About

Shuai Huang is a Professor in the Department of Industrial & Systems Engineering at the University of Washington. His research focuses on Biomedical Informatics and Medical Education, with additional expertise in Engineering and health, Engineering and manufacturing, Applied Statistics & Production Systems, and AI in Healthcare. He is also an Adjunct Professor in Engineering and health, and Engineering and manufacturing. His work involves applying machine learning and statistical methods to healthcare and medical decision-making, contributing to advancements in biomedical informatics and medical education.

Research topics

• Computer Science
• Artificial Intelligence
• Machine Learning
• Natural Language Processing
• Mathematics
• Transport engineering
• Medicine
• Statistics
• Engineering
• Medical emergency
• Human–computer interaction

Selected publications

• The ‘Roller Coaster Effect’ in Clear Aligner Premolar Extraction Cases: Posterior Open Bite and Cephalometric Risk Factors

  Orthodontics and Craniofacial Research · 2026-05-09

  article

  OBJECTIVES: This study aimed to investigate the occurrence of the roller coaster effect (RCE) in premolar extraction cases treated with clear aligner therapy (CAT), and to identify cephalometric risk factors. MATERIALS AND METHODS: Sixty five patients treated with CAT after four first premolar extractions were retrospectively analysed, including 34 with mild crowding (Mi) and 31 with moderate-severe crowding (Mo-Se). Posterior functional cusp distances (FCDs) were measured at the second premolar, mandibular first molar mesiobuccal (MB) and distobuccal (DB) cusps, and second molar on pre- and post-treatment digital models using a standardised 3D coordinate system. RESULTS: As mesial molar movement was minimal in this cohort, post-treatment FCDs increased significantly in the Mi group at all posterior sites, with the greatest changes at the first molars (MB: 1.64 ± 1.39 mm; DB: 1.48 ± 1.55 mm; both p < 0.001), while the Mo-Se group showed an increase only at the first molar MB cusp (p < 0.001). Comparisons indicated greater FCDs increases in the Mi group than in the Mo-Se group (all p < 0.05). Correlation analysis revealed positive associations between greater upper incisor proclination (U1-NA, U1-SN) and posterior FCDs increases. The regression analysis further identified a smaller ANB angle as an independent risk factor for RCE. CONCLUSIONS: Posterior FCDs served as a simple and clinically applicable indicator of RCE, most pronounced at the first molars and in mild crowding cases. Smaller ANB angles and increased maxillary incisor proclination were associated with higher RCE susceptibility, emphasising prudent case selection in extraction-based clear aligner therapy.

  PublisherDOI

• Alpinetin alleviates sepsis-induced cardiomyopathy by suppressing serine/threonine protein phosphatase 1γ-mediated ROS accumulation

  European Journal of Pharmacology · 2026-05-12

  articleSenior authorCorresponding
  PublisherDOI

• Melatonin regulates proline metabolism by modulating CmDREB1A/E to enhance cold tolerance in cantaloupe fruit

  Plant Physiology and Biochemistry · 2025-05-14 · 6 citations

  article1st author
  PublisherDOI

• A comprehensive comparison on clustering methods for multi-slice spatially resolved transcriptomics data analysis

  Briefings in Bioinformatics · 2025-08-22

  articleOpen access

  Spatial transcriptomics (ST) data, by providing spatial information, enable simultaneous analysis of gene expression distributions and their spatial patterns within tissue. Clustering or spatial domain detection represents an essential methodology for ST data, facilitating the exploration of spatial organizations with shared gene expression or histological characteristics. Traditionally, clustering algorithms for ST have focused on individual tissue sections. However, the emergence of numerous contiguous tissue sections derived from the same or similar tissue specimens within or across individuals has led to the development of multi-slice clustering methods. In this study, we assess seven single-slice and four multi-slice clustering methods on two simulated datasets and four real datasets. Additionally, we investigate the effectiveness of preprocessing techniques, including spatial coordinate alignment (e.g. PASTE) and gene expression batch effect removal (e.g. Harmony), on clustering performance. Our study provides a comprehensive comparison of clustering methods for multi-slice ST data, serving as a practical guide for method selection in various scenarios.

  PublisherDOI

• Call for Papers— <i>INFORMS Journal on Data Science</i> Virtual Special Issue on Artificial Intelligence and Data Science for Healthcare

  INFORMS Journal on Data Science · 2025-10-10

  articleSenior author
  PublisherDOI

• Stock overpricing, underwriting fees, and stock price crash risk

  Finance research letters · 2025-06-15

  article
  PublisherDOI

• Research on the Evaluation of Ecosystem Cultural Service Quality in Qu County Congren Valley Forest Park from the Perspective of Tourist Perception

  Proceedings of Business and Economic Studies · 2025-07-14

  articleOpen accessSenior author

  This study focuses on the ecosystem cultural service quality of Qu County Congren Valley Forest Park from the perspective of tourist perception. Using the Importance-Performance Analysis (IPA) questionnaire survey method and SPSS data analysis techniques, we systematically evaluate tourists’ cognitive differences and improvement paths regarding the cultural service value of the scenic area. Based on the nonmaterial characteristics of ecosystem cultural services, combined with the unique Congren culture and natural landscape resources of Congren Valley, we designed a five-dimensional scale including natural landscape and ecological protection, cultural display and interpretation services, cultural activity participation and experience, infrastructure and supporting services, and safety management. This covers tourists’ evaluations of the importance of elements such as cultural displays, interpretation systems, interactive activities, and facility support, as well as their actual satisfaction feedback. Through descriptive statistical analysis, reliability and validity testing, factor analysis, and IPA matrix analysis, we reveal the core contradictions and improvement directions perceived by tourists. The study found that the convenience of facilities such as signage, rest areas, toilets, roads, and the cleanliness of the scenic area are key areas for improvement. Additionally, different age groups perceive differences in the cultural service quality of the Congren Valley Forest Park ecosystem. The study concludes that tourists have a high level of concern for the convenience of scenic infrastructure and sanitary environment. Improving these facilities can help increase tourist satisfaction and the overall service quality of the scenic area. Simultaneously, meeting the needs of segmented markets and constructing a three-in-one service system of “deep excavation of cultural symbols–digital storytelling–immersive scenes” is recommended.

