Deep Learning-Powered Nanoplasmonic Biosensing Approach Enables Ultrasensitive Extracellular Vesicles Profiling for...

Area of Science:

• Biotechnology
• Nanotechnology
• Artificial Intelligence in Medicine

Background:

• Nanoplasmonic metasurfaces offer high sensitivity for biosensing but face challenges in data processing.
• Conventional biosensing methods struggle with efficient analysis of complex biological data.
• Early cancer detection, particularly for pancreatic ductal adenocarcinoma (PDAC), remains a critical clinical need.

Purpose of the Study:

• To develop an ultrasensitive biosensing strategy for small extracellular vesicle (sEV) analysis in serum using a novel KAN-enabled metasurface chip.
• To leverage deep learning for enhanced data processing and analysis in nanoplasmonic biosensing for cancer detection.
• To improve the accuracy and efficiency of cancer screening and clinical management through advanced biosensing technology.

Main Methods:

• Development of a metasurface chip integrated with Kolmogorov-Arnold networks (KAN) for sEV detection (metaEVchip).
• Application of KAN-powered deep learning for analyzing full-spectrum data from serum samples.
• Validation of the biosensing strategy using data from 600 pancreatic ductal adenocarcinoma (PDAC) patients and 1200 controls.

Main Results:

• The KAN-enabled metasurface chip achieved an exceptional area under the curve (AUC) of 0.99 in an external validation set.
• The strategy demonstrated superior performance compared to traditional biosensing methods.
• KAN's ability to simultaneously capture multi-dimensional spectral features was identified as key to enhanced data processing and accuracy.

Conclusions:

• The proposed KAN-enabled metasurface chip offers a highly sensitive and accurate platform for sEV analysis in serum.
• This AI-driven nanoplasmonic biosensing approach significantly advances cancer detection capabilities, particularly for PDAC.
• The technology establishes a new paradigm for cancer screening, potentially improving clinical management for various malignancies.