文献类型：期刊文献

英文题名：Darkfield illumination enhancing artificial-intelligence-assisted digital microfluidics for on-site pathogen detection

作者：Song, Zerui[1,2];Xu, Jingsong[3];Guo, Kunlun[1,2];Wu, Zhanli[1,2];Feng, Boyi[1,2];Tang, Wanxin[4];Wang, Hua[3];Gu, Zhen[1,2];Wang, Huifeng[1,2]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, State Key Lab Bioreactor Engn, Shanghai 200237, Peoples R China;[3]Shanghai Jiao Tong Univ, Renji Hosp, Sch Med, Dept Lab Med, Shanghai 200127, Peoples R China;[4]Shanghai Normal Univ, Coll Chem & Mat Sci, Shanghai 200234, Peoples R China

年份：2026

卷号：1382

外文期刊名：ANALYTICA CHIMICA ACTA

收录：;EI(收录号：20254419430253);WOS:【SCI-EXPANDED(收录号:WOS:001612334000005)】；

基金：This research is supported by the National Key Research and Development Program of China (2024YFA0917700) , the National Nat-ural Science Foundation of China (No. 62103148, No. 82472374) , and the Shanghai Science and Technology Commission (contract number: 23121900300) .

语种：英文

外文关键词：Digital microfluidics; Darkfield illumination; Droplet processing; Mycoplasma pneumoniae

摘要：Background: Artificial-intelligence (AI) based computer vision emerges as an important part in the digital microfluidic (DMF) system for realization of high-level automatic control of droplets for on-site pathogen detection at the nano-micro scale. However, the low optical contrast and indistinct of droplets in the oil media by conventional imaging approaches limits the performance of the AI-based droplet recognition and feature extractions in practical applications. Results: Herein, we present an integrated Indium-Tin-Oxide (ITO)-DMF platform featuring a switchable dual-mode optical system. The system utilizes a darkfield imaging mode for robust, AI-driven process control, which is enabled by a darkfield illumination-based transmissive configuration that significantly enhances droplet contrast. This enhancement proved essential for deep-learning models, yielding exceptional performance in object detection (mAP50 = 98.3 %) and semantic segmentation (mIoU = 98.2 %). This high-fidelity visual feedback is critical for precise droplet manipulation, which precedes endpoint analysis where the system switches to an epi-fluorescence mode. As a functional validation, this AI-assisted control enabled the automated, multi-step detection of Mycoplasma pneumoniae from clinical samples in under 11 min on-chip reaction time, achieving 100 % concordance with qPCR results. Significance: This study demonstrates that integrating darkfield illumination into DMF platforms is a powerful and essential strategy to empower AI-driven visual feedback for robust, automated process control. The proposed ITO-DMF platform with darkfield illumination is promising to be a versatile and highly extensible solution for AI-assisted DMF applications, such as point-of-care testing, DNA storage, single cell analysis and synthetic biology.