Multi-branch neural network for hybrid metrology improvement - Université Grenoble Alpes
Article Dans Une Revue Small Année : 2021

Multi-branch neural network for hybrid metrology improvement

Résumé

Abstract Because of antibiotics misuse, the dramatic growth of antibioresistance threatens public health. Tests are indeed culture‐based, and require therefore one to two days. This long time‐to‐result implies the use of large‐spectrum antibiotherapies as a first step, in absence of pathogen characterization. Here, a breakthrough approach for a culture‐less fast assessment of bacterial response to stress is proposed. It is based on non‐destructive on‐chip optical tweezing. A laser loads an optical nanobeam cavity whose evanescent part of the resonant field acts as a nano‐tweezer for bacteria surrounding the cavity. Once optically trapped, the bacterium‐nanobeam cavity interaction induces a shift of the resonance driven by the bacterial cell wall optical index. The analysis of the wavelength shift yields an assessment of viability upon stress at the single‐cell scale. As a proof of concept, bacteria are stressed by incursion, before optical trapping, at different temperatures (45, 51, and 70 °C). Optical index changes correlate with the degree of thermal stress allowing to sort viable and dead bacteria. With this disruptive diagnosis method, bacterial viability upon stress is probed much faster (typically less than 4 h) than with conventional culture‐based enumeration methods (24 h).
Fichier non déposé

Dates et versions

hal-04626700 , version 1 (27-06-2024)

Identifiants

Citer

Paul Digraci, Maxime Besacier, Patrice Gergaud, Guido Rademaker, Jérôme Rêche. Multi-branch neural network for hybrid metrology improvement. Small, 2021, 18 (4), pp.46. ⟨10.1117/12.2612798⟩. ⟨hal-04626700⟩
19 Consultations
0 Téléchargements

Altmetric

Partager

More