Ziqian Xie
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
Ziqian Xie has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Degui Zhi
McWilliams School of Biomedical Informatics, University of Texas Health Science Center at Houston, Houston, USA
11 shared publications - Laila Rasmy
University of Texas Health Science Center in Houston
4 shared publications - Xingzhong Zhao
McWilliams School of Biomedical Informatics, University of Texas Health Science Center at Houston, Houston, USA
3 shared publications - Bingyu Mao
McWilliams School of Biomedical Informatics, The University of Texas Health Science Center at Houston, 7000 Fannin, Suite 600, Houston, TX, United States, 1 713-500-3629, 1 713-500-3929
2 shared publications - Cui Tao
Department of Artificial Intelligence and Informatics, Mayo Clinic, Jacksonville, FL 32224, USA
2 shared publications - Ming-Hsiu Wu
The University of Texas Health Science Center at Houston
2 shared publications
Latest Publications
- Open Access
- 23 October 2026
- Published by Springer Nature in Communications Biology
- Vol. 9 (1) , 1394
A review of self-supervised imaging phenotypes for genome-wide association studies
Biobank-scale imaging-genomic cohorts enable genome-wide association studies (GWASs) of organ structure and function. Traditional approaches rely on expert-defined image-derived phenotypes (IDPs), which capture only a fraction of the biological variation in high-content images. Powered by artificial intelligence, self-supervised learning (SSL) learns representations directly from intrinsic image structures, creating scalable, data-driven traits. Here, we review the technical frameworks and challenges of SSL-based imaging genetics and their applications in brain MRI, cardiac MRI, and retinal imaging. Early studies suggest that these data-driven approaches can complement traditional IDPs and may expand the phenotypic space available for discovery of heritable imaging traits, although broader validation across cohorts, ancestries, scanners, and imaging protocols remains needed.
- 23 October 2026
- Published by Springer Nature in Communications Biology
- Vol. 9 (1) , 1394
Open AccessA review of self-supervised imaging phenotypes for genome-wide association studies
Biobank-scale imaging-genomic cohorts enable genome-wide association studies (GWASs) of organ structure and function. Traditional approaches rely on expert-defined image-derived phenotypes (IDPs), which capture only a fraction of the biological variation in high-content images. Powered by artificial intelligence, self-supervised learning (SSL) learns representations directly from intrinsic image structures, creating scalable, data-driven traits. Here, we review the technical frameworks and challenges of SSL-based imaging genetics and their applications in brain MRI, cardiac MRI, and retinal imaging. Early studies suggest that these data-driven approaches can complement traditional IDPs and may expand the phenotypic space available for discovery of heritable imaging traits, although broader validation across cohorts, ancestries, scanners, and imaging protocols remains needed.
- Open Access
- 9 June 2026
- Published by Springer Nature in Communications Biology
- Vol. 9 (1)
Vision transformer autoencoders captures local and non-local features in brain imaging to reveal novel genetic associations
Linking genetic variation to human brain structure is a key step toward understanding the biological basis of cognition and disease. Progress in this area, however, has been limited by a major challenge: imaging features are often predefined, restricting the discovery of novel associations. Here, we present a framework that applies a Vision Transformer (ViT)-based autoencoder to derive 128-dimensional representations from T1-weighted brain MRI scans of 6,130 UK Biobank participants, which we call unsupervised learning derived image phenotypes from ViT (ViT-UDIP). These ViT-UDIP phenotypes are used in genome-wide association studies (GWAS) of 22,867 UK Biobank participants to identify significant genetic variants, which were further aggregated into genetic loci. The ViT-based approach uncovers a total of 63 loci and out of which 24 were not detected by the CNN-based method. Importantly, feature interpretation reveals that the model captured local as well as non-local anatomical patterns such as left-right hemisphere symmetry within brain MRI data by leveraging its attention mechanism and positional embeddings. This ability of capturing non-local patterns distinguishes the ViT from the previous CNN model. Together, these results demonstrate the value of transformer-based architectures in discovering novel and robust imaging phenotypes for genetic discovery.
