About

Lei Xing is a professor at Stanford University and is associated with the Center for Artificial Intelligence in Medicine & Imaging (AIMI). He holds the title of Jacob Haimson and Sarah S. Donaldson Professor and is also a Professor, by Courtesy, of Electrical Engineering. His work focuses on the application of artificial intelligence in medicine and imaging, contributing to the advancement of healthcare through innovative AI-driven solutions. As a key figure at AIMI, he is involved in research, education, and leadership efforts aimed at integrating AI technologies into medical practice and imaging sciences.

Research topics

• Artificial Intelligence
• Computer Science
• Machine Learning
• Data Mining
• Medicine
• Materials science
• Biology
• Internal medicine
• Software engineering
• Engineering
• Computer vision
• Pathology
• Radiology
• Nuclear medicine
• Data science
• Chemistry
• Biochemistry
• Biomedical engineering
• Biophysics

Selected publications

• Emerging FLASH therapy platforms for stereotactic radiosurgery and body radiotherapy.

  PubMed · 2026-01-01

  article

  The integration of FLASH radiotherapy with stereotactic techniques presents a promising avenue for improving therapeutic outcomes through normal tissue sparing while maintaining tumor control. However, significant technical challenges must be addressed for successful clinical implementation. This review evaluates emerging platforms and technical requirements for combining FLASH delivery with stereotactic radiosurgery (SRS) and stereotactic body radiotherapy (SBRT). While electrons have enabled extensive preclinical FLASH research, their limited penetration depth makes them unsuitable for most stereotactic applications. Photon-based systems face significant engineering challenges in achieving both FLASH dose rates (>40 Gy/s) and the beam characteristics necessary for stereotactic delivery, particularly regarding heat management and multi-angle treatment capabilities. Proton and heavy ion systems offer advantages through the Bragg peak but require substantial development to overcome technical limitations in beam delivery and scanning speeds. We evaluate emerging platforms including novel accelerator designs, beam monitoring systems, and delivery techniques aimed at clinical translation. Critical technical requirements are discussed, including specialized dosimetry systems capable of ultra-high dose rate measurements, quality assurance protocols, treatment planning systems that optimize both spatial and temporal aspects of delivery, and novel image guidance strategies.

  Publisher
• RETRACTED

  Retraction Note: BMP2 expression in oral squamous cell carcinoma and its effects on SCC9 cell biological behavior

  Scientific Reports · 2026-03-09

  article
  PublisherOA PDFDOI

• Advancing In-Context Learning for Efficient and Stable Medical Report Generation

  IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026-01-01

  articleSenior author

  Vision-language models (VLMs) have shown strong generalization across multimodal tasks, but adapting them to medical report generation (MRG) often demands extensive paired image-text data that are limited due to data privacy and annotation cost. In-context learning (ICL) offers a promising training-free alternative, yet standard ICL approaches rely on long demonstration prompts that are computationally inefficient and often yield inconsistent or clinically inaccurate descriptions. To address these challenges, we propose Principal In-Context Vectors (PCVs), a compact latent-guidance framework that distills multimodal demonstrations into stable semantic representations. By extracting hidden states from auto-regressive VLMs and applying principal component analysis (PCA), we identify robust semantic directions that remain stable under input perturbations. These PCVs are then injected into new queries to steer generation toward accurate and clinically meaningful outputs without any model tuning. Extensive experiments on four MRG benchmark datasets show that our approach can enhance both zero-shot and fully supervised generation quality across diverse settings, including cross-center, cross-disease, and longitudinal scenarios. This work provides a lightweight and scalable approach to adapt pre-trained VLMs for practical clinical deployment.

  PublisherDOI

• Magnetic and Electrical Properties of La2−xBixNiMnO6 (x = 0.2, 0.5 and 1.0) Synthesized by High-Temperature and High-Pressure Method

  Symmetry · 2026-04-17

  articleOpen access1st authorCorresponding

  Polycrystalline La2−xBixNiMnO6 (x = 0.2, 0.5, 1.0) samples were synthesized via a high-temperature and high-pressure method, with their structural, magnetic, and electrical properties systematically characterized. X-ray diffraction (XRD) confirms a monoclinic double perovskite structure (space group P21/n) for all samples, while Bi3+ induces a lattice volume expansion trend inferred from XRD peak shifts due to its larger ionic radius than La3+. Magnetically, all exhibit ferromagnetism and soft magnetic features, with magnetization decreasing as Bi content increases. The x = 0.2 and 0.5 samples show two distinct Curie temperatures, both decreasing with Bi substitution, whereas the higher Curie temperature vanishes in the x = 1.0 sample, likely due to Bi-induced structural changes. Electrically, all display semiconducting behavior (resistivity: x = 0.5 > x = 0.2 > x = 1.0) and negative magnetoresistance (MR) at 200 K, peaking at 12% (x = 0.5) and 7.5% (x = 1.0). For the x = 1.0 sample, negative magnetoresistance strengthens with decreasing temperature (130–200 K), with magnetoresistance-field (MR-H) curves showing herringbone and arc shapes.

