Jiahong Ouyang

Longitudinal MRIs enable noninvasive tracking of the gradual deterioration of brain structure and function caused by neurological diseases and environmental influences over time. We first proposed a CNN+RNN based framework with a longitudinal pooling layer and a consistency loss function to achieve longitudinally and clinically consistent diagnosis prediction. [paper][code]

However, the longitudinal analysis is complicated by the complex covariance structure characterizing a mixture of time-varying and static effects across visits. It generally requires a large number of ground-truth labels, which are often missing or expensive to obtain. To mitigate this challenge, LSSL, a longitudinal self-supervised learning method, was proposed by our group to model the brain aging by a linear direction in the latent space.

But it has two main drawbacks: (1) a linear direction might not have the capacity to model complex longitudinal changes; (2) it can only model diseases/disorders that can be considered accelerated or daccelarated aging. To resolve the first probelm, we further proposed to regularize a smooth trajectory vector field in the latent space. [paper][code] To solve the second problem, we proposed to disentangle a disease progression from the normal brain aging. [paper] [code]

Multi-modal MR images are widely used in neuroimaging applications to provide complementary information about the brain structures. Multi-modal deep learning analysis can be benefit from explicitly disentangling anatomical (shape) and modality (appearance) representations. We proposed a novel margin-based similarity loss and a modified conditional convolution to enable efficient fully disentanglement. We further introduced a fusion function to obtain a set of modality-invariant features for downstream tasks. [paper] [code] [talk]

Positron emission tomography (PET) is a widely used imaging technique in many clinical applications including tumor detection and neurological disorder diagnosis. To obtain high-quality images, the amount of injected radiotracer in current protocols leads to the risk of radiation exposure in scanned subjects. Decreasing this injected dose will lead to low quality images that can further affecting the disease diagnosis. We propose one of the pioneer work that synthesizes high quality and pathologically accurate amyloid PET images solely from 1% ultra-low-dose PET using cGAN with task-specific perceptual loss. [paper] We further attempt to push the boundary to zero-dose PET, i.e., using widely available and non-invasive MRI to synthesize PET images. We propose a U-Net based model with symmetry-aware spatial-wise attention module and channel attention-wise module to accurately capture the abnormality. [abstract]

Predicting Hypoperfusion Lesion and Target Mismatch in Stroke from Diffusion-weighted MRI Using Deep Learning
Y. Yu, S. Christensen, J. Ouyang, F. Scalzo, D. Liebeskind, M. Lansberg, G. Albers, G. Zaharchuk
Radiology, 2022

Self-supervised Learning of Neighborhood Embedding for Longitudinal MRI
J. Ouyang, Q. Zhao, E. Adeli, G. Zaharchuk, K. Pohl
Medical Image Analysis, 2022

Automated Detection of Arterial Landmarks and Vascular Occlusions in Patients with Acute Stroke Receiving Digital Subtraction Angiography Using Deep Learning
J. Khankari, Y. Yu, J. Ouyang, R. Hussein, H. Do, J. Heit, G. Zaharchuk
Journal of NeuroInterventional Surgery, 2022

Disentangling Normal Aging from Severity of Disease via Weak Supervision on Longitudinal MRI
J. Ouyang, Q. Zhao, E. Adeli, G. Zaharchuk, K. Pohl
IEEE Transactions on Medical Imaging, 2022 [code]

Deep Learning Evaluation of Biomarkers from Echocardiogram Videos
J. Hughes, N. Yuan, B. He, J. Ouyang, ..., D. Ouyang, J. Zou
EBioMedicine, 2021

Tissue at Risk and Ischemic Core Estimation Using Deep Learning in Acute Stroke
Y. Yu, Y. Xie, T. Thamm, E. Gong, J. Ouyang, S. Christensen, M. Marks, M. Lansberg, G. Albers, G. Zaharchuk
American Journal of Neuroradiology, 2021 (Annual Lucien Levy Best Research Article Award)

Longitudinal Pooling & Consistency Regularization to Model Disease Progression from MRIs
J. Ouyang, Q. Zhao, E. Sullivan, A. Pfefferbaum, S. Tapert, E. Adeli, K. Pohl
Journal of Biomedical and Health Informatics (JBHI), 2020 [code]