- 9 June 2026
- Published by Springer Nature in Communications Biology
- Vol. 9 (1)
Open AccessVision transformer autoencoders captures local and non-local features in brain imaging to reveal novel genetic associations
Linking genetic variation to human brain structure is a key step toward understanding the biological basis of cognition and disease. Progress in this area, however, has been limited by a major challenge: imaging features are often predefined, restricting the discovery of novel associations. Here, we present a framework that applies a Vision Transformer (ViT)-based autoencoder to derive 128-dimensional representations from T1-weighted brain MRI scans of 6,130 UK Biobank participants, which we call unsupervised learning derived image phenotypes from ViT (ViT-UDIP). These ViT-UDIP phenotypes are used in genome-wide association studies (GWAS) of 22,867 UK Biobank participants to identify significant genetic variants, which were further aggregated into genetic loci. The ViT-based approach uncovers a total of 63 loci and out of which 24 were not detected by the CNN-based method. Importantly, feature interpretation reveals that the model captured local as well as non-local anatomical patterns such as left-right hemisphere symmetry within brain MRI data by leveraging its attention mechanism and positional embeddings. This ability of capturing non-local patterns distinguishes the ViT from the previous CNN model. Together, these results demonstrate the value of transformer-based architectures in discovering novel and robust imaging phenotypes for genetic discovery.
- Open Access
- 3 June 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Genetic architecture of white matter microstructure captured by unsupervised deep representation learning of fractional anisotropy maps
Fractional anisotropy (FA) from diffusion MRI is a widely used marker of white matter (WM) integrity, but conventional FA-based genetic studies typically rely on tract- or atlas-defined averages that may obscure spatially distributed WM variation and limit genetic discovery. Here, we propose a deep learning framework, termed unsupervised deep representation of WM (UDR-WM), which uses voxel-wise FA maps to derive brain-wide unsupervised deep imaging phenotypes (UDIP-FA) without prior anatomical assumptions. Compared with traditional FA phenotypes, UDIP-FA shows greater sensitivity to aging and substantially higher SNP-based heritability. Multivariate GWAS identified 939 lead SNPs across 586 loci, mapping to 3,480 UDIP-FA-associated genes. These genes are enriched in glial cells, especially astrocytes and oligodendrocytes, and form disease-relevant modules in protein interaction and co-expression networks implicating myelination and axonal structure. UDIP-FA is genetically associated with multiple brain disorders, cognitive traits, and polygenic risk. Together, our results suggest that UDIP-FA provides a biologically meaningful view of white matter, complementing conventional ROI-based FA measures and offering a more refined way to study its genetic architecture.
- 3 June 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessGenetic architecture of white matter microstructure captured by unsupervised deep representation learning of fractional anisotropy maps
Fractional anisotropy (FA) from diffusion MRI is a widely used marker of white matter (WM) integrity, but conventional FA-based genetic studies typically rely on tract- or atlas-defined averages that may obscure spatially distributed WM variation and limit genetic discovery. Here, we propose a deep learning framework, termed unsupervised deep representation of WM (UDR-WM), which uses voxel-wise FA maps to derive brain-wide unsupervised deep imaging phenotypes (UDIP-FA) without prior anatomical assumptions. Compared with traditional FA phenotypes, UDIP-FA shows greater sensitivity to aging and substantially higher SNP-based heritability. Multivariate GWAS identified 939 lead SNPs across 586 loci, mapping to 3,480 UDIP-FA-associated genes. These genes are enriched in glial cells, especially astrocytes and oligodendrocytes, and form disease-relevant modules in protein interaction and co-expression networks implicating myelination and axonal structure. UDIP-FA is genetically associated with multiple brain disorders, cognitive traits, and polygenic risk. Together, our results suggest that UDIP-FA provides a biologically meaningful view of white matter, complementing conventional ROI-based FA measures and offering a more refined way to study its genetic architecture.
