A homogeneous plasmon-enhanced Raman biosensor integrated with LASSO and deep learning approach to advance label-free and rapid diagnosis of urolithiasis

Abstract

Accurate detection and diagnosis of urolithiasis are critical to follow-up clinical treatments. Current stones diagnosis commonly relies on on-spot urinalysis assisted with extra gold standard examinations, however, the required hospital testing is cumbersome and time-consuming. Here, we develop a homogeneous plasmon-enhanced Raman biosensor integrated with the least absolute shrinkage and selection operator (LASSO) method and an artificial neural network (ANN) algorithm, namely LASSO-ANN-PERB, to sensitively assess spectroscopic variations of human urine samples for label-free and rapid stone screening. In a practical scenario loaded with low-volume Raman features of urine from healthy subjects and patients, the integrated biosensor realizes an excellent stones identification accuracy as well as can largely shorten the time consumption throughout the workflow. These practically demonstrated merits suggest the potential of integrated biosensor with the unique ability to enable noninvasive urine fluid biopsy for further development of rapid stones diagnostics, which may promisingly advance the existing clinical routine toward a convenient one.