Method and system for automated assessment of antibiotic susceptibility

Abstract

To provide an imaging system and method providing automated microbial growth detection for antibiotic sensitivity testing.SOLUTION: A processing system having an image sensor for capturing images of an inoculated culture plate having antibiotic disks disposed on culture media captures images of the plate at separate times. From comparison between a first image and a second image, the system generates pixel characteristic data for pixels of the second image. The pixel characteristic data may be indicative of plate growth. The system may access growth modeling data concerning the antibiotic disks and generate simulated image data with a growth model function. The growth model function uses the growth modeling data. The simulated image data simulates growth on the plate relative to the disks. The system compares the simulated image and the pixel characteristic data to identify pixel regions of the second image that differ from the simulated image.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] [Cross-reference of related applications] This application claims the benefit as of the filing date of U.S. Provisional Patent Application No. 62 / 564,727, filed on 28 September 2017, the disclosures of which constitute part of this specification by reference. [Background technology]

[0002] There is growing interest in digital imaging of culture plates for the detection of microbial growth. A technique for imaging plates for the detection of microbial growth is described in International Publication No. 2015 / 114121 (which in whole forms part of this specification by reference). Using such a technique, laboratory personnel can use high-quality digital images for plate inspection, eliminating the need to read plates by direct visual inspection. The shift of laboratory workflows and decision-making to the examination of digital images of culture plates may also improve efficiency. Operators can mark images for further detailed examination by the operator or another person with the appropriate skills. Additional images can also be taken and used to guide secondary processes.

[0003] For example, imaging can be used in agar diffusion tests. Agar diffusion tests determine the susceptibility of bacterial microorganisms to antibiotics. Such tests (sometimes called antibiotic susceptibility tests (ASTs)) typically involve applying several antibiotic discs to a culture medium (e.g., agar) on a plate uniformly coated with the bacteria of test. Different discs may contain different concentrations of a specific antibiotic and / or several different antibiotics. The plate is incubated to allow time for bacterial growth. The plate is then observed. Bacterial growth around the periphery of each antibiotic disc serves as an indicator of the effectiveness of the specific antibiotic in that disc. For example, an effective antibiotic in a particular disc may have a large area where the tested bacteria do not grow, while an ineffective antibiotic in a particular disc may have no area where the tested bacteria do not grow.

[0004] The size of the no-growth zone can provide an indicator of the minimum inhibitory concentration (TPI) of antibiotics in the peripheral disk. For example, in the case of agar, antibiotics migrate away from the disk over time after it is placed. This migration causes the antibiotic concentration to diffuse depending on the distance from the disk, the diffusion rate of the antibiotic, and the culture medium. The antibiotic concentration is highest near the disk and decreases with increasing distance from the disk. Typically, the minimum inhibitory concentration can be considered the lowest concentration furthest from the disk, which includes no bacterial growth.

[0005] Such decisions can help in selecting appropriate antibiotics and their dosages for various bacterial infections. This helps clarify the purpose of modern microbiological imaging systems. Achieving these objectives as quickly as possible helps to deliver results to patients rapidly and to provide such results and analyses economically. Automating laboratory workflows and decision-making can improve the speed and cost-effectiveness that can achieve these goals.

[0006] While considerable progress has been made in imaging techniques for detecting evidence of microbial growth, there is still a need to expand the application of such imaging techniques to support automated workflows. The apparatus and methods for examining culture plates for indicators of microbial growth are difficult to automate, partly due to the highly visual nature of plate examination. Therefore, there is a need to develop techniques that can automate the interpretation of culture plate images (e.g., identification of growth, susceptibility testing, antibiotic susceptibility analysis, etc.) and determine the next steps to be performed based on the automated interpretation. [Overview of the Initiative]

[0007] One aspect of this disclosure relates to a method for antibiotic susceptibility testing in a processor. This method involves preparing a culture plate inoculated with a biological sample. The culture plate comprises a culture medium and at least one antibiotic disk placed thereon. First and second image data of the culture plate are generated by an image sensor. The first and second image data represent first and second acquisition images of the culture plate, respectively. The first and second acquisition images are captured at different time points using the image sensor. The image sensor is controlled to collect desired image information (i.e., color, intensity, etc.). Pixel characteristic data is generated for pixels of the second image data from a comparison of the first and second image data. The pixel characteristic data indicates microbial growth over time on the culture plate. Next, modeling data for microbial growth is accessed. The modeling data models microbial growth for combinations of culture medium, microorganism, antibiotic, and antibiotic concentrations on and in the culture medium. The antibiotic concentration in the culture medium is a function of the antibiotic concentration on the disk, time, and distance from the disk. Simulated image data is generated using the growth model function. The simulated image data simulates microbial growth on an inoculated culture plate based on at least one of several antibiotic disks placed on the culture medium, the antibiotic concentration on at least one disk, the culture medium, and the antibiotic concentration in the culture medium as a function of time and distance from the antibiotic disk. By comparing the simulated image data with pixel characteristic data, one or more pixel regions of the second image data that are different from one or more pixel regions of the simulated image data are identified.