- Open Access
- 1 June 2026
- Published by JMIR Publications Inc. in JMIR AI
- Vol. 5, e81103
Improving Vancomycin Therapeutic Drug Monitoring With a Deep Learning–Based Two-Compartment Predictive Model: Development and Validation Study
Background: Vancomycin is a widely used antibiotic that requires therapeutic drug monitoring (TDM) for optimized individual dosage. A deep learning–based model, pharmacokinetic recurrent neural network–1 compartment model (PKRNN-1CM), has shown the advantage of leveraging time-series electronic health record data for individualized estimation of vancomycin pharmacokinetic (PK) parameters. While 1-compartment PK models are commonly used because of their simplicity and previous trough-based clinical practices for dose adjustment, the pre–deep learning literature suggests the superiority of 2-compartment models. Objective: This study introduces the pharmacokinetic recurrent neural network–2 compartment model (PKRNN-2CM), a novel deep learning–based model designed to improve vancomycin TDM by integrating a 2-compartment PK framework. Methods: PKRNN-2CM combines recurrent neural network–driven PK parameter estimation with a 2-compartment PK model to predict vancomycin concentration trajectories. Training on both simulated data and real-world electronic health record data allows for a comprehensive evaluation of its performance. Results: Experiments based on simulated data highlight PKRNN-2CM’s superiority over the simpler 1-compartment model, PKRNN-1CM, in predicting vancomycin concentration measurements (root mean square error 3.04 vs 4.50). Application to a real dataset from 5483 patients showcases significant improvement over PKRNN-1CM (root mean square error 5.55 vs 5.65; 2-sample 2-tailed unpaired t test; P =.01), with potential further gains expected with nontrough level measurements. Our simulation also indicates that PKRNN-2CM offers a better estimate of the average area under the concentration-time curve to minimum inhibitory concentration ratio, a more clinically relevant measure. Conclusions: PKRNN-2CM is an important improvement in vancomycin TDM, demonstrating enhanced accuracy and performance compared to the PKRNN-1CM model. This deep learning model holds potential for future individualized vancomycin TDM optimization and broader applications in diverse clinical scenarios.
- 1 June 2026
- Published by JMIR Publications Inc. in JMIR AI
- Vol. 5, e81103
Open AccessImproving Vancomycin Therapeutic Drug Monitoring With a Deep Learning–Based Two-Compartment Predictive Model: Development and Validation Study
Background: Vancomycin is a widely used antibiotic that requires therapeutic drug monitoring (TDM) for optimized individual dosage. A deep learning–based model, pharmacokinetic recurrent neural network–1 compartment model (PKRNN-1CM), has shown the advantage of leveraging time-series electronic health record data for individualized estimation of vancomycin pharmacokinetic (PK) parameters. While 1-compartment PK models are commonly used because of their simplicity and previous trough-based clinical practices for dose adjustment, the pre–deep learning literature suggests the superiority of 2-compartment models. Objective: This study introduces the pharmacokinetic recurrent neural network–2 compartment model (PKRNN-2CM), a novel deep learning–based model designed to improve vancomycin TDM by integrating a 2-compartment PK framework. Methods: PKRNN-2CM combines recurrent neural network–driven PK parameter estimation with a 2-compartment PK model to predict vancomycin concentration trajectories. Training on both simulated data and real-world electronic health record data allows for a comprehensive evaluation of its performance. Results: Experiments based on simulated data highlight PKRNN-2CM’s superiority over the simpler 1-compartment model, PKRNN-1CM, in predicting vancomycin concentration measurements (root mean square error 3.04 vs 4.50). Application to a real dataset from 5483 patients showcases significant improvement over PKRNN-1CM (root mean square error 5.55 vs 5.65; 2-sample 2-tailed unpaired t test; P =.01), with potential further gains expected with nontrough level measurements. Our simulation also indicates that PKRNN-2CM offers a better estimate of the average area under the concentration-time curve to minimum inhibitory concentration ratio, a more clinically relevant measure. Conclusions: PKRNN-2CM is an important improvement in vancomycin TDM, demonstrating enhanced accuracy and performance compared to the PKRNN-1CM model. This deep learning model holds potential for future individualized vancomycin TDM optimization and broader applications in diverse clinical scenarios.
