Machine-Learning-Assisted Catalytic Sensor Array for Rapid Diagnosis of Aortic Dissection

Acute aortic dissection (AD) is a time-critical cardiovascular emergency that is clinically difficult to distinguish from other causes of acute chest pain. Rapid blood diagnostics with high specificity for AD are therefore urgently needed. Herein, we report a machine-learning-assisted sensor array constructed from manganese porphyrin (Mn-TCPP) and two framework assemblies (PFC-73-Mn, PCN-224-Mn) for the precise triage of AD. The sensor array generates disease-specific colorimetric response patterns through differential catalytic inhibition upon exposure to plasma. This sensing capability is governed by the protein corona effect, in which the disease-related plasma proteins form distinct coronal layers on the surface of sensing elements and modulate the permeation of substrates to the catalytic active sites. The sensor array combined with an XGBoost algorithm successfully distinguished AD patients from other causes of acute chest pain in a cohort of 181 individuals with an accuracy of 86.5% and an area under the curve (AUC) of 0.933. Fusion with clinical biomarkers using a Random Forest classifier further improved performance to 91.9% accuracy and 0.985 AUC. This study provides a rapid, robust tool for AD diagnosis using 20 μL of plasma in 15 min and offers mechanistic insight into the sensor array for disease triage based on blood samples.