HepatoAI^LFA: Machine-Learning-Assisted Nano-enhanced Point-of-Care System for Personalized Precise Diagnosis of Hepatocellular Carcinoma.

Abstract

Early diagnosis significantly improves survival rates for hepatocellular carcinoma (HCC), yet traditional methods face limitations, including specialized instruments/personnel and prolonged reporting cycles. While lateral flow immunoassay (LFA) offers a promising alternative, its sensitivity and accuracy remain challenged. Here, we present HepatoAI^LFA, a machine-learning-assisted nano-enhanced point-of-care system that enables sensitive biomarker detection and digitalized, personalized HCC risk assessment. Taking enzyme-engineered metal-polydopamine frameworks as nanoprobes, a visually amplified LFA is developed for picogram-level detection of α-fetoprotein (AFP) and des-γ carboxyprothrombin (DCP). To assist LFA diagnosis, a machine-learning-based diagnostic model termed F4-ASAD is constructed by incorporating AFP, DCP, and two key predictors, age and sex. End-users input age, sex, and biomarker concentrations into a self-developed F4-ASAD-based online calculator via a smartphone to instantly obtain the HCC risk probability. This integrated system achieves 91.1% accuracy, 93.3% sensitivity, and 86.7% specificity, showing significant potential to streamline clinical workflows and advance precise disease diagnosis.