Chenyu Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Chenyu Wang has a total of 148 co-authors. Here are the top co-authors by number of shared publications:
Michael Barnett
University of Sydney
15 shared publications- Michael Barnett
Brain and Mind Centre, University of Sydney, Sydney, Australia
14 shared publications Mariano Cabezas
Brain and Mind Centre The University of Sydney 2050 Camperdown NSW Australia
8 shared publicationsDongang Wang
University of Sydney
7 shared publicationsGeng Zhan
Brain and Mind Centre The University of Sydney 2050 Camperdown NSW Australia
7 shared publicationsKain Kyle
Brain and Mind Centre The University of Sydney 2050 Camperdown NSW Australia
7 shared publications
Latest Publications
- Open Access
- 30 September 2026
- Published by Springer Nature in Journal of Neurology
- Vol. 273 (10) , 634
Spinal cord imaging in multiple sclerosis: from diagnosis to disease progression
Spinal cord (SC) involvement in multiple sclerosis (MS) is common and a major determinant of long-term disability in both relapsing and progressive disease phenotypes. Despite this, SC pathology remains under-assessed in both clinical practice and therapeutic trials, largely due to technical challenges inherent to spinal SC magnetic resonance imaging (MRI) conferred by the cord’s small anatomical size and sensitivity to physiological motion and susceptibility artefact. This review synthesises current understanding of SC involvement in MS, integrating advances in pathophysiology, imaging, and clinical application. We describe the characteristic MRI features of MS SC lesions and their key inflammatory and non-inflammatory differentials, and examine the role of SC imaging in diagnosis, particularly in the context of the 2024 revisions to the McDonald criteria. We highlight the prognostic importance of SC lesions and atrophy, including their association with progression independent of relapse activity, and review the limited and heterogeneous evidence regarding the effects of disease-modifying therapies on SC pathology. We further evaluate current MRI acquisition strategies and consensus guidelines, and discuss emerging quantitative and advanced imaging techniques that detect microstructural and biochemical abnormalities beyond conventional imaging. Together, these developments support a more systematic integration of SC imaging into MS clinical care and trial design.
- 30 September 2026
- Published by Springer Nature in Journal of Neurology
- Vol. 273 (10) , 634
Open AccessSpinal cord imaging in multiple sclerosis: from diagnosis to disease progression
Spinal cord (SC) involvement in multiple sclerosis (MS) is common and a major determinant of long-term disability in both relapsing and progressive disease phenotypes. Despite this, SC pathology remains under-assessed in both clinical practice and therapeutic trials, largely due to technical challenges inherent to spinal SC magnetic resonance imaging (MRI) conferred by the cord’s small anatomical size and sensitivity to physiological motion and susceptibility artefact. This review synthesises current understanding of SC involvement in MS, integrating advances in pathophysiology, imaging, and clinical application. We describe the characteristic MRI features of MS SC lesions and their key inflammatory and non-inflammatory differentials, and examine the role of SC imaging in diagnosis, particularly in the context of the 2024 revisions to the McDonald criteria. We highlight the prognostic importance of SC lesions and atrophy, including their association with progression independent of relapse activity, and review the limited and heterogeneous evidence regarding the effects of disease-modifying therapies on SC pathology. We further evaluate current MRI acquisition strategies and consensus guidelines, and discuss emerging quantitative and advanced imaging techniques that detect microstructural and biochemical abnormalities beyond conventional imaging. Together, these developments support a more systematic integration of SC imaging into MS clinical care and trial design.
- Open Access
- 1 July 2026
- Published by SAGE Publications in Multiple Sclerosis Journal
- Vol. 32 (11) , 1145-1154
Impact of multiple sclerosis disease-modifying therapies on chronic lesion tissue expansion
Background: Chronic lesion tissue expansion (CLTE) reflects slow, concentric growth of established multiple sclerosis (MS) lesions and is linked to central brain atrophy and disability progression. Whether existing MS disease-modifying therapies (DMTs) differentially influence this aspect of progressive MS biology remains unclear. Objectives: To compare the effect of current DMTs on CLTE in a multicentre, real-world MS cohort. Methods: We conducted a retrospective, observational study using linked clinical data from MSBase and the MSBase Imaging Repository. Data from patients aged ⩾18 years with ⩾3 longitudinal MRI scans, and 7 therapies with ⩾100 stable treatment epochs were included. Therapy effects on CLTE, new T2 lesions, and brain atrophy were assessed using epoch-based covariate-adjusted generalised estimating equation models, with fingolimod as a comparator. Result: The cohort included 564 patients contributing 1648 stable treatment epochs. After adjustment for demographic, clinical, and imaging covariates, B-cell depleting therapy was the only DMT associated with significantly lower CLTE (β = −4.03, p = 0.017) relative to fingolimod. CLTE was independently associated with age, baseline lesion volume, and centre effects. Conclusions: B-cell depletion is associated with reduced CLTE, a promising biomarker of progressive MS biology.