[0008] An example of pixel characteristic data is contrast data. Contrast data may also be pixel intensity values. Contrast data may include opacity data, color data, and blur data.

[0009] Pixel characteristic data may include distance data, which represents the distance to at least one of multiple antibiotic disks. For example, the distance data is the distance from the pixel to the center of at least one of the multiple antibiotic disks.

[0010] The growth model function may model one or more of the maximum growth, minimum growth, mean growth, median growth, or percentile growth. In one example, the growth model function characterizes contrast data as a function of radial distance from at least one disk of multiple antibiotic disks. The growth model function may include a diffusion map for the diffusion of antibiotics into the culture medium.

[0011] The method may include the step of detecting an image of the disk itself in the first or second image data, and analyzing the disk image data so that indicia data can be detected. Accessing the growth modeling data may include positioning the growth modeling data using the indicia data. The growth modeling data may include concentration information for the antibiotic load of at least one disk of a plurality of antibiotic disks.

[0012] In one example, the growth model function uses one or more of the following parameters: i) a growth time parameter representing the elapsed time of growth in the culture plate at the time of acquisition of the second image data, ii) a diffusion coefficient parameter, and iii) a dimensional parameter.

[0013] In one example, the generated simulated image data includes an image mask having a first image pixel representing a no-growth zone corresponding to the location of at least one of several antibiotic disks on a culture plate, and a second image pixel representing a growth zone that extends radially from the no-growth zone and begins at a point at a radial distance from the aforementioned location, where the radial distance is determined from a growth model function and represents the estimated inhibition zone limit of at least one disk.

[0014] When comparing simulated image data with pixel characteristic data, difference image data is generated using the contrast between pixels in the image mask and the pixel characteristic data. Then, a region of the second image data is evaluated based on the difference image data. For temporal contrast (i.e., the difference in pixel intensity over time), the first image data can be used as a pre-growth reference for the augmentation process.

[0015] In one embodiment, the method is executed by a computer. Specifically, the above method can be performed using instructions executed by a processor.

[0016] A system for antibiotic susceptibility testing is also described herein. The system includes an image sensor configured to capture an image of a culture plate on which culture medium is placed on which at least one antibiotic disk is placed. A biological sample is inoculated onto the culture plate and the culture plate is within the field of view of the image sensor. The system includes a processor coupled to the image sensor. The system also includes a computer medium containing programming instructions that, when executed by the processor, control the processor for antibiotic susceptibility testing. For example, the programming instructions control the processor to perform the method described herein.

[0017] In another embodiment, the system includes an image sensor configured to capture an image of a culture plate on which a culture medium is placed, with at least one plurality of antibiotic disks placed on top of it. When the culture plate is within the field of view of the image sensor, a biological sample is inoculated onto the culture plate, and the image sensor generates first image data and second image data, the first and second image data representing first and second acquisition images of the culture plate containing the plurality of antibiotic disks, respectively, and the first and second acquisition images are captured at first and second time points, respectively. The system includes a processor and memory. The processor receives the first and second image data and accesses the memory. The memory stores growth modeling data. The processor generates pixel characteristic data for pixels in the second image data based on a comparison of the first and second image data. The pixel characteristic data indicates microbial growth on the culture plate. The processor also accesses growth modeling data, which models microbial growth as a function of the culture medium, microorganism, antibiotic, and antibiotic concentration. The processor also generates simulated image data using a growth model function, which uses growth modeling data to simulate microbial growth on a culture plate for at least one disk of multiple antibiotic disks. The processor then compares the simulated image data with pixel characteristic data to identify one or more pixel regions of a second image data that are different from the simulated image data. [Brief explanation of the drawing]

