About

Shu-mei Shih is the Irving and Jean Stone Professor of Humanities at UCLA, serving in the Department of Comparative Literature, Asian Languages and Cultures, and Asian American Studies. She was the inaugural holder of the Edward W. Said Professorship in Comparative Literature from 2019 to 2022 and has served as the President of the American Comparative Literature Association from 2021 to 2022. Her scholarly work has significantly contributed to the development of Sinophone Studies, with her book 'Visuality and Identity: Sinophone Articulations across the Pacific' (2007) being credited with inaugurating this field. The book has been translated into multiple languages, including Mandarin Chinese and Korean. Her research interests encompass comparative modernism, transnationalism, critical race studies, critical theory, and Taiwan studies, among others. She has held visiting professorships at numerous universities worldwide and has delivered keynote and plenary lectures globally. Her work has been translated into several languages, and she serves on numerous editorial boards. At UCLA, she co-directed the Mellon Postdoctoral Fellowship in the Humanities, supported by a substantial grant, and continues to work on monographs related to Sinophone divergences, empire, race, and comparative literature in a relational world.

Research topics

• Medicine
• Radiology
• Nuclear medicine
• Mathematics
• Computer Science
• Artificial Intelligence
• Computer vision
• Anatomy
• Statistics

Selected publications

• Breath-hold, Four-Dimensional, Ferumoxytol-Enhanced MUSIC: Initial Results in Adult Congenital Heart Disease

  Journal of Cardiovascular Magnetic Resonance · 2026-01-01

  articleOpen access
  PublisherDOI

• Cross-cohort federated learning for pediatric abdominal adipose tissue segmentation and quantification using free-breathing 3D MRI

  Radiology Advances · 2026-01-01

  articleOpen access

  Abstract Background Pediatric abdominal visceral and subcutaneous adipose tissue (VAT, SAT) quantified on magnetic resonance imaging (MRI) can assess risk for metabolic diseases. However, the complex structure of VAT in children and the lack of sufficient MRI datasets pose challenges for developing automated segmentation methods. Purpose To achieve accurate and rapid automated segmentation of pediatric abdominal VAT and SAT on motion-robust free-breathing (FB) 3D Dixon MRI by developing a cross-cohort federated learning (FL) framework that leverages adult datasets. Materials and Methods 3D FB-MRI datasets were prospectively acquired in children 6-18 years old (single center, 2 scanners; 2016-2023) and used to train 3D neural network models for segmenting abdominal VAT and SAT. The FL model was trained across the pediatric cohort and a separate adult cohort (5 centers, 7 scanners) without requiring direct data sharing. Segmentation performance of the FL model was assessed by Dice scores with respect to references and compared with standalone local training and joint training with full data access. Quantification of VAT and SAT volume and proton-density fat fraction (PDFF) was compared against references using intraclass correlation coefficients (ICCs) and Bland-Altman analysis. Differences between training approaches were analyzed using the Kruskal-Wallis test followed by paired Wilcoxon signed-rank tests. Results The FL model, trained and tested with 134 children (mean age, 13.3 years ± 2.7 [standard deviation]; 71 males) and 920 adults (50.4 years ± 14.0; 677 females), achieved mean Dice scores of 91.09% (VAT) and 95.55% (SAT), outperforming standalone training (VAT: P < .001) and performing comparably to joint training (VAT: P = .21). Volume quantification demonstrated strong agreement (VAT: ICC = 0.99, SAT: ICC = 1.00). PDFF quantification showed small mean differences (VAT: 0.21%, SAT: −1.19%). Inference time was <3 seconds for each subject. Conclusion The proposed FL framework achieved accurate and rapid automated segmentation and quantification of pediatric abdominal VAT and SAT on 3D FB-MRI. Abbreviations AT = adipose tissue, BMI = body mass index, FB = free-breathing, FL = federated learning, FN = false negative, FP = false positive, ICC = intraclass correlation coefficient, JT = jointly trained, MASLD = metabolic dysfunction-associated steatotic liver disease, MD = mean difference, PDFF = proton-density fat fraction, SAT = subcutaneous adipose tissue, SFF = signal fat fraction, VAT = visceral adipose tissue Summary The proposed cross-cohort federated learning framework enables accurate and rapid automated segmentation of pediatric abdominal adipose tissue using free-breathing 3D MRI, safeguarding data privacy across cohorts while maintaining high performance. Key Results In a cohort of 6- to 18-year-old children, the proposed federated learning (FL) framework achieved mean Dice scores of 91.09% and 95.55% for segmenting abdominal visceral and subcutaneous adipose tissue (VAT, SAT) on free-breathing 3D Dixon MRI. FL demonstrated accurate adipose tissue volume quantification, with intraclass correlation coefficients of 0.99-1.00 compared to references. FL achieved accurate VAT and SAT proton-density fat fraction quantification with mean differences of 0.21% and −1.19% versus references.