- 1 July 2026
- Published by SAGE Publications in Multiple Sclerosis Journal
- Vol. 32 (11) , 1145-1154
Open AccessImpact of multiple sclerosis disease-modifying therapies on chronic lesion tissue expansion
Background: Chronic lesion tissue expansion (CLTE) reflects slow, concentric growth of established multiple sclerosis (MS) lesions and is linked to central brain atrophy and disability progression. Whether existing MS disease-modifying therapies (DMTs) differentially influence this aspect of progressive MS biology remains unclear. Objectives: To compare the effect of current DMTs on CLTE in a multicentre, real-world MS cohort. Methods: We conducted a retrospective, observational study using linked clinical data from MSBase and the MSBase Imaging Repository. Data from patients aged ⩾18 years with ⩾3 longitudinal MRI scans, and 7 therapies with ⩾100 stable treatment epochs were included. Therapy effects on CLTE, new T2 lesions, and brain atrophy were assessed using epoch-based covariate-adjusted generalised estimating equation models, with fingolimod as a comparator. Result: The cohort included 564 patients contributing 1648 stable treatment epochs. After adjustment for demographic, clinical, and imaging covariates, B-cell depleting therapy was the only DMT associated with significantly lower CLTE (β = −4.03, p = 0.017) relative to fingolimod. CLTE was independently associated with age, baseline lesion volume, and centre effects. Conclusions: B-cell depletion is associated with reduced CLTE, a promising biomarker of progressive MS biology.
- Open Access
Rest2Visual: Predicting Visually Evoked fMRI from Resting-State Scans
Understanding how spontaneous brain activity relates to stimulus-driven neural responses is a fundamental challenge in cognitive neuroscience. While task-based functional magnetic resonance imaging (fMRI) captures localized stimulus-evoked brain activation, its acquisition is costly, time-consuming, and difficult to scale across populations. In contrast, resting-state fMRI (rs-fMRI) is task-free and abundant, but lacks direct interpretability. We introduce Rest2Visual, a conditional generative model that predicts visually evoked fMRI (ve-fMRI) from resting-state input and 2D visual stimuli. It follows a volumetric encoder--decoder design, where multiscale 3D features from rs-fMRI are modulated by image embeddings via adaptive normalization, enabling spatially accurate, stimulus-specific activation synthesis. To enable model training, we construct a large-scale triplet dataset from the Natural Scenes Dataset (NSD), aligning each rs-fMRI volume with stimulus images and their corresponding ve-fMRI activation maps. Quantitative evaluation shows that the predicted activations closely match ground truth across standard similarity and representational metrics, and support successful image reconstruction in downstream decoding. Notably, the predicted maps preserve subject-specific structure, demonstrating the model's capacity to generate individualized functional surrogates. Our results provide compelling evidence that individualized spontaneous neural activity can be transformed into stimulus-aligned representations, opening new avenues for scalable, task-free functional brain modeling.
Open AccessRest2Visual: Predicting Visually Evoked fMRI from Resting-State Scans
Understanding how spontaneous brain activity relates to stimulus-driven neural responses is a fundamental challenge in cognitive neuroscience. While task-based functional magnetic resonance imaging (fMRI) captures localized stimulus-evoked brain activation, its acquisition is costly, time-consuming, and difficult to scale across populations. In contrast, resting-state fMRI (rs-fMRI) is task-free and abundant, but lacks direct interpretability. We introduce Rest2Visual, a conditional generative model that predicts visually evoked fMRI (ve-fMRI) from resting-state input and 2D visual stimuli. It follows a volumetric encoder--decoder design, where multiscale 3D features from rs-fMRI are modulated by image embeddings via adaptive normalization, enabling spatially accurate, stimulus-specific activation synthesis. To enable model training, we construct a large-scale triplet dataset from the Natural Scenes Dataset (NSD), aligning each rs-fMRI volume with stimulus images and their corresponding ve-fMRI activation maps. Quantitative evaluation shows that the predicted activations closely match ground truth across standard similarity and representational metrics, and support successful image reconstruction in downstream decoding. Notably, the predicted maps preserve subject-specific structure, demonstrating the model's capacity to generate individualized functional surrogates. Our results provide compelling evidence that individualized spontaneous neural activity can be transformed into stimulus-aligned representations, opening new avenues for scalable, task-free functional brain modeling.