[0018] [Figure 1] This is a schematic diagram of an example of a system for image-based antibiotic susceptibility testing according to one aspect of the present disclosure. [Figure 2] This flowchart shows an example of an automated process for image-based antibiotic susceptibility testing according to one aspect of the present disclosure. [Figure 3] This is an image of a growth plate with an antibiotic disk, showing the plate before incubation and / or with minimal growth time. [Figure 4] This is an image of a growth plate, such as the growth plate of Figure 3, comprising an antibiotic disk, showing the plate after incubation and / or after a substantial growth time in which bacterial growth has occurred. [Figure 5] This is an image of the minimum inhibition zone for an antibiotic disk on a growth plate after incubation and / or after a substantial growth time in which bacterial growth has occurred. [Figure 6] This is a converted diagram of the image of Figure 5, wherein the image is converted by polar transform to show the minimum inhibition distance from the disk. [Figure 7] This is an image of a plurality of antibiotic disks on a plate. [Figure 7A] Figures 7A to 7F are grayscale images showing mapping of intensity values for pixels of each disk in Figure 7 as a function of radial distance from each disk, wherein shading represents distance from the disk, and intensity decreases as the distance from the disk increases. [Figure 7B] It is a grayscale image showing mapping of intensity values for pixels of each disk in Figure 7 as a function of radial distance from each disk, wherein shading represents distance from the disk, and intensity decreases as the distance from the disk increases. [Figure 7C] It is a grayscale image showing mapping of intensity values for pixels of each disk in Figure 7 as a function of radial distance from each disk, wherein shading represents distance from the disk, and intensity decreases as the distance from the disk increases. [Figure 7D] It is a grayscale image showing mapping of intensity values for pixels of each disk in Figure 7 as a function of radial distance from each disk, wherein shading represents distance from the disk, and intensity decreases as the distance from the disk increases. [Figure 7E] It is a grayscale image showing mapping of intensity values for pixels of each disk in Figure 7 as a function of radial distance from each disk, wherein shading represents distance from the disk, and intensity decreases as the distance from the disk increases. [Figure 7F] Figure 7 is a grayscale image showing the mapping of intensity values ​​for each disk's pixels as a function of the radial distance from each disk. Here, shading represents the distance from the disk, with intensity decreasing as the distance from the disk increases. [Figure 8] This graph maps proliferation (e.g., contrast) as a function of the radial distance from the antimicrobial disk D1 in Figure 4, which is determined according to one or more central pixel values ​​(e.g., intensity, opacity, color, blur, etc.). [Figure 9] This is an image representation of the graph in Figure 8 for a single disk of the plate, with image information related to other disks of the plate and areas of the image outside the plate boundary masked. [Figure 10] This graph shows the growth model function of a growth model that maps pixel intensity values ​​according to the radial distance from the model antimicrobial disk for the modeled bacteria. [Figure 11] Figure 10 is a graph showing the characteristic points of the model at various distances. [Figure 12] These are simulated images generated according to a growth model function (which may be multiple) (for example, a value of 1 or greater (intensity) for each disk) in which no interaction occurs between the modeled antibiotic disk and the modeled bacteria. [Figure 13] These are simulated images generated according to growth model functions (e.g., one or more for each disk) that show interactions between a portion of the modeled antibiotic disk and the modeled bacteria. [Figure 14] This figure shows the pixel characteristic information of an observed image obtained by an image sensor, obtained through contrast analysis comparing multiple images (for example, before and after the growth period). [Figure 15] This figure shows a comparison of pixel characteristic information with a simulated image to identify pixel regions (in an image mask, etc.) for further evaluation and / or analysis. [Figure 16]This figure shows a model of growth that occurs when a disk contains an antibiotic. Here, growth is strongly regulated within a 10 mm radius of the disk, and the growth response to the antibiotic disk is modeled as a function of growth regulation with respect to the distance from the edge of the disk. [Figure 17] This figure shows examples of inhibition zones formed around antibiotic disks against the bacteria Enterococcus faecium, Enterococcus gallinarum, and Leuconostoc species. [Figure 18] This figure shows microbial growth on a culture dish containing an antibiotic disc when the disc does not regulate growth (i.e., there is no inhibition). [Figure 19] This figure shows the various microbial growth regulation effects on various microorganisms brought about by various antibiotic discs. [Figure 20] This figure shows the regulation of microbial growth affected by adjacent disks. [Figure 21A] This figure shows the transformation of growth regulation brought about by one or more antibiotic disks to a two-dimensional representation of regulation as a function of distance from each of the two disks. [Figure 21B] This figure shows the transformation of growth regulation brought about by one or more antibiotic disks to a two-dimensional representation of regulation as a function of distance from each of the two disks. [Figure 21C] This figure shows the transformation of growth regulation brought about by one or more antibiotic disks to a two-dimensional representation of regulation as a function of distance from each of the two disks. [Figure 21D] This figure shows the transformation of growth regulation brought about by one or more antibiotic disks to a two-dimensional representation of regulation as a function of distance from each of the two disks. [Figure 21E] This figure shows the transformation of growth regulation brought about by one or more antibiotic disks to a two-dimensional representation of regulation as a function of distance from each of the two disks. [Figure 21F]This figure shows the transformation of growth regulation brought about by one or more antibiotic disks to a two-dimensional representation of regulation as a function of distance from each of the two disks. [Modes for carrying out the invention]

[0019] This disclosure provides apparatus and methods for identifying and analyzing microbial growth on plate media for antibiotic susceptibility testing. Many of the methods described herein can be fully or partially automated and incorporated, for example, as part of a fully or partially automated laboratory workflow.

[0020] The systems described herein can be implemented in optical systems for imaging microbiological samples. Many such commercially available systems exist but are not described in detail herein. One example is the BD Kiestra® ReadA Compact intelligent incubation and imaging system. Other examples of systems include those described in International Publication No. 2015 / 114121 and U.S. Patent Application Publication No. 2015 / 0299639 (which in whole form form part of this specification). Such optical imaging platforms are known to those skilled in the art and are not described in detail herein.

[0021] Figure 1 is a schematic diagram of an example of an antibiotic susceptibility testing system 100 having a processing module 110 and an image acquisition device 120 (e.g., a camera) for high-quality imaging of plate media. The system typically also accesses one or more data storage media 130, such as data memory, which may include processor control instructions that control the processor to perform, for example, any of the processes or methods described herein. The memory may include antibiotic disk data, antibiotic concentration data, distance data, image data, simulated image data, growth model data, growth model functions, bacterial data, etc. The processing module and image acquisition device can be further connected to and interact with other system components, such as an incubation module (not shown) that incubates the plate media to enable the growth of cultures inoculated onto the plate media. Such connections can be fully or partially automated using a truck system that receives samples for incubation, transports samples to the incubator, and then transports them between the incubator and the image acquisition device.

