Wu Peiying, Zhang Rui, Porte Céline, Kiessling Fabian, Lammers Twan, Rezvantalab Sima, Mihandoost Sara, Pallares Roger M
Institute for Experimental Molecular Imaging, RWTH Aachen University Hospital Aachen 52074 Germany
Fraunhofer Institute for Digital Medicine MEVIS Bremen 28359 Germany.
Nanoscale Adv. 2025 May 27. doi: 10.1039/d5na00265f.
Gold nanostars (AuNS) are nanoparticles with spiky structures and morphology-dependent optical features. These include strong extinction coefficients in the visible and near-infrared regions of the spectrum, which are commonly exploited for biomedical imaging and therapy. AuNS can be obtained seedless protocols with Good's buffers, which are beneficial because of their simplicity and the use of biocompatible reagents. However, AuNS growth and optical properties are affected by various experimental factors during their seedless synthesis, which affects their performance in diagnosis and therapy. In this study, we develop a workflow based on machine learning models to predict AuNS optical properties. This approach includes data collection, feature selection, data generation, and model selection, resulting in predictions of the first and second localized surface plasmon resonance positions within 9 and 15% of their true values (root-mean-squared percentage error), respectively. Our results highlight the benefits of using machine learning models to infer the optical properties of AuNS from their synthesis conditions, potentially improving nanoparticle design and production for better disease diagnosis and therapy.
金纳米星（AuNS）是具有尖刺结构和形态依赖光学特性的纳米颗粒。这些特性包括在光谱的可见光和近红外区域具有很强的消光系数，这一特性通常用于生物医学成像和治疗。可以使用古德缓冲液通过无种子方案获得AuNS，由于其操作简单且使用生物相容性试剂，因此这种方法很有益处。然而，在无种子合成过程中，AuNS的生长和光学性质会受到各种实验因素的影响，这会影响它们在诊断和治疗中的性能。在本研究中，我们开发了一种基于机器学习模型的工作流程来预测AuNS的光学性质。该方法包括数据收集、特征选择、数据生成和模型选择，分别得出第一和第二局域表面等离子体共振位置的预测值，其与真实值的偏差在9%和15%以内（均方根百分比误差）。我们的结果突出了使用机器学习模型从AuNS的合成条件推断其光学性质的好处，这可能会改善纳米颗粒的设计和生产，以实现更好的疾病诊断和治疗。