- Open Access
- 25 February 2025
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 12 (4) , 842-850
Longitudinal Trajectories of Digital Cognitive Biomarkers for Multiple Sclerosis
Background: Cognitive impairment is one of the most common and debilitating symptoms of relapsing–remitting multiple sclerosis (RRMS). Digital cognitive biomarkers require less time and resources and are rapidly gaining popularity in clinical settings. We examined the longitudinal trajectory of the iPad‐based Processing Speed Test (PST) and predictors of PST scores.Methods: We prospectively enrolled RRMS patients between 2017 and 2021 across six Australian MS centres. Longitudinal data was analysed with mixed effect modelling and latent class mixed models. We then examined whether latent class group membership predicted confirmed decrease in correct PST responses.Results: We recruited a total of 1093 participants, of which 724 had complete baseline data with a median follow up duration of 2 years. At a population level, PST trajectory was stable. A small practice effect was present up to the 4th visit. Age, baseline disability, T2 lesion volume, male sex and depression were associated with lower correct PST responses, whilst years of education and full/part‐time employment were associated with more correct PST responses.We identified four latent class trajectories of PST. The worst latent class was typified by low baseline PST and lack of a practice effect. Being in the worst latent class was associated with a greater hazard of time to sustained 5% decrease in PST (HR 2.84, 95% CI 1.16–6.94, p = 0.02).Conclusion: Worse baseline cognitive performance and lack of a practice effect predicted future cognitive decline in RRMS.
- 25 February 2025
- Published by Wiley in Annals of Clinical and Translational Neurology
- Vol. 12 (4) , 842-850
Open AccessLongitudinal Trajectories of Digital Cognitive Biomarkers for Multiple Sclerosis
Background: Cognitive impairment is one of the most common and debilitating symptoms of relapsing–remitting multiple sclerosis (RRMS). Digital cognitive biomarkers require less time and resources and are rapidly gaining popularity in clinical settings. We examined the longitudinal trajectory of the iPad‐based Processing Speed Test (PST) and predictors of PST scores.Methods: We prospectively enrolled RRMS patients between 2017 and 2021 across six Australian MS centres. Longitudinal data was analysed with mixed effect modelling and latent class mixed models. We then examined whether latent class group membership predicted confirmed decrease in correct PST responses.Results: We recruited a total of 1093 participants, of which 724 had complete baseline data with a median follow up duration of 2 years. At a population level, PST trajectory was stable. A small practice effect was present up to the 4th visit. Age, baseline disability, T2 lesion volume, male sex and depression were associated with lower correct PST responses, whilst years of education and full/part‐time employment were associated with more correct PST responses.We identified four latent class trajectories of PST. The worst latent class was typified by low baseline PST and lack of a practice effect. Being in the worst latent class was associated with a greater hazard of time to sustained 5% decrease in PST (HR 2.84, 95% CI 1.16–6.94, p = 0.02).Conclusion: Worse baseline cognitive performance and lack of a practice effect predicted future cognitive decline in RRMS.
- Open Access
- 3 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1
Fibre Population-guided Pre-training for 3D Spatial Super-Resolution on Multimodal Brain Diffusion MR Imaging
- 3 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1
Open AccessFibre Population-guided Pre-training for 3D Spatial Super-Resolution on Multimodal Brain Diffusion MR Imaging
- 23 October 2024
- Published by Springer Nature
- p. 742-752
Symmetry Awareness Encoded Deep Learning Framework for Brain Imaging Analysis
- 23 October 2024
- Published by Springer Nature
- p. 742-752
Symmetry Awareness Encoded Deep Learning Framework for Brain Imaging Analysis
- Open Access
Symmetry Awareness Encoded Deep Learning Framework for Brain Imaging Analysis
The heterogeneity of neurological conditions, ranging from structural anomalies to functional impairments, presents a significant challenge in medical imaging analysis tasks. Moreover, the limited availability of well-annotated datasets constrains the development of robust analysis models. Against this backdrop, this study introduces a novel approach leveraging the inherent anatomical symmetrical features of the human brain to enhance the subsequent detection and segmentation analysis for brain diseases. A novel Symmetry-Aware Cross-Attention (SACA) module is proposed to encode symmetrical features of left and right hemispheres, and a proxy task to detect symmetrical features as the Symmetry-Aware Head (SAH) is proposed, which guides the pretraining of the whole network on a vast 3D brain imaging dataset comprising both healthy and diseased brain images across various MRI and CT. Through meticulous experimentation on downstream tasks, including both classification and segmentation for brain diseases, our model demonstrates superior performance over state-of-the-art methodologies, particularly highlighting the significance of symmetry-aware learning. Our findings advocate for the effectiveness of incorporating symmetry awareness into pretraining and set a new benchmark for medical imaging analysis, promising significant strides toward accurate and efficient diagnostic processes. Code is available at https://github.com/bitMyron/sa-swin.