[0022] The processing module 110 can instruct other components of the system 100 to perform tasks based on the processing of various types of information. The processor 110 may be hardware that performs one or more operations. The processor 110 may be any standard processor such as a central processing unit (CPU), or it may be a dedicated processor such as an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA). Although one processor block is shown, the system 100 may also include multiple processors that may or may not operate in parallel, or other dedicated logic circuits and memories for storing and tracking information about sample containers or AST plates in the incubator and / or image acquisition device 120. In this regard, the processing device can track and / or store various types of information about the AST plate in system 100, including, but not limited to, the location of the AST plate in the system (incubator or image acquisition device, its position and / or orientation therein, etc.), incubation time, pixel information of the acquired image, type of AST plate sample (e.g., concentration of antibiotic(s) and type of bacteria), type of culture medium, and preliminary handling information (e.g., hazardous specimens). In this regard, the processor may be able to automate all or part of the various routines described herein. In one embodiment, processor control instructions that control the routines described herein may be stored in a non-transient computer-readable medium (e.g., a software program).

[0023] Figure 2 is a flowchart illustrating an example of an automated laboratory routine 200 for performing antibiotic susceptibility testing. Routine 200 can be executed by an automated microbiology laboratory system such as Kiestra® Total Lab Automation or Kiestra® Work Cell Automation (both manufactured by Becton, Dickenson & Co). An example of such a system is an interconnected module in which each module is configured to perform one or more steps of routine 200. In this example, routine 200 may be understood to include a growth and imaging process 202 and an image evaluation process 204.

[0024] Proliferation and imaging process 202 In 206, a culture plate containing agar medium is prepared for AST. The culture plate or other suitable container is prepared and the biological sample (e.g., bacteria) is inoculated. The culture plate may be an optically transparent container so that the biological sample and antibiotic disc inside the container can be observed by illuminating them from various angles. Inoculation can follow a predetermined pattern or process for uniformly distributing the bacteria onto the culture medium. Automated methods for inoculating plates are known to those skilled in the art. In 206, one or more antibiotic samples, such as antibiotic wafers in the form of discs, are applied to the culture medium.

[0025] In the 208, a first digital image can be captured by the system's image sensor. It is preferable to capture such an image during or near the initialization of the culture plate, before any growth.

[0026] In step 210, the culture medium is incubated to allow for the growth of the biological sample.

[0027] Next, in 212, for example, additional digital images (one or more) of the culture medium and biological sample can be acquired at predetermined points in time relative to the initialization of the AST plate. Digital imaging of the culture medium may be performed multiple times during the incubation process (e.g., at the start of incubation, at a point during incubation, and at the end of incubation) so that changes in the culture medium can be observed and analyzed. The timing may be based on the properties of the AST plate, such as the type and concentration of the antibiotic(s) being tested. Imaging of the culture medium may include removing the culture medium from the incubator. If multiple images of the culture medium are taken after different incubation times, the culture medium may be returned to the incubator for further incubation between imaging sessions.

[0028] Image evaluation process 204 After imaging, the AST plate is analyzed based on information from the acquired digital image(s). The analysis of the digital image(s) may include the analysis of pixel information contained in the image(s). In some cases, pixel information can be analyzed on a pixel-by-pixel basis. In other examples, pixel information can be analyzed on a block-by-block basis. In even further examples, pixels can be analyzed based on their entire region, and the pixel information of individual pixels within a region can be obtained by combining the information of individual pixels, selecting sample pixels, or using other statistical methods such as statistical histogram operations, which are described in more detail below. Operations described in this disclosure as applying to “pixels” can similarly be applied to blocks or other classifications of pixels, and the term “pixel” is intended to include such applications as herein.

[0029] Typically, the analysis may involve determining whether growth (or lack thereof) is detected with respect to the antibiotic disk in the culture medium. From an image analysis perspective, growth can be detected by identifying the imaged object in the image (based on the difference between the object and its adjacent environment) and then identifying changes within the object over time. As will be described in more detail herein, both these differences and changes are in the form of "contrast." Contrast can be represented, for example, by one or more of the following at the pixel level: intensity value, color value (which may be multiple), grayscale value, opacity value, and blur value.

[0030] For example, in 214, the processor can receive data representing first and second capture images of a particular plate. The first and second images are taken at different points in time. Here, the “first” image is understood to be an image of the plate taken before the second image, but not necessarily the image of the plate immediately preceding the second image. Other images of the plate may be taken between the first and second images. Typically, a considerable incubation time is placed between the first and second images (see Figure 4 showing the disk in Figure 3, which has post-incubation growth G, but shows no growth NG around disks D1, D4, D5, and D6, and partially no growth PNG around D3 and D2). In some cases, the first image may be a pre-incubation image, for example, an image taken in process 208 (see Figure 3 showing disks D1-D6 and no growth on the plate image). In some cases, the first image may be an image taken at least some incubation time later, for example, during process 212.

