About

Allan Hsiao is an Assistant Professor of Economics at Stanford University, working within the Department of Economics. His research focuses on questions in environmental and development economics, utilizing tools from empirical industrial organization and international trade. He explores issues related to environmental, resource, and energy economics, development economics, and international trade, contributing to the understanding of how economic policies and market dynamics impact environmental and developmental outcomes. Hsiao's academic work is centered on applying empirical methods to analyze complex economic phenomena in these fields. He is actively involved in teaching and mentoring students at both undergraduate and graduate levels, and he maintains a research agenda aimed at pushing forward the frontier of knowledge in economics related to environmental and development issues.

Research topics

• Political Science
• Medicine
• Law
• Medical physics
• Internal medicine
• Cardiology
• Radiology

Selected publications

• Deep Learning Automated Measurement of Shunt Severity with Estimation of Uncertainty in 4D Flow MRI

  Radiology Cardiothoracic Imaging · 2026-02-01

  articleOpen accessSenior author

  A fully automated deep learning–based system can be trained to measure systemic and pulmonary blood flow and shunt fraction using four-dimensional flow MRI with accuracy and reliability similar to that of expert physicians.

  PublisherOA PDFDOI

• Abstract 1090: Multiomic analysis of Taiwanese NSCLC: insights from a predominantly non-smoking cohort

  Cancer Research · 2025-04-21

  articleSenior author

  Abstract Introduction: We curated multiomic data including whole exome sequencing (WES) and high-resolution chest CT from patients diagnosed with NSCLC, and evaluated the potential for a multitask deep learning algorithm to predict tumor mutational status and WES signatures from CT. Materials and Methods: Taipei Veterans General Hospital (TVGH) initiated a multi-omic databank integrating clinical, radiological, pathological, and genomic data from NSCLC patients. A predefined pipeline identified INDEL and SNP mutations with high or moderate impact using Ensembl Variant Effect Predictor (VEP). PCA reduced dimensionality of the WES binary data, and t-SNE was used for clustering and visualization. Chest CT was performed using scanners with ≥ 20 detector rows, at end-inspiration with 120 kV, automated exposure control, and a 50 cm field of view. Patients were split 80% for training and 20% for validation. 3D CT image data centered on lesions were rotated, resampled and cropped. A 3D ResNet-18 network, pre-trained on Kinetics-400, was used for predicting common NSCLC mutations (EGFR, TP53), and the first and second principal component coefficients. Results: A total of 317 patients were included, with 74% non-smokers.(Table 1) Of 109 LDCT detected nodules, only 27 were from smokers. PCA of 12, 023 gene mutations identified five principal components. t-SNE visualized the clustered distribution of the top three genes with the highest weights: EGFR, TP53, and RBM10, along with tumor mutation burden (TMB) and EGFR co-mutation status. Performance of the deep learning algorithm in the validation set (n=63) showed: EGFR mutation prediction: AUROC 71%, PPV 71%; TP53 mutation prediction: AUROC 76%, NPV 94%, sensitivity 0.84, specificity 0.72.Pearson correlation for prediction of the coefficients of the genetic signature principal components from CT images were 0.37 and 0.47. Conclusion: Deep learning algorithms can infer genetic mutation signatures from CT radiologic features. Further optimization is needed. Citation Format: Hsu-Ching Huang, Samira Masoudi, Daniel Halmos, Chien-jung Huang, Yu-Chao Wang, Yi-Chen Yeh, Yung-Hung Luo, Han-Shui Hsu, Yuh-Min Chen, Chun-Ku Chen, Sandip Patel, Albert Hsiao. Multiomic analysis of Taiwanese NSCLC: insights from a predominantly non-smoking cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1090.

  PublisherDOI

• Pilot Multireader Study of Next Generation Chest X-ray Tomosynthesis for the Detection of Lung Cancer

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01

  articleSenior author

  Abstract RATIONALE Lung cancer screening (LCS) with annual low-dose computed tomography (LDCT) decreases mortality by identifying early-stage cancer but uptake is limited, especially in medically under-resourced areas. Chest x-ray tomosynthesis (CXRT) is an alternative imaging approach that can be manufactured at one sixth the cost of CT but has shown limited potential for lung cancer detection in prior studies. We investigated a novel, mobile CXRT device designed to overcome limitations of prior devices in a multireader pilot study of its performance in a composite population of patients undergoing LDCT, enriched with patients undergoing diagnostic chest CT for new lung cancer and suspicious lung nodules. METHODS: A composite population of patients undergoing screening and diagnostic chest CT between October 2023 and May 2024 were prospectively recruited to undergo imaging with next-generation CXRT device. Excluding subjects recruited for operator training, dose optimization, and radiologist training, 65 were included in a multireader study. Four subspecialty-trained cardiothoracic radiologists independently evaluated CXRT images for pulmonary lesions, blinded to patient data and CT scans. Two prespecified sets of analysis were performed to assess the accuracy of CXRT for detection of 1) biopsy-confirmed cancer and 2) pulmonary nodules on CT with ≥10 mm, ≥8 mm, and ≥6 mm long-axis size thresholds. RESULTS: Of the 65 subjects recruited for the multireader study, 20 had biopsy-proven malignancy on follow-up (median 262 days, maximum 343 days, minimum 146 days) while the remaining had benign exams. Readers showed substantial concordance (κ= 0.44-0.69) for detection of malignancy (N=20) with high sensitivity (0.85-1.0) and moderate specificity (0.56-0.60). There was moderate-high sensitivity for nodules, most of which were benign, across nodule size thresholds ≥10 mm (N=31, Sn 0.65-0.90, Sp 0.74-0.91), ≥8 mm (N=36, Sn 0.61-0.86, Sp 0.72-0.83), and ≥6 mm (N=41, Sn 0.66-0.85, Sp 0.67-0.79). CONCLUSIONS: This is the first human trial of a novel next-generation CXRT device, conducted in an at-risk pilot population enriched with participants with suspected or recurrent lung cancer. The high sensitivity of CXRT for detecting lung cancer signals its potential future use as a first-line LCS tool whereby positive screens would be referred for diagnostic CT. Notably, false negative rates with LDCT have been estimated to be 8-15% based on retrospective analysis of large LCS trials. The current multireader trial demonstrates the feasibility of CXRT for lung cancer detection and sets initial performance benchmarks to inform future studies and efforts at optimization.