  PublisherOA PDFDOI

• A loosely coupled serial digital image correlation method based on deep learning

  Measurement · 2025-05-06 · 2 citations

  articleSenior author
  PublisherDOI

• Orthogonal design-based investigation of dose, time point, and treatment course on the toxicity-efficacy transition of triptolide in collagen-induced arthritis mice

  Frontiers in Pharmacology · 2025-08-12 · 1 citations

  articleOpen access1st authorCorresponding

  In recent years, the potential application of Tripterygium wilfordii Hook f. (TWHF) in the treatment of rheumatoid arthritis (RA) has garnered increasing attention in both academic research and clinical practice. However, the effective use of rheumatoid arthritis is limited in clinical practice by its severe toxic side effects. We conducted a comprehensive analysis of the effects of triptolide dose, treatment course, and time point on the clinical efficacy and safety of treating experimental arthritis. This work employed an orthogonal design grouping and three-factor, three-level dose, treatment course, and time point modeling of collagen-induced arthritis (CIA) in female C57BL/6 mice. Using smears from exfoliated mouse vaginal cells, the estrous cycle was observed. Mice blood was tested by the enzyme linked immunosorbent assay (ELISA) for ovarian hormones such as estradiol (E2) and follicle stimulating hormone (FSH), as well as inflammatory markers such as interleukin-6 (IL-6) and interleukin-17A (IL-17A). In CIA mice, triptolide changed serum E2 and FSH levels, the estrous cycle, arthritis scores, IL-6, and IL-17A levels. The inhibitory effect of triptolide on IL-17A was significantly influenced by the time point of administration. For triptolide therapy of CIA mice, a high benefit-low risk dosage schedule is 150 μg/kg/d −23:00–6 weeks. Therefore, in clinical applications, optimizing the TWHF dosing regimen (including dose, time point, and treatment course) may help to minimize ovarian toxicity while retaining therapeutic efficacy.

  PublisherOA PDFDOI

• SmartHome-Bench: A Comprehensive Benchmark for Video Anomaly Detection in Smart Homes Using Multi-Modal Large Language Models

  ArXiv.org · 2025-06-15

  preprintOpen accessSenior author

  Video anomaly detection (VAD) is essential for enhancing safety and security by identifying unusual events across different environments. Existing VAD benchmarks, however, are primarily designed for general-purpose scenarios, neglecting the specific characteristics of smart home applications. To bridge this gap, we introduce SmartHome-Bench, the first comprehensive benchmark specially designed for evaluating VAD in smart home scenarios, focusing on the capabilities of multi-modal large language models (MLLMs). Our newly proposed benchmark consists of 1,203 videos recorded by smart home cameras, organized according to a novel anomaly taxonomy that includes seven categories, such as Wildlife, Senior Care, and Baby Monitoring. Each video is meticulously annotated with anomaly tags, detailed descriptions, and reasoning. We further investigate adaptation methods for MLLMs in VAD, assessing state-of-the-art closed-source and open-source models with various prompting techniques. Results reveal significant limitations in the current models' ability to detect video anomalies accurately. To address these limitations, we introduce the Taxonomy-Driven Reflective LLM Chain (TRLC), a new LLM chaining framework that achieves a notable 11.62% improvement in detection accuracy. The benchmark dataset and code are publicly available at https://github.com/Xinyi-0724/SmartHome-Bench-LLM.

  PublisherOA PDFDOI

Recent grants

• Collaborative Research: Collaborative Degradation Analysis for Enterprise-Level Maintenance Management via Dynamic Segmentation

  NSF · $195k · 2015–2019

• AF: Small: Collaborative Research: Personalized Environmental Monitoring of Type 1 Diabetes (T1D): A Dynamic System Perspective

  NSF · $158k · 2017–2022

• Collaborative Research: Data-Driven Smart Monitoring of Alzheimer's Disease via Data Fusion, Personalized Prognostics, and Selective Sensing

  NSF · $232k · 2014–2017

Frequent coauthors

• Xiaoning Qian

  Texas A&M University

  39 shared

• Zhangyang Wang

  15 shared

• Xiangyu Chang

  13 shared

• Ameer Hamza Shakur

  The University of Texas Southwestern Medical Center

  12 shared

• Byung-Jun Yoon

  Texas A&M University

  11 shared

• Ian K. Blaby

  Lawrence Berkeley National Laboratory

  11 shared

• Maria J. Soto

  Lawrence Berkeley National Laboratory

  11 shared

• Francis J. Alexander

  National Taiwan University of Science and Technology

  10 shared

Awards & honors

• Professional Achievement Award, Division of Data Analytics a…
• Teaching Award, Division of Data Analytics and Information S…
• Faculty Appreciation for Career Education & Training (FACET)…
• Best Paper Award (First Runner-up), IEEE Transactions on Aut…
• Best Applications Paper Award, IISE Transactions, 2016