[0031] In 216, the processor can generate pixel characteristic data for pixels in the second image from a contrast analysis that includes a comparison of the first image data with the second image data. As will be described in more detail herein, this may include a comparison of the second image data with the first image data (e.g., background characteristics) to assess the growth level achieved in the second image data. For example, by ignoring antibiotic disks in the image, the pixel characteristic data may represent indicators of the most different identical plate regions in the latest image when compared with a previous image (e.g., an image before incubation).

[0032] Optionally, such comparisons may be performed on a region-by-region basis for each antibiotic disk on the plate. Such a process may include the detection of antibiotic disks and / or the detection of specific regions (or multiple regions) of an image by marking the antibiotic disks. This may optionally include the evaluation of a normalized version of the image. For example, the processor may scan a first and / or second image or its normalized version by recognizing the features of markers on the antibiotic disks. The recognized marker features can serve as an index for accessing memory, such as a database, containing information about a particular antibiotic disk on the plate. For example, the information may include the size of the disk, the shape of the disk, the location of the disk, the antibiotic name, the concentration, etc.

[0033] For each disk, the process can optionally evaluate the growth level regulation as a function of the distance to the edge(s) of the antibiotic disk. For example, pixel-level contrast information, which becomes pixel characteristic data, can be obtained for the pixels of the second image of the plate relative to the first image of the plate. This characteristic data may include one or more of the following for each pixel of the image being evaluated: intensity value, color value(s), grayscale value, opacity value, and blur value. Optionally, these values ​​may be characterized according to their distance from a particular disk. Such distances are shown in Figure 9. For example, pixels can be grouped as a function of distance in a distance map or the like.

[0034] An example of such pixel characteristic data can be examined with reference to Figures 5 and 6. Figure 5 includes a portion of an image of an AST growth plate showing region R containing one antibiotic disk 502. A non-growth region 504 exists around disk 502. Growth region G exists further away from the non-growth region and disk 502. The image also shows the plate edge 506. Figure 6 shows the polar coordinate transformation of contrast information that can be obtained by comparing the image in Figure 5 with the corresponding region (not shown) of a previous image. In Figure 6, pixels are displayed as intensity values. The polar coordinate transformation transforms the disk 502 portion of the image in Figure 5 into the bright band 608 (highest intensity) on the left side of the image. During the transformation, the non-growth region 504 is transformed into a non-growth band 604 with a lower intensity than the bright band 608. The growth region 610 showing growth G has a higher intensity than the non-growth band 604 and a lower intensity than the bright band 608. Edge detection can optionally be used to detect edges 612 between pixels in the light zone and non-growth zones, and / or between pixels in the non-growth zone and pixels in the growth zone. As shown in the image in Figure 6, edge detection can be used to determine the edges between pixels in the non-growth zone 604 and pixels in the growth zone to indicate or determine the minimum inhibition distance and bacterial growth on the plate for a particular antibiotic disk 502.

[0035] By using polar coordinate transformations for the image pixels, intensity values ​​can be grouped as a function of distance from disk to form a distance map (see, for example, Figure 8). For example, each column of n pixels (x c )(x c ,y0.. n ) can be averaged, where xc represents a constant pixel distance from the disk center or disk edge. Then, each distance x c A distance map can be provided by combining the averages for each. Other distance maps can be similarly formed, for example, by finding the maximum, minimum, average, center, and / or arbitrary percentile values ​​of pixels for each column. In addition to these intensity distance maps, other such maps can be formed using the color(s), grayscale, opacity, and blur values ​​associated with the pixels of the image. Polar coordinate transformations can be used to facilitate distance-dependent pixel characterization as described herein, but in some versions, using polar coordinates allows for characterization of pixels based on distance from disk without transforming contrast data. Polar transformations of image pixels are described in more detail below.

[0036] In some versions, a map (or multiple maps) of plates can be used for further evaluation of a particular AST plate to be subsequently tested. However, the map can also be stored in a database, similarly or alternatively, along with information (e.g., type, concentration, etc.) about the specific antibiotic disk (or multiple disk) on the plate and information about the specific bacteria being tested, thereby allowing the map to be used for modeling, as will be discussed in more detail herein.

[0037] The process for evaluating a specific AST plate image continues as shown in Figure 2, but the system can access growth model data for any subset or all of the disks of the AST plate being tested. Such access may include selecting data from memory, such as a database on the data storage medium 130. Access to memory may be based on the recognition of the markings on the AST plate disks as described above, and / or on the markings of a particular AST plate and its contents associated with that plate. In some versions, access may retrieve data from a database(s), which may include, for example, one or more distance maps, antibiotic concentration data, growth model functions, etc., for one or more disks of the AST plate and optionally for the bacteria being tested. Such growth model data, including maps or functions, may be obtained according to a growth model as described in more detail herein, and may further be based on observational image data, such as map functions from other AST plates. Typically, each disk may have a specific distance map or a set of distance maps associated with the disk. Thus, the distance maps may vary depending on a particular antibiotic, the concentration of the antibiotic on the disk, and the time elapsed since the antibiotic was placed on the disk. Optionally, the distance map of a particular antibiotic disk may also depend on the type and concentration of adjacent antibiotic disks (which may be multiple) in the AST plate.

