摘要
This study introduces an innovative AI-assisted system for diagnosing biliary atresia (BA) in infants, which achieves wide process window tolerance by integrating various process parameters for model training. This results in improved sensor accuracy and clinical validation that surpasses traditional linear regression calibration analysis. The AI system also establishes a buffer-to-blood correlation bridge, enhancing the correlation between purified sample tests and blood samples, which is cost-effective by reducing the need for blood samples. The predictive sensing capabilities of the AI model further enhance the sensor's prediction accuracy by utilizing fabrication process data, improving overall performance. Validation of 20 clinical samples using AI demonstrated an improvement in Pearson's correlation between our system and the standard ELISA test, increasing to 0.96. A single drop of blood for diagnosis makes the entire system more direct and less stressful for both infants and their parents. Additionally, the accelerated diagnostic timeline shortens the diagnosis duration from several hours to just 𝟏 𝟎 minutes. This advancement surpasses traditional dried blood spot (DBS) testing, demonstrating that our system enhances the early diagnosis of BA in newborns and holds promise for widespread application in healthcare.