling pharmacologic-free CVR estimation. We curate a single-center dataset of Moyamoya ASL perfusion imaging, comprising pre/post-ACZ scan pairs from 194 patients. We design a post-ACZ conditional Autoencoder (cAE) network to regress the middle axial post-ACZ slice from the corresponding pre-ACZ slice using a combined L1 and SSIM loss. We evaluate our method against three diffusion-based formulations (conditional DDPM, Cold Diffusion, and Residual Diffusion). On a holdout test set of 49 patients, the proposed post-ACZ cAE achieves the highest reconstruction fidelity (SSIM $\approx$ 0.79), outperforming diffusion-based baselines in MAE, SSIM, and PSNR. Region-wise analysis of CBF percentage change in affected versus healthy MCA territories showed that the generated post-ACZ model outputs followed ground truth patterns of cerebrovascular reserve. Our findings demonstrate the feasibility of non-invasive CVR assessment using MRI for high-risk patients. Our data-driven approach could reduce reliance on ACZ challenges in routine clinical workflow and expand access to CVR testing to evaluate brain health.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-goyal26a, title = {Generating Post-Acetazolamide Cerebral Blood Flow MRI for High-Risk Stroke Patients}, author = {Goyal, Rydham and Gonzalez, Camila and Alexander, Sasha and Zou, Aja and Moseley, Michael E and Zhao, Moss Y and Steinberg, Gary K}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {3899--3910}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/goyal26a/goyal26a.pdf}, url = {https://proceedings.mlr.press/v315/goyal26a.html}, abstract = {Cerebrovascular reserve (CVR) quantifies the brain’s ability to augment cerebral blood flow in response to a vasodilatory stimulus. It is a key biomarker in Moyamoya disease and other steno-occlusive cerebrovascular disorders. Clinically, CVR is typically assessed by administering acetazolamide (ACZ) and acquiring post-ACZ perfusion maps, but this workflow is time-consuming, costly, and contraindicated in a subset of patients. In this work, we investigate whether deep learning can predict post-ACZ perfusion directly from baseline arterial spin labeling (ASL) MRI, enabling pharmacologic-free CVR estimation. We curate a single-center dataset of Moyamoya ASL perfusion imaging, comprising pre/post-ACZ scan pairs from 194 patients. We design a post-ACZ conditional Autoencoder (cAE) network to regress the middle axial post-ACZ slice from the corresponding pre-ACZ slice using a combined L1 and SSIM loss. We evaluate our method against three diffusion-based formulations (conditional DDPM, Cold Diffusion, and Residual Diffusion). On a holdout test set of 49 patients, the proposed post-ACZ cAE achieves the highest reconstruction fidelity (SSIM $\approx$ 0.79), outperforming diffusion-based baselines in MAE, SSIM, and PSNR. Region-wise analysis of CBF percentage change in affected versus healthy MCA territories showed that the generated post-ACZ model outputs followed ground truth patterns of cerebrovascular reserve. Our findings demonstrate the feasibility of non-invasive CVR assessment using MRI for high-risk patients. Our data-driven approach could reduce reliance on ACZ challenges in routine clinical workflow and expand access to CVR testing to evaluate brain health.} }
Endnote
%0 Conference Paper %T Generating Post-Acetazolamide Cerebral Blood Flow MRI for High-Risk Stroke Patients %A Rydham Goyal %A Camila Gonzalez %A Sasha Alexander %A Aja Zou %A Michael E Moseley %A Moss Y Zhao %A Gary K Steinberg %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-goyal26a %I PMLR %P 3899--3910 %U https://proceedings.mlr.press/v315/goyal26a.html %V 315 %X Cerebrovascular reserve (CVR) quantifies the brain’s ability to augment cerebral blood flow in response to a vasodilatory stimulus. It is a key biomarker in Moyamoya disease and other steno-occlusive cerebrovascular disorders. Clinically, CVR is typically assessed by administering acetazolamide (ACZ) and acquiring post-ACZ perfusion maps, but this workflow is time-consuming, costly, and contraindicated in a subset of patients. In this work, we investigate whether deep learning can predict post-ACZ perfusion directly from baseline arterial spin labeling (ASL) MRI, enabling pharmacologic-free CVR estimation. We curate a single-center dataset of Moyamoya ASL perfusion imaging, comprising pre/post-ACZ scan pairs from 194 patients. We design a post-ACZ conditional Autoencoder (cAE) network to regress the middle axial post-ACZ slice from the corresponding pre-ACZ slice using a combined L1 and SSIM loss. We evaluate our method against three diffusion-based formulations (conditional DDPM, Cold Diffusion, and Residual Diffusion). On a holdout test set of 49 patients, the proposed post-ACZ cAE achieves the highest reconstruction fidelity (SSIM $\approx$ 0.79), outperforming diffusion-based baselines in MAE, SSIM, and PSNR. Region-wise analysis of CBF percentage change in affected versus healthy MCA territories showed that the generated post-ACZ model outputs followed ground truth patterns of cerebrovascular reserve. Our findings demonstrate the feasibility of non-invasive CVR assessment using MRI for high-risk patients. Our data-driven approach could reduce reliance on ACZ challenges in routine clinical workflow and expand access to CVR testing to evaluate brain health.
APA
Goyal, R., Gonzalez, C., Alexander, S., Zou, A., Moseley, M.E., Zhao, M.Y. & Steinberg, G.K.. (2026). Generating Post-Acetazolamide Cerebral Blood Flow MRI for High-Risk Stroke Patients. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:3899-3910 Available from https://proceedings.mlr.press/v315/goyal26a.html.

Related Material