[0038] Examples of distance maps obtained by growth models can be examined by referring to the map examples in Figures 10 and 11. In the example in Figure 10, the distance map, similar to the map described above, yields an estimate of a radial profile showing bacterial growth as a function of distance from the center of a particular disk. For this particular map, changes in intensity are associated with the distance from the edge of the antibiotic disk (e.g., pixel distance). For example, higher intensity values ​​may be an indicator of modeling a lack of growth, and lower intensity values ​​may be an indicator of modeling bacterial growth. In this example, the intensity values ​​of the function can be considered as the mean intensity values. However, other functions may employ minimum, maximum, median, and / or percentile values. As shown in Figure 11, the function may have characteristic points (indicated by arrows) that can show various features of the relationship between bacteria and a particular concentration of antibiotic from the disk (and potentially adjacent disks) as a function of distance from the antibiotic disk. For example, points in the shown function may indicate features such as the edge of the antibiotic disk, the greatest effect on growth (i.e., the maximum amount of growth regulation and its distance from the disk), and the distance from the disk at which 50% of the growth response regulation occurs.

[0039] In some versions, the difference between the model distance map and the observed distance map can be detected and shown by comparing these model maps with observation maps created for a specific AST plate in process 216. Such a comparison process may include discrepant analysis. In some such versions, the comparison may include the generation of simulated images that follow the model maps. For this reason, in process 220, the system can generate simulated image data using a growth model function with the growth modeling data accessed in 218. The simulated image data then simulates growth (and / or no growth) on the AST growth plate with respect to one or more disks of the plate.

[0040] For an image map (either the target image or an image in a library of images of plates with different patterns of microbial growth regulation for specific microorganisms, antibiotics, and antibiotic concentrations), growth can be simulated as growth regulation observable when AB disks are not present (or are infinitely far from AB disks). For example, C no AB This is the measured contrast produced by the growing organism after incubation time t, in the absence of an AB disk (i.e., the same as growth occurring at an infinite distance from any antibiotic disk (a bandwidth not blocked at all by the disk)).

[0041] The observed contrast is represented by the following sigmoid function:

number

[0042] The following can be used to optimize the match between measured contrast and simulated contrast:

Number

[0043] For example, as shown in Figure 12, an AST plate image can be generated from a series (one or more) of model distance functions, wherein the modeled distance function comprises modeled proliferation. Such functions may each include an opacity value function, a color function, and a blur function that apply a specific value of a map to each pixel of the simulated image according to its distance from the disk and the map function(s). In the example of Figure 12, proliferation of specific bacteria on a specific disk and plate is modeled by a model function (not shown) such that the bacteria are not affected by or resistant to the antibiotic on the disk, so proliferation is shown even around the disk (i.e., there is no inhibition zone). In contrast, Figure 13 shows a simulated image having inhibition zones at several sites, since different inhibitory effects of the antibiotic(s) and / or their concentrations are modeled by the model function for a specific disk.

[0044] Many examples of images showing growth regulation / inhibition of adjacent antibiotic disks can be found at http: / / cdstest.net / manual / plates / . Such images are presented for illustrative purposes in Figures 17–20 and 21A–21F. As can be observed in Figure 17, the dark ring around the antibiotic disk is the inhibition zone. Blurred areas indicate diffuse growth. Brighter areas around the disk, as shown in disk VA5 in plate 13.2.C, indicate a reduction in the inhibition zone (i.e., partial regulation as opposed to complete regulation). When the background extends to the periphery of the disk, as in disks VA5 and TEC15 in plate 13.2D, this image indicates that growth for this microorganism (Leuconostoc) was not regulated / inhibited (at these concentrations) by these antibiotic disks.

[0045] There are three distinctly different scenarios for modeling growth. In the first scenario, growth is not simply regulated by the antibiotic disk. This pattern (i.e., lack of regulation) is shown in Figure 18. When growth is regulated by a single disk, regulation / inhibition of growth manifests in a broad pattern (depending on the specific microorganism and the specific antibiotic). A series of non-limiting examples of such growth regulation are shown in Figure 19.

[0046] Growth regulation influenced by multiple antibiotic disks can also take on various patterns depending on the type of microorganism, the antibiotics, and the distance from the disks. Such patterns are shown in Figure 20. As described above, diverse growth regulation patterns can arise depending on various factors (i.e., the microorganism, the antibiotics contained in the disks, the concentration of antibiotics on the disks, the proximity of disks on the plate, etc.). Therefore, models that simulate the regulation of microbial growth by antibiotics selected to interpret image data will vary in complexity depending on the interactions being modeled.

[0047] To evaluate areas on a plate where microbial growth is regulated by multiple disks, it is useful to transform the pixel intensity as a function of the distance from the two disks. Referring to Figure 21A, an example of regulation provided by disk D1 is shown as a bright ring around the disk. Disk D2 does not provide regulation. The image transformation is shown on the left. The distance from disk D2 is taken on the x-axis, and the distance from disk D1 is taken on the y-axis. The intensity of pixels within the regulation band (700) is shown as 700 in the two-dimensional transformation. There is a band in the transformation that does not represent the combination of distances from the two disks, which is shown by 701. Figure 21B shows the two-dimensional transformation where each of the two disks provides regulation (i.e., one disk does not affect the regulation provided by the other disk). The regulation bands (710, 711) for disks D1 and D2 are shown on the right, and the regulation bands (710, 711) in the two-dimensional transformation are shown on the left.

