Lavonda Li

All Publications

• Machine-Learning Microfluidic Minute-Scale Microorganism Metrics Monitoring(M6). Advanced science (Weinheim, Baden-Wurttemberg, Germany) Yang, N., Ding, J., Chen, S., Yan, L., Ding, S., Li, L., Sun, J., Liu, H., Li, T., Liu, N., Wei, M., Zhu, X., Zou, X., Yuan, S., Zhang, X. 2026: e21106

  Abstract

  On-site monitoring of microorganisms remains challenging because of low concentrations, strong background interference, and dynamic aerosol diffusion, particularly for aerosol-transmitted pathogens. Here, we report a rapid detection platform that integrates a Puri-focusing microfluidic chip, electrochemical impedance spectroscopy (EIS), and machine learning for the analysis of airborne microorganisms. Guided by fluid-dynamic design and laminar-flow focusing, the chip achieved a 95.8% separation efficiency for 5 µm target particles. African swine fever virus (ASFV) was used as a model pathogen. Impedance features, including modulus, real and imaginary components, and phase angle, were extracted from aerosol samples and analyzed using multiple machine learning classifiers. Five-fold cross-validation identified Random Forest (RF) as the optimal model, achieving 95.2% classification accuracy. The platform reached a system-level detection limit of 188 TCID50/mL for air-sampled aerosols and showed high concordance with enzyme-linked immunosorbent assay (ELISA) results. Each detection cycle required less than 1 minute. This integrated strategy offers a feasible route for rapid on-site monitoring of aerosol-transmitted microorganisms in public health, agriculture, livestock farming, and production safety.

  View details for DOI 10.1002/advs.202521106

  View details for PubMedID 41984488

• EchoAtlas: A Conversational, Multi-View Vision-Language Foundation Model for Echocardiography Interpretation and Clinical Reasoning. medRxiv : the preprint server for health sciences Chao, C. J., Asadi, M., Li, L., Ramasamy, G., Pecco, N., Wang, Y. C., Poterucha, T., Arsanjani, R., Kane, G., Oh, J. K., Banerjee, I., Langlotz, C., Fei-Fei, L., Adeli, E., Erickson, B. J. 2026

  Abstract

  Echocardiography is the most widely used cardiac imaging modality, yet artificial intelligence-enabled interpretation remains limited by the inability of existing models to integrate visual assessment, quantitative measurement, and clinical reasoning within a unified framework. Here we present EchoAtlas, the first autoregressive vision-language model developed for echocardiographic interpretation. Trained on over 12.9 million question-answer pairs derived from approximately 2 million echocardiogram videos, EchoAtlas achieves 0.966 accuracy on multiple-choice questions in our internal test set and establishes a new state-of-the-art on the public MIMIC-EchoQA benchmark (0.699 vs. 0.508 previously). EchoAtlas also provides accurate quantitative measurements, segment-level regional wall motion assessment, longitudinal comparison, and diagnostic reasoning across diverse question formats - capabilities not previously demonstrated in this domain. These results highlight the potential of autoregressive vision-language models as a foundation for interactive echocardiographic interpretation, representing an early step toward scalable, auditable artificial intelligence systems in cardiology practice.

  View details for DOI 10.64898/2026.03.14.26348388

  View details for PubMedID 41891021

  View details for PubMedCentralID PMC13015684

• Artificial intelligence-powered nanomedicine. Chemical Society reviews Luo, G., Jiang, X., Hu, C., Li, L., Yan, L., Xiao, G., Duo, Y., Zhang, X. 2026