- Open Access
Disease Risk Prediction Using Structured EHR Data: Can Generalist Large Language Models Match Specialized Clinical Foundation Models? A Comparative Evaluation with Fine-Tuning
Background: Electronic health records (EHRs) with clinical decision support tools are now ubiquitous in healthcare organizations. Clinical foundation models (CFMs) pretrained on large-scale, heterogeneous structured EHR data have emerged as a powerful approach to improve predictive performance and generalizability. Meanwhile, large language models (LLMs) pretrained on broad data sources are being applied to an expanding range of healthcare tasks. However, it remains unclear whether generalist LLMs can match specialized CFMs for disease risk prediction using structured clinical data. Methods: We compared CFMs (Med-BERT, CLMBR) against fine-tuned generalist LLMs (Mistral, LLaMA-2/3/3.1), a clinical LLM (Me-LLaMA), and LLM-generated embeddings paired with simple classifiers (using DeepSeek, Qwen3, and GPT-OSS) on two disease risk prediction tasks: heart failure risk among diabetic patients (DHF) and pancreatic cancer diagnosis (PaCa). Evaluations spanned multi-site EHR data, claims data, and an open-source single-institution benchmark (EHRSHOT). Performance was assessed using the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). Results: On larger EHR and claims cohorts (>30,000 patients), fine-tuned CFMs outperformed fine-tuned LLMs by a small but statistically significant margin (<1% AUROC). The clinical LLM performed comparably to generalist LLMs despite being smaller. On the open-source PaCa cohort (3,810 patients, 199 cases), LLMs achieved slightly higher AUROCs that were not statistically significant (LLaMA-3.1-70B 86.1% vs. Med-BERT 85.3%, p=0.27), but CFMs achieved significantly higher AUPRC (Med-BERT 55.9% vs. LLaMA-3.1-70B 41.1%, p=0.001). Notably, LLM-generated trajectory embeddings paired with logistic regression or a simple MLP, without any LLM fine-tuning, achieved the best overall performance, with AUROC exceeding 90% (Qwen3) and AUPRC reaching 66% (GPT-OSS 20B). Conclusion: LLM-generated embeddings with lightweight classifiers outperformed both fine-tuned CFMs and fine-tuned LLMs on AUROC and AUPRC. While these results demonstrate the potential of generalist models to match or surpass specialized CFMs, their substantially greater computational cost and variable AUPRC performance in the fine-tuning setting warrant caution. We provide a reproducible evaluation framework and codebase to support continued benchmarking.
Open AccessDisease Risk Prediction Using Structured EHR Data: Can Generalist Large Language Models Match Specialized Clinical Foundation Models? A Comparative Evaluation with Fine-Tuning
Background: Electronic health records (EHRs) with clinical decision support tools are now ubiquitous in healthcare organizations. Clinical foundation models (CFMs) pretrained on large-scale, heterogeneous structured EHR data have emerged as a powerful approach to improve predictive performance and generalizability. Meanwhile, large language models (LLMs) pretrained on broad data sources are being applied to an expanding range of healthcare tasks. However, it remains unclear whether generalist LLMs can match specialized CFMs for disease risk prediction using structured clinical data. Methods: We compared CFMs (Med-BERT, CLMBR) against fine-tuned generalist LLMs (Mistral, LLaMA-2/3/3.1), a clinical LLM (Me-LLaMA), and LLM-generated embeddings paired with simple classifiers (using DeepSeek, Qwen3, and GPT-OSS) on two disease risk prediction tasks: heart failure risk among diabetic patients (DHF) and pancreatic cancer diagnosis (PaCa). Evaluations spanned multi-site EHR data, claims data, and an open-source single-institution benchmark (EHRSHOT). Performance was assessed using the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). Results: On larger EHR and claims cohorts (>30,000 patients), fine-tuned CFMs outperformed fine-tuned LLMs by a small but statistically significant margin (<1% AUROC). The clinical LLM performed comparably to generalist LLMs despite being smaller. On the open-source PaCa cohort (3,810 patients, 199 cases), LLMs achieved slightly higher AUROCs that were not statistically significant (LLaMA-3.1-70B 86.1% vs. Med-BERT 85.3%, p=0.27), but CFMs achieved significantly higher AUPRC (Med-BERT 55.9% vs. LLaMA-3.1-70B 41.1%, p=0.001). Notably, LLM-generated trajectory embeddings paired with logistic regression or a simple MLP, without any LLM fine-tuning, achieved the best overall performance, with AUROC exceeding 90% (Qwen3) and AUPRC reaching 66% (GPT-OSS 20B). Conclusion: LLM-generated embeddings with lightweight classifiers outperformed both fine-tuned CFMs and fine-tuned LLMs on AUROC and AUPRC. While these results demonstrate the potential of generalist models to match or surpass specialized CFMs, their substantially greater computational cost and variable AUPRC performance in the fine-tuning setting warrant caution. We provide a reproducible evaluation framework and codebase to support continued benchmarking.