[0048] A more complex interaction is shown in Figure 21C. Similar to the previous example, adjustment bands 720 and 721 are observed. However, there is clearly an overlapping bandwidth in the adjustments provided by the disk. Since the adjustment bands extend within each other, an overlap in influence appears in the left-hand transformation.

[0049] Referring to Figure 21D, a situation is shown where the disks do not individually induce growth regulation. However, a regulatory zone 730 exists at a certain distance from both disks, indicating that growth regulation is induced not by each disk individually, but by the combination of the two antibiotics. By transforming the image data in this way, the pattern of a particular combination effect of multiple disks can be more easily compared with existing patterns, and the image data from the plate being examined can be interpreted.

[0050] In some cases, the degree or range of adjustment may vary depending on the distance from the disk. Figure 21E shows the first adjustment band (740) and the second adjustment band (741) provided by disk D1. The first band 740 is the band of complete blocking, and the second band is the band of partial blocking. The difference in intensity is also evident from the transformed data on the left, as the pixel intensity changes as a function of the range of adjustment. For simplicity of explanation, no adjustment by D2 is provided in Figure 21E.

[0051] Figure 21F shows a transformation in which the adjustment 750 brought about by D2 absorbs D1, which itself does not bring about adjustment. While it may be difficult to identify this from the three-dimensional image, it is clear from the polar coordinate transformation data that pixels directly adjacent to D1 (shown as region 751 in the transformation data) have an intensity that matches the unadjusted state (i.e., the distance d1 from D1 is zero for many pixels that have an intensity that matches the zero-growth adjustment).

[0052] To make susceptibility testing more efficient, such simulated images may be applied by system 100. For example, in process 222, the system can compare a simulated image generated by the system with an image of an AST plate taken up by the system. Such a comparison can be useful for contrasting the taken-up image and the simulated image by difference analysis. For example, Figure 14 shows an image (or contrast image) taken up by the system that shows growth G and various inhibition regions (RI1, RI2, RI3, RI4, and RI5) around some disks. Such an image may be called an observation image and can be processed in relation to previous images to enhance the contrast showing growth observation, as already discussed. This observation image can then be contrasted with the simulated image. For example, as shown in Figure 15, an image mask 1520 can be generated to highlight areas where there are differences between the observation image (Figure 14) and the simulated image (Figure 13), and to emphasize the difference between modeled growth and observed growth. For example, the image mask can be generated by comparing pixels on a pixel-by-pixel basis. In such an example, if a given pixel (x,y) in the observed image has a value that is the same as or not significantly different from the corresponding pixel in the simulated image (e.g., within the difference threshold), the corresponding pixel in the image mask can be set to a desired value (e.g., black). Otherwise, the corresponding pixel in the image mask can be set to a different value (e.g., white) to highlight the areas with differences. In some cases, a mask image can be generated using pixels that scale the difference such that areas with larger differences are more prominent than areas with smaller differences.

[0053] Difference analysis of models against such observational images can serve as a basis for verifying the model. This can help identify / highlight areas not adequately explained by the model. This can also help simplify the automated detection and further analysis of antibiotic disk interactions. In this regard, as shown with respect to Figures 13, 14, and 15, disks D1, D4, and D6 have the expected antibiotic effects (i.e., the modeling in Figure 13 is approximately equivalent to the observation in Figure 14) and have typical inhibition regions. However, the partial inhibition region RI4 between disks D3 and D2 seen in Figure 14 indicates some unexpected interaction between the antibiotic in D2 and the antibiotic in D3. This results in the difference region 1522 presented within the mask in Figure 15. Such difference images or masks can serve as tools to aid in the visual evaluation of AST plates, such as being presented to the laboratory assistant on a display alongside (e.g., nearby) or overlaid on the observational image. Such evaluations can also serve as a basis for comparison with a library of simulated or image data. Such a comparison allows for an evaluation of the plate without the need for operator intervention to assist in interpretation. Therefore, the system 100 may be configured with a display (e.g., a monitor screen) for presenting such imaging along with an evaluation of the image's meaning (if the system is capable of providing such an evaluation). In this way, specific synergistic effects between two different antibiotics can be detected using the data from the difference images.

[0054] The process of system 100 described herein may also enable incremental modeling of the AST response. For example, the next level in the model can be corrected for further detection by adjusting the model using the difference detected between the model and the observation after each step. Any final difference between the model and the true image may highlight shortcomings of the model for automated interpretation of the AST. For this reason, the system can learn to improve or update the model for automated detection.

[0055] The system can improve one of the following automatic detections: * Growth characteristics (when growth regulation induced by one or more antibiotics is not observed) * Detailed interpretation of growth regulation induced by a given antibiotic * Heterogeneity of microbial populations in the case of various susceptible / tolerant subpopulations or mixed organisms * Complex resistance / susceptibility patterns resulting from the synergistic effects of two or more antibiotics.