  PublisherDOI

• Spiral cine displacement encoding with stimulated echoes (DENSE) MRI at 0.55 T

  Journal of Cardiovascular Magnetic Resonance · 2026-01-01

  articleOpen access1st authorCorresponding
  PublisherDOI

• CoDe: A Self-Supervised Consistency Model Framework for MRI Denoising

  2026-04-08

  article

  Magnetic Resonance Imaging (MRI) is great for visualizing soft tissues and quantitative assessment of tissue properties, but its acquisitions are often limited by signal noise originating from the patient's thermal and physiological processes, motion, and hardware limitations/imperfections. We propose CoDe, a self-supervised consistency model (CM) framework that achieves efficient one-step MRI denoising. Our approach consists of two components: a noise estimation model that predicts the noise level of the input image and a CM performing one-step denoising to recover clean images. Additionally, we introduce a Random Matrix Theory (RMT)-based regularization that leverages physical noise statistics to enhance structural fidelity. Experiments on public brain and in-house prostate diffusion MRI datasets demonstrate that CoDe achieves superior image quality compared with existing methods while maintaining fast, one-step inference. Code is available at: https://github.com/JimmyHou123/CoDe-model.

  PublisherDOI

• Enhanced <scp>2D</scp> Spiral Cine <scp>DENSE MRI</scp> Using Low‐Rank Denoising for Improved Apparent Signal‐to‐Noise Ratio, Spatial Resolution, Efficiency, Accuracy, and Accessibility

  Magnetic Resonance in Medicine · 2026-03-09

  articleOpen access1st author

  ABSTRACT Purpose To develop a denoising technique for displacement encoding with stimulated echoes (DENSE) MRI that improves spatial resolution, efficiency, and accuracy, and enhances accessibility by implementing DENSE MRI at 0.55 T. Methods We developed a low‐rank denoising technique, which leverages multidimensional spiral cine DENSE MRI data for empirical noise estimation via Monte Carlo simulation combined with automatic noise suppression. Thirty‐six subjects (16 healthy, 20 with heart disease) were scanned at 3 T with breath‐hold standard‐resolution 2D cine DENSE (2.8 × 2.8 mm 2 ) in a short‐axis slice of the heart. In 10 healthy subjects, high‐resolution DENSE with 1.2 × 1.2 mm 2 was acquired. Apparent signal‐to‐noise ratio (SNR), phase SNR, scan efficiency (SNR per heartbeat per unit voxel size), and standard deviation of segmental circumferential myocardial strain (E cc ) were compared with Wilcoxon signed‐rank tests ( p &lt; 0.05 considered significant). High‐resolution and standard‐resolution DENSE results were compared using Bland–Altman analysis. Lastly, we scanned seven healthy subjects at 0.55 T and 3 T, and compared E cc results. Results Apparent magnitude SNR, phase SNR and scan efficiency were significantly improved after denoising in both standard‐resolution and high‐resolution DENSE (all p &lt; 0.01). Bland–Altman analysis showed denoised high‐resolution DENSE E cc had smaller mean differences (non‐denoised: 0.028 vs. denoised: 0.009) and narrower limits of agreement (non‐denoised: [−0.072, 0.127] vs. denoised: [−0.048, 0.065]), indicating improved accuracy. Strain measurements from denoised DENSE at 0.55 T showed good agreement with those from 3 T, demonstrating feasibility of DENSE MRI at 0.55 T. Conclusion Our proposed denoising technique may allow DENSE MRI with improved spatial resolution, efficiency, and accuracy, and enhanced accessibility at 0.55 T.

  PublisherDOI

• Role of Low-Field Magnetic Resonance Imaging in the Lungs: Opportunities and Challenges

  Journal of Computer Assisted Tomography · 2026-05-15

  articleCorresponding

  Clinical utilization of lung MRI has not kept pace with MRI of other body regions. This is due to a combination of technical and perceptual factors related to lung imaging. With increasing access to and application of low-field MRI, there are unique opportunities for lung imaging. This review will first explain the role of conventional lung MRI, next delineate the opportunities afforded by low-field lung MRI and clinical knowledge to date, and third, address remaining challenges and future directions.