- Open Access
- 24 April 2026
- Published by Springer Nature in Research Square
Learning heritable multimodal brain representation via contrastive learning
Magnetic resonance imaging (MRI)-derived phenotypes (IDP) has enabled the discovery of numerous genomic loci associated with brain structure and function. However, most existing IDPs and learned representations are derived from a single imaging modality, missing complementary information across modalities and potentially limiting the scope of genetic discovery. Here, we introduce a multimodal contrastive learning framework to derive heritable representations from paired T1- and T2-weighted MRIs. Unlike single-modality reconstruction-based models, we designed a momentum-based contrastive learning framework. The derived heritable representations show improved correlation with traditional IDPs and better predict age and brain disorders. Notably, genome-wide association studies (GWAS) of the learned representations reveal a substantially higher overlap of genetic loci across modalities, indicating improved alignment of their underlying genetic architecture. Analysis of the GWAS loci identified shared protein and drug targets, yielding meaningful biological insights. Overall, our framework learns shared representations across brain imaging modalities that exhibit anatomical and genetic coherence.
- 24 April 2026
- Published by Springer Nature in Research Square
Open AccessLearning heritable multimodal brain representation via contrastive learning
Magnetic resonance imaging (MRI)-derived phenotypes (IDP) has enabled the discovery of numerous genomic loci associated with brain structure and function. However, most existing IDPs and learned representations are derived from a single imaging modality, missing complementary information across modalities and potentially limiting the scope of genetic discovery. Here, we introduce a multimodal contrastive learning framework to derive heritable representations from paired T1- and T2-weighted MRIs. Unlike single-modality reconstruction-based models, we designed a momentum-based contrastive learning framework. The derived heritable representations show improved correlation with traditional IDPs and better predict age and brain disorders. Notably, genome-wide association studies (GWAS) of the learned representations reveal a substantially higher overlap of genetic loci across modalities, indicating improved alignment of their underlying genetic architecture. Analysis of the GWAS loci identified shared protein and drug targets, yielding meaningful biological insights. Overall, our framework learns shared representations across brain imaging modalities that exhibit anatomical and genetic coherence.
- Open Access
Towards Precision Protein-Ligand Affinity Prediction Benchmark: A Complete and Modification-Aware DAVIS Dataset
Advancements in AI for science unlocks capabilities for critical drug discovery tasks such as protein-ligand binding affinity prediction. However, current models overfit to existing oversimplified datasets that does not represent naturally occurring and biologically relevant proteins with modifications. In this work, we curate a complete and modification-aware version of the widely used DAVIS dataset by incorporating 4,032 kinase-ligand pairs involving substitutions, insertions, deletions, and phosphorylation events. This enriched dataset enables benchmarking of predictive models under biologically realistic conditions. Based on this new dataset, we propose three benchmark settings-Augmented Dataset Prediction, Wild-Type to Modification Generalization, and Few-Shot Modification Generalization-designed to assess model robustness in the presence of protein modifications. Through extensive evaluation of both docking-free and docking-based methods, we find that docking-based model generalize better in zero-shot settings. In contrast, docking-free models tend to overfit to wild-type proteins and struggle with unseen modifications but show notable improvement when fine-tuned on a small set of modified examples. We anticipate that the curated dataset and benchmarks offer a valuable foundation for developing models that better generalize to protein modifications, ultimately advancing precision medicine in drug discovery. The benchmark is available at: https://github.com/ZhiGroup/DAVIS-complete
Open AccessTowards Precision Protein-Ligand Affinity Prediction Benchmark: A Complete and Modification-Aware DAVIS Dataset
