Zixuan Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Zixuan Liu has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Linda G. Shapiro
Paul G. Allen School of Computer Science and Engineering, University of Washington, Seattle, WA, USA
2 shared publications - Sheng Wang
University of Washington
2 shared publications - Aaron Y. Lee
John F. Hardesty M.D. Department of Ophthalmology and Visual Sciences, Washington University in St Louis School of Medicine, St. Louis MO
1 shared publication Addie Woicik
Paul G. Allen School of Computer Science and Engineering, University of Washington, Seattle, WA, USA
1 shared publication- Belinda Zeng
Amazon SP, Seattle, WA, USA
1 shared publication Catherine A. Cukras
F. Hoffmann-La Roche Ltd, Basel, Switzerland
1 shared publication
Latest Publications
- 24 April 2026
- Published by Springer Nature in Nature Biomedical Engineering
A three-dimensional multi-modal foundation model for optical coherence tomography
Vision loss caused by retinal diseases remains a leading global cause of disability. Optical coherence tomography (OCT) is an imaging technique that is used for diagnosing retinal diseases. Computational models can use OCT images for various diagnostic and prognostic tasks, but most existing approaches fail to fully leverage the rich three-dimensional (3D) structure of OCT data and lack the capability to integrate other retinal imaging modalities into the analysis. Here, to address these limitations, we present OCTCube-M, a 3D OCT-based multi-modal framework designed for the integrated analysis of 3D OCT and 2D en face (EF) images. OCTCube-M exploits COEP, an effective multi-modal contrastive learning method, to integrate OCT with other retinal imaging modalities, such as fundus autofluorescence imaging and infrared retinal imaging (IR). Using the OCTCube-M framework, we developed three models: OCTCube (uni-modal), OCTCube-IR (bi-modal) and OCTCube-EF (tri-modal). OCTCube, a 3D foundation model pre-trained on 26,605 3D OCT volumes comprising 1.62 million 2D OCT slices, achieved state-of-the-art performance in predicting 8 retinal diseases while demonstrating robust generalizability across cohorts, devices and modalities. OCTCube-IR extends OCTCube by incorporating 26,685 pairs of OCT and IR images, enabling accurate cross-modality retrieval and joint analysis of these two modalities. OCTCube-EF, trained on over 4 million 2D OCT slices and 400 thousand EF retinal images, excels in predicting the growth rate of geographic atrophy across datasets collected from 6 multi-centre clinical trials across 23 countries. Collectively, OCTCube-M is a 3D multi-modal foundation model framework for integrating OCT and other retinal imaging modalities. It demonstrated substantial advancements in cross-site, cross-device, cross-modality and systemic disease prediction, while offering substantial utility in geographic atrophy clinical trials.
- 24 April 2026
- Published by Springer Nature in Nature Biomedical Engineering
A three-dimensional multi-modal foundation model for optical coherence tomography
Vision loss caused by retinal diseases remains a leading global cause of disability. Optical coherence tomography (OCT) is an imaging technique that is used for diagnosing retinal diseases. Computational models can use OCT images for various diagnostic and prognostic tasks, but most existing approaches fail to fully leverage the rich three-dimensional (3D) structure of OCT data and lack the capability to integrate other retinal imaging modalities into the analysis. Here, to address these limitations, we present OCTCube-M, a 3D OCT-based multi-modal framework designed for the integrated analysis of 3D OCT and 2D en face (EF) images. OCTCube-M exploits COEP, an effective multi-modal contrastive learning method, to integrate OCT with other retinal imaging modalities, such as fundus autofluorescence imaging and infrared retinal imaging (IR). Using the OCTCube-M framework, we developed three models: OCTCube (uni-modal), OCTCube-IR (bi-modal) and OCTCube-EF (tri-modal). OCTCube, a 3D foundation model pre-trained on 26,605 3D OCT volumes comprising 1.62 million 2D OCT slices, achieved state-of-the-art performance in predicting 8 retinal diseases while demonstrating robust generalizability across cohorts, devices and modalities. OCTCube-IR extends OCTCube by incorporating 26,685 pairs of OCT and IR images, enabling accurate cross-modality retrieval and joint analysis of these two modalities. OCTCube-EF, trained on over 4 million 2D OCT slices and 400 thousand EF retinal images, excels in predicting the growth rate of geographic atrophy across datasets collected from 6 multi-centre clinical trials across 23 countries. Collectively, OCTCube-M is a 3D multi-modal foundation model framework for integrating OCT and other retinal imaging modalities. It demonstrated substantial advancements in cross-site, cross-device, cross-modality and systemic disease prediction, while offering substantial utility in geographic atrophy clinical trials.
- Open Access
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1523-1533
ForeSeer: Product Aspect Forecasting Using Temporal Graph Embedding
Developing text mining approaches to mine aspects from customer reviews has been well-studied due to its importance in understanding customer needs and product attributes. In contrast, it remains unclear how to predict the future emerging aspects of a new product that currently has little review information. This task, which we named product aspect forecasting, is critical for recommending new products, but also challenging because of the missing reviews. Here, we propose ForeSeer, a novel textual mining and product embedding approach progressively trained on temporal product graphs for this novel product aspect forecasting task. ForeSeer transfers reviews from similar products on a large product graph and exploits these reviews to predict aspects that might emerge in future reviews. A key novelty of our method is to jointly provide review, product, and aspect embeddings that are both time-sensitive and less affected by extremely imbalanced aspect frequencies. We evaluated ForeSeer on a real-world product review system containing 11,536,382 reviews and 11,000 products over 3 years. We observe that ForeSeer substantially outperformed existing approaches with at least 49.1% AUPRC improvement under the real setting where aspect associations are not given. ForeSeer further improves future link prediction on the product graph and the review aspect association prediction. Collectively, Foreseer offers a novel framework for review forecasting by effectively integrating review text, product network, and temporal information, opening up new avenues for online shopping recommendation and e-commerce applications.