  PublisherDOI

• ICD Artifacts Reduction for Radiotherapy of Ventricular Tachycardia Using 3D Wideband Late Gadolinium-Enhanced MRI: A Preliminary Study in Healthy Participants and Patient

  International Journal of Radiation Oncology*Biology*Physics · 2025-09-01

  article
  PublisherDOI

• Accelerated 2D Cine DENSE MRI using Golden-Angle Spiral Acquisition and Compressed Sensing Reconstruction

  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

  article1st authorCorresponding

  Motivation: To accelerate the acquisition time of 2D spiral cine DENSE MRI for strain imaging. Goal(s): To develop a golden-angle-ordered spiral cine DENSE sequence and a compressed sensing MRI reconstruction method and investigate the performance retrospectively undersampled datasets. Approach: We used a compressed sensing model with constraints in the spatial domain and also along the cardiac phase domain for image reconstruction. We compared the image quality and the strain measurements from data acquired with linear ordering and golden-angle ordering. Results: We demonstrated 1.5-fold accelerated 2D cine DENSE MRI. The proposed work may facilitate efficient DENSE data acquisition for strain imaging in clinical settings. Impact: The proposed work may facilitate efficient DENSE data acquisition for strain imaging in clinical settings.

  PublisherDOI

• Assessing the Reliability of SE-EPI MRE for Liver Stiffness Estimation Under Free-Breathing Instruction Compared to Breath-Holding

  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: Magnetic resonance elastography (MRE) provides non-invasive liver fibrosis assessment, but breath-holding (BH) requirements limit its use in certain patients. Goal(s): To compare liver stiffness (LS) estimates from SE-EPI MRE acquired under BH and free-breathing (FB) instruction. Approach: Forty-three adults underwent both BH and FB SE-EPI MRE. We assessed the Structural Similarity Index Measure (SSIM) among MRE images and measurable region-of-interest (ROI) size. Agreement between LS estimates was evaluated using Bland-Altman and regression analyses. Results: While FB had lower SSIM values and smaller ROIs, mean LS did not significantly differ from BH SE-EPI, with minimal bias (0.00kPa) and an intra-class correlation coefficient of 0.988. Impact: Free-breathing SE-EPI MRE without explicit motion compensation provides reliable mean LS estimates comparable to breath-holding scans, enabling rapid liver stiffness assessment in patients with difficulty performing breath-holding.

  PublisherDOI

• Volumetric thermometry in moving tissues using stack‐of‐radial <scp>MRI</scp> and an image‐navigated multi‐baseline proton resonance frequency shift method

  Magnetic Resonance in Medicine · 2025-09-10 · 1 citations

  articleOpen access

  PURPOSE: To develop and evaluate a volumetric proton resonance frequency shift (PRF)-based thermometry method for monitoring thermal ablation in moving tissues. METHODS: A golden-angle-ordered 3D stack-of-radial MRI sequence was combined with an image-navigated multi-baseline (iNAV-MB) PRF method to reconstruct motion-compensated 3D temperature maps with high spatiotemporal resolution and volumetric coverage. Two radial MRI reconstruction techniques, k-space weighted image contrast filter (KWIC) and golden-angle radial sparse parallel (GRASP) MRI, were implemented and compared within a sliding window reconstruction framework. Ex vivo motion phantom experiments were performed with high-intensity focused ultrasound ablation to evaluate motion tracking and temperature accuracy using input motion waveforms and temperature probe readings as references. In vivo non-heating swine experiments were conducted to assess temperature mapping stability in 3D liver regions of interest. RESULTS: , and effective temporal resolution of 0.98 s/volume. In ex vivo high-intensity focused ultrasound experiments, motion tracking achieved correlation coefficients of 0.951 and 0.973, and temporal mean absolute errors were 1.80°C and 1.44°C using KWIC and GRASP, respectively. In vivo experiments demonstrated improvements in voxel-wise temperature temporal SD from a median of 8.03 to 3.85°C (KWIC) and from a median of 7.23 to 2.37°C (GRASP), compared to single-baseline PRF. CONCLUSION: The proposed stack-of-radial iNAV-MB volumetric PRF thermometry framework can reliably track respiratory motion and map ablation-associated temperature change. This framework has the potential to improve MRI-guided thermal ablation in moving tissues.

  PublisherOA PDFDOI

Frequent coauthors

• Holden H. Wu

  44 shared

• Xiaodong Zhong

  Resonance Research (United States)

  21 shared

• Sevgi Gökçe Kafalı

  19 shared

• Kara L. Calkins

  University of California, Los Angeles

  15 shared

• Shahnaz Ghahremani

  10 shared

• Vibhas Deshpande

  9 shared

• Bradley D. Bolster

  7 shared

• Zhaohuan Zhang

  Jinan University

  7 shared

Awards & honors

• Fulbright-Hays Foundation Fellowship
• American Philosophical Society Fellowship
• American Council of Learned Societies Fellowship
• Mellon Foundation Grant for Cultures in Transnational Perspe…