Advancements in AI for science unlocks capabilities for critical drug discovery tasks such as protein-ligand binding affinity prediction. However, current models overfit to existing oversimplified datasets that does not represent naturally occurring and biologically relevant proteins with modifications. In this work, we curate a complete and modification-aware version of the widely used DAVIS dataset by incorporating 4,032 kinase-ligand pairs involving substitutions, insertions, deletions, and phosphorylation events. This enriched dataset enables benchmarking of predictive models under biologically realistic conditions. Based on this new dataset, we propose three benchmark settings-Augmented Dataset Prediction, Wild-Type to Modification Generalization, and Few-Shot Modification Generalization-designed to assess model robustness in the presence of protein modifications. Through extensive evaluation of both docking-free and docking-based methods, we find that docking-based model generalize better in zero-shot settings. In contrast, docking-free models tend to overfit to wild-type proteins and struggle with unseen modifications but show notable improvement when fine-tuned on a small set of modified examples. We anticipate that the curated dataset and benchmarks offer a valuable foundation for developing models that better generalize to protein modifications, ultimately advancing precision medicine in drug discovery. The benchmark is available at: https://github.com/ZhiGroup/DAVIS-complete
- Open Access
- 7 April 2025
- Published by Springer Nature in Communications Chemistry
- Vol. 8 (1) , 1-9
A Folding-Docking-Affinity framework for protein-ligand binding affinity prediction
Accurate protein-ligand binding affinity prediction is crucial in drug discovery. Existing methods are predominately docking-free, without explicitly considering atom-level interaction between proteins and ligands in scenarios where crystallized protein-ligand binding conformations are unavailable. Now, with breakthroughs in deep learning AI-based protein folding and binding conformation prediction, can we improve binding affinity prediction? This study introduces a framework, Folding-Docking-Affinity (FDA), which folds proteins, determines protein-ligand binding conformations, and predicts binding affinities from three-dimensional protein-ligand binding structures. Our experimental results indicate that FDA performs comparably to state-of-the-art docking-free methods. We anticipate that our proposed framework serves as a starting point for integrating binding structures for more accurate binding affinity prediction.
- 7 April 2025
- Published by Springer Nature in Communications Chemistry
- Vol. 8 (1) , 1-9
Open AccessA Folding-Docking-Affinity framework for protein-ligand binding affinity prediction
Accurate protein-ligand binding affinity prediction is crucial in drug discovery. Existing methods are predominately docking-free, without explicitly considering atom-level interaction between proteins and ligands in scenarios where crystallized protein-ligand binding conformations are unavailable. Now, with breakthroughs in deep learning AI-based protein folding and binding conformation prediction, can we improve binding affinity prediction? This study introduces a framework, Folding-Docking-Affinity (FDA), which folds proteins, determines protein-ligand binding conformations, and predicts binding affinities from three-dimensional protein-ligand binding structures. Our experimental results indicate that FDA performs comparably to state-of-the-art docking-free methods. We anticipate that our proposed framework serves as a starting point for integrating binding structures for more accurate binding affinity prediction.
- Open Access
- 6 February 2024
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 13 (3) , e029900
Dynamic Prognosis Prediction for Patients on DAPT After Drug‐Eluting Stent Implantation: Model Development and Validation
Background: The rapid evolution of artificial intelligence (AI) in conjunction with recent updates in dual antiplatelet therapy (DAPT) management guidelines emphasizes the necessity for innovative models to predict ischemic or bleeding events after drug‐eluting stent implantation. Leveraging AI for dynamic prediction has the potential to revolutionize risk stratification and provide personalized decision support for DAPT management. Methods and Results: We developed and validated a new AI‐based pipeline using retrospective data of drug‐eluting stent‐treated patients, sourced from the Cerner Health Facts data set (n=98 236) and Optum's de‐identified Clinformatics Data Mart Database (n=9978). The 36 months following drug‐eluting stent implantation were designated as our primary forecasting interval, further segmented into 6 sequential prediction windows. We evaluated 5 distinct AI algorithms for their precision in predicting ischemic and bleeding risks. Model discriminative accuracy was assessed using the area under the receiver operating characteristic curve, among other metrics. The weighted light gradient boosting machine stood out as the preeminent model, thus earning its place as our AI‐DAPT model. The AI‐DAPT demonstrated peak accuracy in the 30 to 36 months window, charting an area under the receiver operating characteristic curve of 90% [95% CI, 88%–92%] for ischemia and 84% [95% CI, 82%–87%] for bleeding predictions. Conclusions: Our AI‐DAPT excels in formulating iterative, refined dynamic predictions by assimilating ongoing updates from patients' clinical profiles, holding value as a novel smart clinical tool to facilitate optimal DAPT duration management with high accuracy and adaptability.