Generalization of Deep Learning Models for Ultra-low-count Amyloid PET/MRI Using Transfer Learning
K. Chen, M. Schurer, J. Ouyang, M. Koran, G. Davidzon, E. Mormino, S. Tieport, K. Hoffmann, O. Sabri, G. Zaharchuk, H. Barthel
European Journal of Nuclear Medicine and Molecular Imaging, 2020

Use of Deep Learning to Predict Final Ischemic Stroke Lesions From Initial Magnetic Resonance Imaging
Y. Yu, Y. Xie, T. Thamm, E. Gong, J. Ouyang, C. Huang, S. Christensen, M. Marks, M. Lansberg, G. Albers, G. Zaharchuk
Journal of the American Medical Association (JAMA) Network Open, 2020

Accurate Tissue Interface Segmentation via Adversarial Pre-segmentation of Anterior Segment OCT Images
J. Ouyang, T. Mathai, J. Galeotti
Biomedical Optics Express, 2019 [patent]

Ultra-low-dose Amyloid PET Reconstruction by Generative Adversarial Network with Feature Matching and Task-specific Perceptual Loss
J. Ouyang, E. Gong, K. Chen, J. Pauly, G. Zaharchuk
Medical Physics, 2019

Weakly Semi-Supervised Detection in Lung Ultrasound Videos
J. Ouyang, L. Chen, G. Li, ..., B. Raju, J. Kruecker, A. Chen
Information Processing for Medical Imaging (IPMI) 2023

Self-Supervised Longitudinal Neighbourhood Embedding
J. Ouyang, Q. Zhao, E. Adeli, E. Sullivan, A. Pfefferbaum, G. Zaharchuk, K. Pohl
Medical Image Computing and Computer Assisted Intervention (MICCAI) 2021 (Oral Presentation, Young Scientist Award Shortlisted, 10 over 1630 papers) [code]

Representation Disentanglement for Multi-modal MR Analysis
J. Ouyang, E. Adeli, K. Pohl, Q. Zhao, G. Zaharchuk
Information Processing for Medical Imaging (IPMI) 2021 (Oral Presentation) [code][talk]

Ultra-low-dose 18F-FDG Brain PET/MR Denoising Using Deep Learning and Multi-contrast Information
J. Xu, E. Gong, J. Ouyang, J. Pauly, G. Zaharchuk, S. Han
Medical Imaging 2020: Image Processing (Oral Presentation)

Task-GAN: Improving Generative Adversarial Network for Image Reconstruction
J. Ouyang, G. Wang, E. Gong, K. Chen, J. Pauly, G. Zaharchuk
MICCAI Machine Learning for Medical Image Reconstruction Workshop (MLMIR) 2019

Fingerprint Pose Estimation based on Faster R-CNN
J. Ouyang, J. Feng, J. Lu, Z. Guo, J. Zhou
International Joint Conference on Biometrics (IJCB) 2017 (Oral Presentation, Best Student Paper Award Runner-Up)

System and Method for Removing Noise and/or Artifacts from an OCT Image Using a Generative Adversarial Network
J. Ouyang, T. Mathai, J. Galeotti

• NeuroImage
  
• Medical Image Analysis (MEDIA)
  
• IEEE Transactions on Medical Imaging (TMI)
  
• Scientific Reports
  
• Pattern Recognition
  
• Knowledge-Based Systems
  
• Journal of Biomedical and Health Informatics (JBHI)
  
• IEEE Transactions on Radiation and Plasma Medical Sciences (TRPMS)
  
• Artifical Intelligence in Medicine
  
• Computer Methods and Programs in Biomedicine
  
• Frontiers in Neuroimaging
  
• IEEE Access
  

• European Conference on Computer Vision (ECCV), 2022
  
• International Conference on Computer Vision (ICCV), 2023
  
• IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022
  
• Medical Image Computing and Computer Assisted Intervention (MICCAI), 2020, 2021, 2022, 2023
  
• International Workshop on PRedictive Intelligence in MEdicine (PRIME), 2022
  
• Medical Applications with Disentanglements, 2022
  
• IEEE Winter Conference on Applications of Computer Vision (WACV), 2019