  Abstract

  The escalating global burden of diseases-including cancer, neurodegenerative, and cardiovascular disorders-poses severe threats to human health and social development. The inherent limitations of conventional diagnostic, imaging, and therapeutic modalities have driven the rapid evolution of nanomedicine, particularly nanotheranostics, which integrates diagnostic and therapeutic functionalities within a single nanoplatform for enhanced precision and safety. Despite remarkable advances in nanotechnology and materials science over recent decades, challenges such as the biological complexity of living systems, incomplete understanding of nano-bio interactions, inefficiencies in nanoparticle synthesis, and limited clinical translation continue to hinder progress. The recent convergence of nanomedicine with artificial intelligence (AI) and computational sciences has opened transformative opportunities to overcome these obstacles. AI-empowered algorithms, including machine learning, deep learning, and generative models, are increasingly being applied to optimize nanoparticle design and synthesis, predict nano-bio interactions, and improve diagnostic and therapeutic efficacy. These approaches not only accelerate materials discovery but also enable data-driven, adaptive nanotheranostic systems capable of autonomous optimization across disease contexts. This review systematically summarizes the current landscape of AI-powered nanomedicine, highlighting advances in nanoparticle design, synthesis, and the development of AI-guided diagnostic and therapeutic nanoplatforms. It further discusses applications in bioimaging, targeted therapy, and clinical translation, while identifying existing challenges and future perspectives in establishing next-generation AI-empowered nanotheranostics. Ultimately, the integration of artificial intelligence and nanotechnology is expected to revolutionize precision medicine by bridging the gap between fundamental nanoscience and clinical implementation, paving the way toward intelligent, personalized healthcare.

  View details for DOI 10.1039/d5cs01406a

  View details for PubMedID 41636234

• Bioorthogonal optimized virus immuno-nanomedicine (BOVIN). Nature communications Peng, W., Du, Y., Cui, L., Xiao, Y., Li, L., Yan, L., Gu, Y., Zhang, L., Li, B., Wang, Z., Wang, H., Dai, X., Teng, Y., Wang, T., Zheng, B., Zhang, X. 2025

  Abstract

  The failure of tumour immunotherapy can be attributed to two primary factors: the limited immunogenicity of tumour cells and the immunosuppressive characteristics of the tumour microenvironment. Recent research indicates that cancer patients infected with respiratory viruses exhibit a positive antitumour response, as the activated host immune system. Here we show a bioorthogonal optimized virus immuno-nanomedicine (BOVIN) that uses nonpathogenic recombinant virus immuno-nanomedicine (VIN) for the targeted therapy of solid tumours. The influenza viruses (A/WSN/1933(WSN)) induce immunogenic cell death (ICD) in tumour cells by triggering a stress response mediated by integrated mitochondrial dysregulation. Furthermore, the surface modification of WSN with lactate oxidase enables the consumption of lactate to produce hydrogen peroxide, which synergistically enhance ICD activation. This leads to the attraction and significant activation of antigen-presenting cells in the tumour region through the exposure of calreticulin and secretion of high mobility group box 1. Additionally, the BOVIN triggers the tumour ICD, significantly enhances CD8+ T-cell tumour infiltration, strengthens the sensitivity of immune checkpoint blockers, and thus activates anti-tumour immune memory to effectively inhibit tumour recurrence after surgery. In conclusion, bioorthogonal nonpathogenic recombinant VIN reverses the immunosuppressed state of malignant cells, offering potential for improved tumour immunotherapy.

  View details for DOI 10.1038/s41467-025-66089-w

  View details for PubMedID 41290673

• Parabiosis, Assembloids, Organoids (PAO). Advanced science (Weinheim, Baden-Wurttemberg, Germany) Hong, Y., Li, L., Yan, L., Bai, L., Su, J., Zhang, X. 2025: e11671

  Abstract

  The research and treatment of major diseases challenge global public health, necessitating advanced disease models. Existing approaches have clear limitations: two-dimensional cell cultures lack multi-organ interactions, clinical trials are costly and ethically constrained, and animal models, focused on single organs, fail to replicate systemic regulation. Parabiosis, which connects two organisms via shared circulation, provides insights into systemic factors and multi-organ interactions but has limited applicability to humans. Furthermore, organoids are three-dimensional structures formed through stem cell self-organization that replicate the functions of individual tissues and advance personalized medicine; however, they cannot model inter-tissue interactions. Assembloids overcome these constraints by integrating diverse organoids, enabling sophisticated simulation of multi-organ dynamics. The integration of these parabiosis, assembloids, organoids (PAO) models with emerging technologies, such as artificial intelligence for precision analytics, CRISPR-based gene editing for disease mechanism elucidation, organ-on-a-chip platforms for dynamic environmental control, and soft robotics for replicating physiological biomechanics, promises to revolutionize disease modeling, regenerative medicine, and precision therapeutics. This review evaluates parabiosis, assembloids, and organoids, highlighting their development, current limitations, and transformative potential when combined with frontier biomedical engineering approaches to address complex human diseases.

  View details for DOI 10.1002/advs.202511671

  View details for PubMedID 41014607