- 6 February 2024
- Published by Wolters Kluwer Health in Journal of the American Heart Association
- Vol. 13 (3) , e029900
Open AccessDynamic Prognosis Prediction for Patients on DAPT After Drug‐Eluting Stent Implantation: Model Development and Validation
Background: The rapid evolution of artificial intelligence (AI) in conjunction with recent updates in dual antiplatelet therapy (DAPT) management guidelines emphasizes the necessity for innovative models to predict ischemic or bleeding events after drug‐eluting stent implantation. Leveraging AI for dynamic prediction has the potential to revolutionize risk stratification and provide personalized decision support for DAPT management. Methods and Results: We developed and validated a new AI‐based pipeline using retrospective data of drug‐eluting stent‐treated patients, sourced from the Cerner Health Facts data set (n=98 236) and Optum's de‐identified Clinformatics Data Mart Database (n=9978). The 36 months following drug‐eluting stent implantation were designated as our primary forecasting interval, further segmented into 6 sequential prediction windows. We evaluated 5 distinct AI algorithms for their precision in predicting ischemic and bleeding risks. Model discriminative accuracy was assessed using the area under the receiver operating characteristic curve, among other metrics. The weighted light gradient boosting machine stood out as the preeminent model, thus earning its place as our AI‐DAPT model. The AI‐DAPT demonstrated peak accuracy in the 30 to 36 months window, charting an area under the receiver operating characteristic curve of 90% [95% CI, 88%–92%] for ischemia and 84% [95% CI, 82%–87%] for bleeding predictions. Conclusions: Our AI‐DAPT excels in formulating iterative, refined dynamic predictions by assimilating ongoing updates from patients' clinical profiles, holding value as a novel smart clinical tool to facilitate optimal DAPT duration management with high accuracy and adaptability.
- Open Access
Genetic InfoMax: Exploring Mutual Information Maximization in High-Dimensional Imaging Genetics Studies
Genome-wide association studies (GWAS) are used to identify relationships between genetic variations and specific traits. When applied to high-dimensional medical imaging data, a key step is to extract lower-dimensional, yet informative representations of the data as traits. Representation learning for imaging genetics is largely under-explored due to the unique challenges posed by GWAS in comparison to typical visual representation learning. In this study, we tackle this problem from the mutual information (MI) perspective by identifying key limitations of existing methods. We introduce a trans-modal learning framework Genetic InfoMax (GIM), including a regularized MI estimator and a novel genetics-informed transformer to address the specific challenges of GWAS. We evaluate GIM on human brain 3D MRI data and establish standardized evaluation protocols to compare it to existing approaches. Our results demonstrate the effectiveness of GIM and a significantly improved performance on GWAS.
Open AccessGenetic InfoMax: Exploring Mutual Information Maximization in High-Dimensional Imaging Genetics Studies
Genome-wide association studies (GWAS) are used to identify relationships between genetic variations and specific traits. When applied to high-dimensional medical imaging data, a key step is to extract lower-dimensional, yet informative representations of the data as traits. Representation learning for imaging genetics is largely under-explored due to the unique challenges posed by GWAS in comparison to typical visual representation learning. In this study, we tackle this problem from the mutual information (MI) perspective by identifying key limitations of existing methods. We introduce a trans-modal learning framework Genetic InfoMax (GIM), including a regularized MI estimator and a novel genetics-informed transformer to address the specific challenges of GWAS. We evaluate GIM on human brain 3D MRI data and establish standardized evaluation protocols to compare it to existing approaches. Our results demonstrate the effectiveness of GIM and a significantly improved performance on GWAS.