Rapid antimicrobial susceptibility testing by image analysis

A method for testing antimicrobial susceptibility, the method comprising:

• receiving an input data structure, the input data structure comprising:
• an input image data set comprising a value for pixel intensity for each pixel of a sample plurality of pixels, the input image data set generated from imaging a sample plurality of wells at one or more time points after incubation using an imaging unit comprising a charge-coupled device (CCD) sensor, a contact image sensor (CIS), or other sensor that uses photodiodes, the sample plurality of wells containing a sample microorganism and a first and a second initial concentrations of a sample antimicrobial, the second initial concentration being greater than the first initial concentration, and
• a sample map representing the sample plurality of wells with values indicating an initial concentration of the sample antimicrobial in each well;
• inputting the input data structure into a model, the model trained by:
• receiving a first plurality of first data structures, each first data structure of the first plurality of first data structures comprising:
• a first image data set comprising a value for pixel intensity for each pixel of a first plurality of pixels, the first image data set generated from imaging a first plurality of wells, the first plurality of wells containing a first microorganism and a plurality of initial concentrations of a first antimicrobial, the first microorganism having a known minimum inhibitory concentration to the first antimicrobial, and
• a first map representing the first plurality of wells with values indicating the initial concentration of the first antimicrobial for each well,
• storing a plurality of first training samples, each including one of the first plurality of first data structures and a first label indicating the known minimum inhibitory concentration of the first microorganism to the first antimicrobial, and
• optimizing, using the plurality of first training samples, parameters of the model based on outputs of the model matching or not matching corresponding labels of the first labels when the first plurality of first data structures is input to the model,
• wherein an output of the model specifies the minimum inhibitory concentration of the first microorganism to the first antimicrobial for a given first data structure;
• and
• determining, using the model, a minimum inhibitory concentration of the sample microorganism to the sample antimicrobial, wherein the determining comprises:
• determining a first value of an image characteristic at a first time point of the one or more time points for a first subset of the sample plurality of wells, the image characteristic being a statistical measure of pixel intensities corresponding to the first subset, and
• determining a second value of the image characteristic at the first time point of the one or more time points for a second subset of the sample plurality of wells, the image characteristic being a statistical measure of pixel intensities corresponding to the second subset,
• wherein the minimum inhibitory concentration is determined using at least one of the first value or the second value.

A system for antimicrobial susceptibility testing, the system comprising:

• a dispensing unit configured to automatically dispense a plurality of concentrations of an antimicrobial to a plurality of locations on a well plate;
• an incubation unit configured to receive the well plate and to maintain a temperature set point;
• an imaging unit comprising a light source and a charge-coupled device (CCD) sensor, a contact image sensor (CIS), or other sensor that uses photodiodes, the imaging unit configured to measure light from the light source reflected by the well plate and to generate an image data set from the measured light, wherein the image data set comprises a value for pixel intensity for each pixel of a plurality of pixels;
• and
• a processor configured to execute a plurality of instructions, the plurality of instructions comprising:
• generating an input data structure comprising the image data set,
• analyzing the image data set to determine a first value of an image characteristic at a first time point of the one or more time points for a first subset of wells of the well plate, the first subset of wells containing a microorganism and a first concentration of the antimicrobial, the image characteristic is a statistical measure of pixel intensities corresponding to the first subset of wells, wherein the analyzing comprises inputting the input data structure to a trained model to generate the first value of the image characteristic,
• analyzing the image data set to determine a second value of the image characteristic at the first time point of the one or more time points for a second subset of wells of the well plate, the second subset of wells containing the microorganism and a second concentration of the antimicrobial, the second concentration being greater than the first concentration, the image characteristic being a statistical measure of pixel intensities corresponding to the second subset of wells, wherein the analyzing comprises inputting the input data structure to the trained model to generate the second value of the image characteristic, and
• determining a classification of a resistance of the microorganism to the antimicrobial using the first value and the second value.