The traditional bacterial identification method of growing colonies on agar plates can take several days to weeks to complete depending on the growth rate of the bacteria. Successfully decreasing this analysis time requires cell isolation followed by identification. One way to decrease analysis time is by combining dielectrophoresis (DEP), a common technique used for cell sorting and isolation, and Raman spectroscopy for cell identification. DEP‐Raman devices have been used for bacterial analysis, however, these devices have a number of drawbacks including sample heating, cell‐to‐electrode proximity that limits throughput and separation efficiency, electrode fouling, or inability to address sample debris. Presented here is a contactless DEP‐Raman device to simultaneously isolate and identify particles from a mixed sample while avoiding common drawbacks associated with other DEP designs. Using the device, a mixed sample of bacteria and 3 μm polystyrene spheres were isolated from each other and a Raman spectrum of the trapped bacteria was acquired, indicating the potential for cDEP‐Raman devices to decrease the analysis time of bacteria.

When developing a Raman spectral library to identify bacteria, differences between laboratory and real world conditions must be considered. For example, culturing bacteria in laboratory settings is performed under conditions for ideal bacteria growth. In contrast, culture conditions in the human body may differ and may not support optimized bacterial growth. To address these differences, researchers have studied the effect of conditions such as growth media and phase on Raman spectra. However, the majority of these studies focused on Gram-positive or Gram-negative bacteria. This article focuses on the influence of growth media and phase on Raman spectra and discrimination of mycobacteria, an acid-fast genus. Results showed that spectral differences from growth phase and media can be distinguished by spectral observation and multivariate analysis. Results were comparable to those found for other types of bacteria, such as Gram-positive and Gram-negative. In addition, the influence of growth phase and media had a significant impact on machine learning models and their resulting classification accuracy. This study highlights the need for machine learning models and their associated spectral libraries to account for various growth parameters and stages to further the transition of Raman spectral analysis of bacteria from laboratory to clinical settings. K E Y W O R D S discriminant analysis, environmental microbiology, mycobacteria, Raman spectroscopy