Open AccessSymmetry Awareness Encoded Deep Learning Framework for Brain Imaging Analysis
The heterogeneity of neurological conditions, ranging from structural anomalies to functional impairments, presents a significant challenge in medical imaging analysis tasks. Moreover, the limited availability of well-annotated datasets constrains the development of robust analysis models. Against this backdrop, this study introduces a novel approach leveraging the inherent anatomical symmetrical features of the human brain to enhance the subsequent detection and segmentation analysis for brain diseases. A novel Symmetry-Aware Cross-Attention (SACA) module is proposed to encode symmetrical features of left and right hemispheres, and a proxy task to detect symmetrical features as the Symmetry-Aware Head (SAH) is proposed, which guides the pretraining of the whole network on a vast 3D brain imaging dataset comprising both healthy and diseased brain images across various MRI and CT. Through meticulous experimentation on downstream tasks, including both classification and segmentation for brain diseases, our model demonstrates superior performance over state-of-the-art methodologies, particularly highlighting the significance of symmetry-aware learning. Our findings advocate for the effectiveness of incorporating symmetry awareness into pretraining and set a new benchmark for medical imaging analysis, promising significant strides toward accurate and efficient diagnostic processes. Code is available at https://github.com/bitMyron/sa-swin.
- Open Access
- 17 April 2024
- Published by Elsevier in Artificial Intelligence in Medicine
- Vol. 152, 102872
Improving multiple sclerosis lesion segmentation across clinical sites: A federated learning approach with noise-resilient training
- 17 April 2024
- Published by Elsevier in Artificial Intelligence in Medicine
- Vol. 152, 102872
Open AccessImproving multiple sclerosis lesion segmentation across clinical sites: A federated learning approach with noise-resilient training
- Open Access
How Much Data are Enough? Investigating Dataset Requirements for Patch-Based Brain MRI Segmentation Tasks
Training deep neural networks reliably requires access to large-scale datasets. However, obtaining such datasets can be challenging, especially in the context of neuroimaging analysis tasks, where the cost associated with image acquisition and annotation can be prohibitive. To mitigate both the time and financial costs associated with model development, a clear understanding of the amount of data required to train a satisfactory model is crucial. This paper focuses on an early stage phase of deep learning research, prior to model development, and proposes a strategic framework for estimating the amount of annotated data required to train patch-based segmentation networks. This framework includes the establishment of performance expectations using a novel Minor Boundary Adjustment for Threshold (MinBAT) method, and standardizing patch selection through the ROI-based Expanded Patch Selection (REPS) method. Our experiments demonstrate that tasks involving regions of interest (ROIs) with different sizes or shapes may yield variably acceptable Dice Similarity Coefficient (DSC) scores. By setting an acceptable DSC as the target, the required amount of training data can be estimated and even predicted as data accumulates. This approach could assist researchers and engineers in estimating the cost associated with data collection and annotation when defining a new segmentation task based on deep neural networks, ultimately contributing to their efficient translation to real-world applications.
Open AccessHow Much Data are Enough? Investigating Dataset Requirements for Patch-Based Brain MRI Segmentation Tasks
Training deep neural networks reliably requires access to large-scale datasets. However, obtaining such datasets can be challenging, especially in the context of neuroimaging analysis tasks, where the cost associated with image acquisition and annotation can be prohibitive. To mitigate both the time and financial costs associated with model development, a clear understanding of the amount of data required to train a satisfactory model is crucial. This paper focuses on an early stage phase of deep learning research, prior to model development, and proposes a strategic framework for estimating the amount of annotated data required to train patch-based segmentation networks. This framework includes the establishment of performance expectations using a novel Minor Boundary Adjustment for Threshold (MinBAT) method, and standardizing patch selection through the ROI-based Expanded Patch Selection (REPS) method. Our experiments demonstrate that tasks involving regions of interest (ROIs) with different sizes or shapes may yield variably acceptable Dice Similarity Coefficient (DSC) scores. By setting an acceptable DSC as the target, the required amount of training data can be estimated and even predicted as data accumulates. This approach could assist researchers and engineers in estimating the cost associated with data collection and annotation when defining a new segmentation task based on deep neural networks, ultimately contributing to their efficient translation to real-world applications.
- 1 December 2023
- Published by Springer Nature
- p. 357-366
Fed-CoT: Co-teachers for Federated Semi-supervised MS Lesion Segmentation
- 1 December 2023
- Published by Springer Nature
- p. 357-366
Fed-CoT: Co-teachers for Federated Semi-supervised MS Lesion Segmentation