AI-supported quantitative anti-müllerian hormone measurement with lateral flow biosensor

Abstract Objectives This study aims to develop a mobile-integrated, paper-based biosensor for the quantitative detection of anti-müllerian hormone (AMH), providing a reliable and objective tool to support reproductive health assessment. Rather than serving as a stand-alone diagnostic for a specific condition, the system focuses on accurate AMH quantification to assist broader hormonal biomarker monitoring in both clinical and decentralized settings. Methods An antibody-based lateral flow biosensor was combined with a deep learning-powered image analysis pipeline. Smartphone-captured images of the test strip were processed using a YOLOv8-based object detection model to localize the test line and extract RGB intensity values. These features were then analyzed by a random forest regression model trained to map image intensities to AMH concentrations. Image standardization and lighting correction algorithms were implemented to ensure measurement robustness under variable field conditions. Results The developed biosensor achieved quantitative AMH detection within a concentration range of 0.05–20 ng/mL, exhibiting a strong correlation with standard laboratory immunoassays (R 2 =0.9966). The YOLOv8-integrated analytical pipeline demonstrated excellent selectivity, repeatability, and stability under different illumination scenarios. Receiver operating characteristic analysis (AUC=0.962) confirmed high diagnostic reliability for AMH level stratification, emphasizing the system’s potential in reproductive endocrinology. Conclusions This AI-integrated biosensor represents a robust and accessible tool for AMH monitoring at the point of care. The combination of paper-based sensing and deep learning-driven quantification offers a scalable platform for hormonal biomarker analysis, paving the way for broader applications in personalized reproductive health management.