  PublisherOA PDFDOI

• Dynamic digital product passport for short-shelf-life products

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-21

  articleOpen access1st authorCorresponding

  Abstract Digital product passport (DPP) can enhance the transparency and efficiency across supply chains for long- and short-shelf-life products. Developing DPPs could enhance the overall sustainability of short-shelf-life products and reduce waste. However, the inherent instability in quality of these products driven by changing environmental and temporal conditions along supply chain, require a dynamic approach.

  PublisherDOI

• A Cross‐Tissue Multiomics Analysis Reveals the Protective Role of TGFBR3 in Postmenopausal Osteoporosis

  International Journal of Genomics · 2026-01-01

  articleOpen access

  Background: Postmenopausal osteoporosis (PMO) develops as a result of pathological cross-tissue interactions. However, current experimental paradigms are constrained by their single-tissue focus, hindering efforts to discover systemwide regulatory genes. Objective: We aimed to discover conserved genetic regulators of PMO by integrating cross-tissue transcriptomic profiles in humans and to characterize their biological functions via combined genetic epidemiology and experimental studies using integrated analytical strategies. Methods: Our analytical framework encompassed transcriptome profiles from human peripheral blood mononuclear cells, bone marrow, and bone tissue. We adopted a tiered strategy involving differential expression analysis, weighted gene coexpression network construction, and machine learning with 108 algorithm combinations for candidate gene selection. A two-sample Mendelian randomization was used to inform causal gene-disease relationships, while the key results were validated in an ovariectomized mouse model of osteoporosis. Mechanistic studies included single-cell transcriptomics, functional enrichment, and immune microenvironment profiling. Results: = 0.037). Osteoporotic mice exhibited considerable downregulation of TGFBR3 expression, which was positively correlated with bone mineral density and mechanical properties as well as bone formation markers and negatively correlated with resorption markers. Cellular localization showed enrichment of TGFBR3 in bone marrow mesenchymal stem cells and T cells from human and mouse bone marrow. Functional analyses suggested that its protective effects involve the modulation of osteogenic differentiation pathways and regulation of the immune microenvironment. Conclusion: This is the first study to identify TGFBR3 as a novel cross-tissue protective regulator of PMO. Our integrated approach covering genomic discovery, causal inference, and experimental validation offers strong support to the hypothesis that TGFBR3 deficiency constitutes a fundamental feature of PMO pathogenesis, while shedding light on its multilevel protective mechanisms.

  PublisherDOI

• A Novel Biphasic Absorbent for Liquid–Solid Phase Separation Induced by Nanoparticle Polarity Difference and Application in DAC

  Environmental Science & Technology · 2026-03-13 · 1 citations

  articleCorresponding

  Liquid absorption represents a technologically mature and immediately scalable approach for direct air capture (DAC), demonstrating a validated effectiveness in atmospheric CO2 removal. However, its practical deployment is hindered by the high energy consumption and cyclic stability. The core advantage of the biphasic solvent system is the low-energy regeneration of CO2 through liquid–solid phase transition, but its application in DAC faces bottlenecks due to environmental humidity sensitivity and susceptibility to oxidative degradation. In this work, we developed a stable nanoparticle-coupled biphasic ionic liquid system ([AEP][1,2-DMI]-DMSO-H2O/MgO) for DAC, where the introduced nanoparticles significantly enhanced the CO2 transport and phase transition kinetics. Through its unique liquid–solid transition mechanism, 79.5% of captured CO2 was concentrated in a minimal solid phase (15.1% of system mass), while maintaining high loading capacity (total loading: 0.65 mol·mol–1, solid phase loading: 0.30 g·g–1). This system can achieve an ultralow regeneration energy (0.94 GJ·t–1CO2) and exceptional cycling stability (>96.5% efficiency retention over 7 cycles). The nanoparticle-enhanced biphasic ionic liquid system developed in this study has led to a breakthrough in new liquid DAC technology, pushing the technology into a new stage of development by unifying ultralow energy consumption and excellent stability.

