Artificial neural network processed linear-light tristimulus and hue parameters of fluorescence for smartphone assisted point-of-care testing device

• Published in: Sensors and actuators. B, Chemical Chemical, 2023-06, Vol.384, p.133659, Article 133659
• Main Authors: Chen, Huiting, Zhuang, Zehong, Guo, Siyun, Xie, Shangfang, Xin, Yu, Chen, Yuying, Ouyang, Sixue, Zhao, Wei, Shen, Kui, Tao, Jia, Zhao, Peng
• Format: Article
• Language: English
• Subjects: Artificial neural network; Hue Parameters; Point-of-care testing; Serum; Smartphone
• ISSN: 0925-4005; 1873-3077
• DOI: 10.1016/j.snb.2023.133659

Smartphone based point-of-care testing (POCT) device has progressed rapidly for its advantages in simulating the main functions of large instruments. The capability of multiple biomarkers detection and the robust data acquiring and processing ability are significant for POCT device to meet the needs of convenient diagnostics. In this work, a POCT device was developed based on the cascade reaction of targets catalyzed by corresponding oxidases and leaf like zeolitic imidazolate framework. It was found that both linear-light tristimulus (RGB) and hue parameters (HSV) of fluorescence image of cascade reaction product were highly related with target concentration. Furthermore, a smartphone application was constructed to analyze the RGB and HSV of fluorescence to predict target level utilizing artificial neural network (ANN) algorithm. The regression values (R) for the training and validation of four targets were all higher than 95 %. The device realized the off-line detection of targets within 50 min in serum sample, and the assay results were comparable with standard methods. It also performed well in identifying the normal and abnormal serum samples. The proposed platform combining with the smartphone application could be used as a handy and costless POCT device for the rapid monitoring of metabolic biomarkers. •The cascade reaction of metabolic biomarkers catalyzed by oxidases and leaf like zeolitic imidazolate framework was utilized.•Artificial neural network (ANN) was used to process RGB and HSV parameters to predict target concentration.•A smartphone application (APP) “CLUG” integrated with the functions of data acquiring and ANN processing was designed.•The small device realized the off-line detection of metabolic biomarkers within 50 min in serum sample