  PublisherDOI

• Deep Learning R-Wave Detection for Electrocardiographic Gating in Cardiac MRI

  Radiology Cardiothoracic Imaging · 2025-11-06

  articleOpen accessSenior author

  A convolutional neural network for R-wave detection in MRI-acquired electrocardiograms achieved higher F1 scores and lower false-positive rates than standard algorithms, particularly at 3.0 T, supporting its potential to improve electrocardiographic gating and cardiac MRI scan quality.

  PublisherOA PDFDOI

• Artificial Intelligence in Cardiovascular MRI

  Journal of Thoracic Imaging · 2025-12-01 · 1 citations

  articleOpen accessSenior author

  In this review, we highlight how artificial intelligence, specifically deep learning, is reshaping every aspect of cardiovascular magnetic resonance imaging: from planning and acquisition to reconstruction, analysis, and clinical report generation. We first introduce core machine learning paradigms and concepts, then survey recent deep learning advances to automate and enhance multiple aspects of MRI. We highlight the range of recent advances to provide a conceptual understanding of how the field has rapidly evolved in the last 10 years, enabling improvements in acquisition speed, spatial resolution, suppression of artifacts, and correction for motion. Automation of postprocessing is providing us a deeper look into detailed analysis of regional cardiac function and measurement of hemodynamics, and a greater ability to automatically integrate interpretation with nonimaging clinical data to support prognostication and management. Advances in artificial intelligence will continue to shape our practice of clinical cardiovascular MRI to provide greater efficiency and enrich our ability to guide the management of patients with cardiovascular disease.

  PublisherOA PDFDOI

• Evaluation of a deep learning model for the automated prescription of aortic valve planes in CMR

  Journal of Cardiovascular Magnetic Resonance · 2025-01-01 · 1 citations

  articleOpen access
  PublisherDOI

• Quantitative 4D-Flow MR Imaging of Abdominopelvic Vasculature in Pelvic Venous Disorders

  medRxiv · 2025-08-24

  preprintOpen access

  Purpose: To determine whether quantitative 4-Dimensional (4D)-Flow MRI could reflect morphologic findings of pelvic venous disorder (PeVD). Methods: Abdominopelvic MRI with 4D-Flow acquired with 3T MRI from 2016-2022 were retrospectively reviewed for morphologic imaging findings: no venous abnormalities (NVA), left common iliac vein compression, left gonadal vein reflux, left renal vein (LRV) compression, and presence of pelvic collaterals. Using 4D-Flow MRI, blood flow was measured for vascular segments from the level of the suprarenal inferior vena cava (IVC) to the common iliac veins. Flow measurements at the LCIV and right common iliac vein (RCIV), the perihilar and juxta-caval renal veins were compared among participants with NVA, LCIV compression, LRV compression with and without LGV reflux, and with LGV reflux without LRV compression. Results: Sixty-six participants with LCIV compression, LRV compression, or LGV reflux displayed significantly diminished flow adjacent to the site of compression or reflux. Compared to participants with NVA, those with LCIV compression with pelvic collaterals showed increased RCIV flow and decreased LCIV flow (LCIV:RCIV flow ratio: 0.49±0.08 vs. NVA:0.92±0.05, p=0.0005). LCIV compression without pelvic collaterals did not significantly differ from NVA (LCIV:RCIV flow ratio: 0.80±0.09 vs. NVA, p>0.1). LRV compression with LGV reflux showed diminished juxta-caval LRV flow and similar perihilar LRV flow compared to LRV compression without LGV reflux ( p=0.03) or NVA (p=0.004) (juxta-caval:perihilar LRV flow ratio: with LGV reflux:0.39±0.17, without LGV reflux:1.3±0.19, NVA:1.3±0.13). Conclusions: Quantitative abdominopelvic 4D-Flow MRI measurements reflected flow diversion away from obstruction in LCIV or LRV compression, particularly in the setting of decompressing venous reservoirs.