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1523-1533
Open AccessForeSeer: Product Aspect Forecasting Using Temporal Graph Embedding
Developing text mining approaches to mine aspects from customer reviews has been well-studied due to its importance in understanding customer needs and product attributes. In contrast, it remains unclear how to predict the future emerging aspects of a new product that currently has little review information. This task, which we named product aspect forecasting, is critical for recommending new products, but also challenging because of the missing reviews. Here, we propose ForeSeer, a novel textual mining and product embedding approach progressively trained on temporal product graphs for this novel product aspect forecasting task. ForeSeer transfers reviews from similar products on a large product graph and exploits these reviews to predict aspects that might emerge in future reviews. A key novelty of our method is to jointly provide review, product, and aspect embeddings that are both time-sensitive and less affected by extremely imbalanced aspect frequencies. We evaluated ForeSeer on a real-world product review system containing 11,536,382 reviews and 11,000 products over 3 years. We observe that ForeSeer substantially outperformed existing approaches with at least 49.1% AUPRC improvement under the real setting where aspect associations are not given. ForeSeer further improves future link prediction on the product graph and the review aspect association prediction. Collectively, Foreseer offers a novel framework for review forecasting by effectively integrating review text, product network, and temporal information, opening up new avenues for online shopping recommendation and e-commerce applications.
- Open Access
Brain-Aware Replacements for Supervised Contrastive Learning in Detection of Alzheimer's Disease
We propose a novel framework for Alzheimer's disease (AD) detection using brain MRIs. The framework starts with a data augmentation method called Brain-Aware Replacements (BAR), which leverages a standard brain parcellation to replace medically-relevant 3D brain regions in an anchor MRI from a randomly picked MRI to create synthetic samples. Ground truth "hard" labels are also linearly mixed depending on the replacement ratio in order to create "soft" labels. BAR produces a great variety of realistic-looking synthetic MRIs with higher local variability compared to other mix-based methods, such as CutMix. On top of BAR, we propose using a soft-label-capable supervised contrastive loss, aiming to learn the relative similarity of representations that reflect how mixed are the synthetic MRIs using our soft labels. This way, we do not fully exhaust the entropic capacity of our hard labels, since we only use them to create soft labels and synthetic MRIs through BAR. We show that a model pre-trained using our framework can be further fine-tuned with a cross-entropy loss using the hard labels that were used to create the synthetic samples. We validated the performance of our framework in a binary AD detection task against both from-scratch supervised training and state-of-the-art self-supervised training plus fine-tuning approaches. Then we evaluated BAR's individual performance compared to another mix-based method CutMix by integrating it within our framework. We show that our framework yields superior results in both precision and recall for the AD detection task.
Open AccessBrain-Aware Replacements for Supervised Contrastive Learning in Detection of Alzheimer's Disease
We propose a novel framework for Alzheimer's disease (AD) detection using brain MRIs. The framework starts with a data augmentation method called Brain-Aware Replacements (BAR), which leverages a standard brain parcellation to replace medically-relevant 3D brain regions in an anchor MRI from a randomly picked MRI to create synthetic samples. Ground truth "hard" labels are also linearly mixed depending on the replacement ratio in order to create "soft" labels. BAR produces a great variety of realistic-looking synthetic MRIs with higher local variability compared to other mix-based methods, such as CutMix. On top of BAR, we propose using a soft-label-capable supervised contrastive loss, aiming to learn the relative similarity of representations that reflect how mixed are the synthetic MRIs using our soft labels. This way, we do not fully exhaust the entropic capacity of our hard labels, since we only use them to create soft labels and synthetic MRIs through BAR. We show that a model pre-trained using our framework can be further fine-tuned with a cross-entropy loss using the hard labels that were used to create the synthetic samples. We validated the performance of our framework in a binary AD detection task against both from-scratch supervised training and state-of-the-art self-supervised training plus fine-tuning approaches. Then we evaluated BAR's individual performance compared to another mix-based method CutMix by integrating it within our framework. We show that our framework yields superior results in both precision and recall for the AD detection task.
- 1 July 2021
- Published by Elsevier in Medical Image Analysis
- Vol. 71, 102051
Longitudinal self-supervised learning
- 1 July 2021
- Published by Elsevier in Medical Image Analysis
- Vol. 71, 102051
Longitudinal self-supervised learning
- Open Access
- 1 January 2021
- Published by MIT Press
Imprecise Fusion Operators for Collective Learning
A decentralised collective learning problem is investigated in which a population of agents attempts to learn the true state of the world based on direct evidence from the environment and belief fusion carried out during local interactions between agents. A parameterised fusion operator is introduced that returns beliefs of varying levels of imprecision. This is used to explore the effect of fusion imprecision on learning performance in a series of agent-based simulations. In general, the results suggest that imprecise fusion operators are optimal when the frequency of fusion is high relative to the frequency with which evidence is obtained from the environment.
- 1 January 2021
- Published by MIT Press
Open AccessImprecise Fusion Operators for Collective Learning
A decentralised collective learning problem is investigated in which a population of agents attempts to learn the true state of the world based on direct evidence from the environment and belief fusion carried out during local interactions between agents. A parameterised fusion operator is introduced that returns beliefs of varying levels of imprecision. This is used to explore the effect of fusion imprecision on learning performance in a series of agent-based simulations. In general, the results suggest that imprecise fusion operators are optimal when the frequency of fusion is high relative to the frequency with which evidence is obtained from the environment.