  PublisherDOI

• Lyar contributes to cell cycle progression and multi-lineage differentiation in mouse embryonic stem cells

  Frontiers in Genetics · 2026-03-20

  articleOpen access

  Ly-1 antibody reactive clone gene ( Lyar ) is involved in the regulation of embryonic stem cell (ESC) self-renewal. To explore the specific role of Lyar in cell cycle progression and embryonic differentiation, we generated Lyar knockout (KO) mouse ESC (mESC) lines using CRISPR/Cas9, and investigated the effects of Lyar deficiency on mESC proliferation, cell cycle, apoptosis and multi-lineage differentiation. We found that Lyar deficiency reduces proliferation, increases apoptosis, and elevates p53 and p21 protein expression. The impaired mESC proliferation is associated with the increased apoptosis and cell cycle progression defect, which is driven by p53-p21 pathway activation. In embryoid body (EB) formation assay, loss of Lyar led to significant downregulation of most germ layer-specific markers in KO mESC clones, including mesoderm ( Gsc , T ), endoderm ( Gata4 , Sox17 ) and ectoderm marker Pax6 . These findings confirm that Lyar is required for cell cycle progression, proliferation, and lineage-specific marker expression during early differentiation, demonstrating that Lyar may serve as a critical regulatory factor in stem cell biology.

  PublisherOA PDFDOI

• Dynamic digital product passport for short-shelf-life products

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-21

  articleOpen access1st authorCorresponding

  Abstract Digital product passport (DPP) can enhance the transparency and efficiency across supply chains for long- and short-shelf-life products. Developing DPPs could enhance the overall sustainability of short-shelf-life products and reduce waste. However, the inherent instability in quality of these products driven by changing environmental and temporal conditions along supply chain, require a dynamic approach.

  PublisherDOI

• Microbial and chemical dynamics in cigar tobacco aging driven by spatiotemporal environmental heterogeneity

  Industrial Crops and Products · 2026-03-19

  articleOpen access

  The natural aging of cigar tobacco leaves is a complex process involving interactions among environmental conditions, microbial community dynamics, and chemical transformations. However, the ecological mechanisms driving microbial succession and chemical changes across different production regions remain poorly understood. In this study, we conducted synchronized aging experiments in four major cigar-producing regions, combining microenvironment monitoring with comprehensive microbial and chemical analyses. Pronounced regional differences in temperature and humidity significantly influenced major chemical traits, including total nitrogen, alkaloids, petroleum ether extracts, and mineral elements. Microbial community structure and diversity also varied significantly across regions and aging stages ( p < 0.05). Random forest regression showed that multiple bacterial and fungal taxa were associated with chemical variation, with fungal community composition and richness exhibiting stronger explanatory power than bacterial diversity. Notably, Fusarium emerged as a representative fungal driver, while bacterial genera such as Massilia and Bacillus displayed significant but relatively weaker associations. Structural equation modeling further revealed that temperature and humidity influenced chemical characteristics indirectly through microbial communities, with bacterial diversity exerting a positive effect and fungal community composition a negative influence ( p < 0.05). Null model indicated that bacterial community assembly was primarily governed by stochastic processes, whereas fungal assembly involved both stochastic and deterministic processes. Overall, these findings provide a mechanistic that connects microenvironmental regulation, microbial community assembly, and chemical evolution, offering insights for region-specific optimization of cigar tobacco aging processes. • Regional temperature and humidity shaped chemical composition of cigar leaves. • Massilia , Solibacillus , Fusarium , and Mortierella contributed to chemical variation. • Fungal structure and richness more influence chemical profiles than bacteria. • Microenvironment affects chemical traits indirectly via microbial communities.

  PublisherDOI

Recent grants

• Dose Reduction and Scatter Correction for Cone Beam Computed Tomography

  NSF · $300k · 2009–2012

• Intraoperative integration of artificial intelligence during cystoscopic surgery

  NIH · $2.3M · 2022–2027

• NIH Grant R01CA133474

  NIH · $960k · 2014

• NIH Grant R01EB016777

  NIH · $2.1M · 2018

• Leveraging deep learning for markerless motion management in radiation therapy

  NIH · $2.1M · 2021–2026

Frequent coauthors

• Hang Lee

  Massachusetts General Hospital

  416 shared

• Bo Yu

  Harbin Medical University

  395 shared

• Ik‐Kyung Jang

  Harvard University

  339 shared

• Shaosong Zhang

  Yunnan Academy of Agricultural Sciences

  289 shared

• Haibo Jia

  Nanjing Medical University

  257 shared

• Shiro Uemura

  Kawasaki Medical School

  253 shared

• Soo–Joong Kim

  Kyung Hee University

  226 shared

• Koji Kato

  Nippon Medical School Hospital

  218 shared