  PublisherOA PDFDOI

• Olfactory Tract BOLD fMRI using Echo Planar Time Resolved Imaging

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: The use of fMRI for assessing olfactory function has been a significant challenge due to the increased blurring and geometric distortion caused by field inhomogeneity and T2* decay. Goal(s): We aim to minimize distortion and motion related artifacts that corrupt olfactory tract fMRI scans by utilizing an EPTI acquisition. Approach: We will acquire activation results using both EPTI and EPI for olfactory tract regions that suffer greatly from signal dropout and distortion for N=6 healthy, volunteers. Impact: Olfactory dysfunction is an early indicator of neurodegenerative diseases, such as Alzheimer's disease and Parkinson's disease.1-4 In this work, we utilize an acquisition technique designed to minimize fMRI artifacts to accurately assess olfactory function potentially allowing for earlier diagnosis.

  PublisherDOI

• Fast Non-contrast MR Angiography Using a Zigzag Centric k_y – k_z k-space Trajectory and Exponential Refocusing Flip Angles with Restoration of Longitudinal Magnetization

  Institutional Repositories DataBase (IRDB) · 2025-01-01

  articleOpen access

  Purpose: Fresh blood imaging (FBI) utilizes physiological blood signal differences between diastole and systole, causing a long acquisition time. The purpose of this study is to develop a fast FBI technique using a centric k_y – k_z k-space trajectory (cFBI) and an exponential refocusing flip angle (eFA) scheme with fast longitudinal restoration. Methods: This study was performed on 8 healthy subjects and 2 patients (peripheral artery disease and vascular disease) with informed consent, using a clinical 3-Tesla MRI scanner. A numeric simulation using extended phase graph (EPG) and phantom studies of eFA were carried out to investigate the restoration of longitudinal signal by lowering refocusing flip angles in later echoes. cFBI was then acquired on healthy subjects at the popliteal artery station to assess the effect of varying high/low flip ratios on the longitudinal restoration effects. In addition, trigger-delays of cFBI were optimized owing to the long acquisition window in zigzag centric k_y – k_z k-space trajectory. After optimizations, cFBI images were compared against standard FBI (sFBI) images in terms of scan time, motion artifacts, Nyquist N/2 artifacts, blurring, and overall image quality. We also performed two-way repeated measures analysis of variance. Results: cFBI with eFA achieved nearly a 50% scan time reduction compared to sFBI. The high/low flip angle of 180/2 degrees with lower refocusing pulses shows fast longitudinal restoration with the highest blood signals, yet also more sensitive to the background signals. Overall, 180/30 degrees images show reasonable blood signal recovery while minimizing the background signal artifacts. After the trigger delay optimization, maximum intensity projection image of cFBI after systole-diastole subtraction demonstrates less motion and N/2 artifacts than that of sFBI. Conclusion: Together with eFA for fast longitudinal signal restoration, the proposed cFBI technique achieved a 2-fold reduction in scan time and improved image quality without major artifacts.

  Publisher

• Automated prescription of left ventricular outflow tract and aortic valve views with a U-Net model

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: Cardiac magnetic resonance examinations require precise positioning of double-oblique planes. This task requires a high level of expertise, limiting the availability of CMR. Goal(s): This study aimed to evaluate a deep learning model to automate the prescription of left ventricular outflow tract and aortic valve views. Approach: A U-Net model was implemented and trained on manually annotated bSSFP Cine images to locate the aortic valve insertion points on 3CH and LVOT views. Results: Trained on >3000 images with 12x augmentation, the model achieved comparable accuracy to expert operators, with mean angle discrepancies of 13.4° for LVOT and 10.2° for AoV. Impact: The proposed model can replicate the prescription skills of expert CMR operators, for LVOT and AoV views. Ongoing work will assess its impact on clinical workflows and validate its performance in diverse patient populations to ensure robustness in practice.

  PublisherDOI

Recent grants

• Training Clinical Scientists in Radiological Imaging

  NIH · $2.9M · 2007–2029

Frequent coauthors

• Seth Kligerman

  University of California, San Diego

  57 shared

• Kyle Hasenstab

  36 shared

• Elizabeth A. Regan

  31 shared

• Shreyas Vasanawala

  Stanford University

  31 shared

• Tara Retson

  University of California, San Diego

  27 shared

• Barry J. Make

  National Jewish Health

  27 shared

• Brian Hurt

  24 shared

• Lewis D. Hahn

  23 shared

Education

• Fellowship, Cardiovascular Imaging, Radiology

  Stanford University

  2014

• Fellowship, Vascular and Interventional Radiology, Radiology

  Stanford University

  2013

• Residency, Diagnostic Radiology, Radiology

  Stanford University

  2012

• Internship, General Surgery

  Stanford University

  2008

• MD, School of Medicine

  University of California San Diego

  2007

• PhD with Specialization in Bioinformatics, Bioengineering

  University of California San Diego

  2005

• MS, Bioengineering

  University of California San Diego

  2003

• BS, Biology

  California Institute of Technology

  2000

• BS, Computer Science, Engineering and Applied Sciences

  California Institute of Technology

  2000