Dynamic Laser Speckle Imaging Meets Machine Learning to Enable Rapid Antibacterial Susceptibility Testing (DyRAST)

Abstract

Rapid antibacterial susceptibility testing (RAST) methods which measure change of a bacterial phenotype in response to a given treatment are of significant importance in healthcare, as they can assist care-givers in timely administration of the right treatment. Various RAST techniques have been reported for tracking bacterial phenotypes, such as size, shape, motion, and metabolic activity. However, they still require bulky and expensive instruments (which hinders their application in resource-limited environments) and/or utilize labeling reagents (which can interfere with antibiotics and add to cost). Furthermore, the existing ultra-rapid methods do not address possible adaptation of gradual adaptation of bacteria to antibiotics, which can lead to false interpretation of resistance when using ultra-rapid methods. In this work, we present a RAST approach leveraging machine learning analysis of time-resolved dynamic laser speckle imaging (DLSI) results to accurately predict the minimum inhibitory concentration (MIC) of a model strain of Escherichia coli in 60 minutes, compared to 6 hours using the currently FDA-approved phenotype-based RAST technique. To demonstrate the DLSI performance, we studied the effect of a -lactam ampicillin and an aminoglycoside gentamicin on Escherichia coli strain K-12. DLSI captures change of bacterial motion/division in response to treatment. The machine learning algorithm was trained and validated using the overnight results of gold standard, broth

Laser Applications for Estimation of Microbial Activity and Investigation of NIR Skin Autofluorescence

“…Speckle analysis has also been applied to biomass growth kinetic measurements in liquid culture, characterization of Colony Forming Units (CFU) morphology, determination of antibiotic susceptibility [2][3][4][5]. As well as laser speckle imaging techniques in combination with Deep Learning (DL) and Artificial Neural Networks (ANN) have demonstrated a fast system response of antibacterial susceptibility evaluation in MIC (Minimum Inhibitory Concentration) tests [6]. In the presented paper, we will prove that dynamic laser speckle…”

“…where 9:;: [5,6] is the Fourier transform of -<1[=]⋅? [=−5], a complex function representing the phase and magnitude of the signal over time and frequency; 4) marking and determination of the times and spatial locations of detected signals; 5) marking the growth start times of each 20 × 20 pix square, knowing the location of each square in space and its distance from the colony center (marked manually in step 1), finally the colony radius was calculated as a function of time (Fig.…”

A machine learning-based multimodal electrochemical analytical device based on eMoSx-LIG for multiplexed detection of tyrosine and uric acid in sweat and saliva

“…ML has also been used to selectively detect heavy metal ions in presence of interferents [29]. Convergence of ML and multimodal electrochemical readout to create fingerprint of analytes can help achieve multiplexing with high accuracy [30]. However, ML-assisted electrochemical biosensing is still in inceptive stage [24], [31].…”

Assessing the properties of a colloidal suspension with the aid of deep learning

“…noisy images [14]) or abstract/recognise useful information from a human-unintelligible set essentially without constructing an explicit model. It seems that so far the number of abstracted classes has been usually limited to several [21][22][23][24].…”