[0056] Development of a growth modeling theory In some versions of this technology, growth modeling data, such as the distance maps already discussed, which can be considered as calibrated diffusion maps, can be obtained based on antibiotic concentration, test / diffusion time, and selected bacteria. When developing such maps for growth models, growth regulation can be measured as a function of "detection" concentration depending on the pixel position on the agar plate as a function of distance to each disk (edges of bacterial growth and non-growth locations).

[0057] The "detection" concentration for each antibiotic shall be determined according to the following formula:

number

[0058] For calibrated diffusion maps of each antibiotic, different disks with given antibiotic loads can be used, and their respective growth patterns can be analyzed for a given susceptible organism. In effect, the limits of each inhibition zone correspond to the same "detection" concentration of the antibiotic by the organism. Therefore, the diffusion equation for each antibiotic can be solved using a series of inhibition zone radii and known antibiotic loads.

[0059] Based on the critical concentration and critical diameter r of the antibiotic at the reading time (e.g., t=24 hours), D and k can be estimated.

[0060] For example, using two disks with initial concentrations C1 and C2, and stopping radii r1 and r2, the following can be achieved:

number

[0061] If C1, C2, C3... are serially diluted by a factor of 2, and C2 = 1 / 2C1, C3 = 1 / 2C2..., then the following occurs:

number

[0062] If the minimum inhibitory concentration ("MIC") or C(r,t) of the organism being tested is provided in parallel using an automated testing system that refines the calibration equation, the calibration conclusions can be incorporated into the calibration equation.

number

number

[0063] If k and B for the antibiotic being tested are known, the estimated concentrations of all antibiotics in the culture medium can be estimated as a function of the distance to the source disk (knowing their initial concentration C0), and appropriate distance maps can be created for various disks. Therefore, growth regulation can be estimated (modeled) as a function of the distance to each antibiotic disk. Furthermore, growth regulation can be arbitrarily estimated (modeled) as a function of the estimated antibiotic concentration.

[0064] Image comparison and contrast In the process described above, the determination of whether proliferation is present on the plate disk is made by comparing images (observation at time t0 versus time t0). x This can be done by observation (and / or observation versus simulation) and the contrast between them can be determined. In this regard, bacteria on the AST plate grow over time. The earlier the time since the bacteria were placed on the plate, the fewer bacteria are detected, and as a result the contrast against the background decreases. In other words, smaller colony sizes produce smaller signals, and smaller signals against a constant background result in smaller contrast. This is reflected in the following equation:

number

[0065] Contrast can play a crucial role in identifying objects within an image. Objects can be detected in an image if their brightness, color, and / or texture differ significantly from their surroundings. If an object is detected, the analysis may include identifying the type of object detected. Such identification may also depend on contrast measurements, such as the smoothness of the edges of the identified object, or the uniformity (or lack thereof) of the object's color and / or brightness. This contrast must be sufficiently high to overcome image noise (background signal) detected by the image sensor.

[0066] Human contrast perception (governed by Weber's Law) is limited. Under optimal conditions, the human eye can detect a 1% difference in light level. The quality and reliability of image measurements (e.g., brightness, color, contrast) can be characterized by the signal-to-noise ratio (SNR) of the measurement, where an SNR value of 100 (or 40 dB) is comparable to human detection capabilities, independent of pixel intensity. Digital imaging techniques using high-SNR imaging information and known SNRs from pixel information can enable the detection of colonies even when they are not yet visible to the human eye.

[0067] However, visually observable or perceptible contrast does not necessarily mean that the observed temporal contrast is a reliable determination of microbial growth regulation near the antibiotic disk. Since this method can determine the range of temporal contrast, it is possible to provide users with an indicator of whether the observed temporal contrast is suitable for a reliable determination of microbial regulation by setting a threshold for temporal contrast (i.e., 1% or more, 2% or more, 3% or more, 4% or more, 5% or more, etc.) that is required before providing users with an indicator that a reliable determination of microbial growth can be made using this method. The threshold to be selected will vary depending on the type of nutrient medium (e.g., Mueller-Hinton agar (MH), Mueller-Hinton agar with 5% sheep blood, Mueller-Hinton chocolate agar, etc.) and the type of microorganism (e.g., Neisseria or Haemophilus, Neisseria gonorrhoeae, Escherichia coli, Salmonella, Shigella, Staphylococcus aureus, etc.). Typically, greater contrast is required if brighter colonies form on brighter agar (e.g., MH) or if similarly colored, darker colonies form on darker agar (e.g., MH chocolate agar). Therefore, the range of temporal contrast required for a reliable determination of microbial growth regulation is intended to be set at different thresholds for different agar and microbial combinations. If the measured range of temporal contrast is detected to meet or exceed the set threshold, the user is notified that the plate is ready for analysis. If the measured range of temporal contrast does not meet or exceed the specified threshold, the user is advised to either continue the sample incubation for another cycle or invalidate the sample due to insufficient sample quality.This is advantageous because, while skilled users can perceive contrast and recognize proliferation, they may not necessarily be able to visually and accurately distinguish between 50% or 80% proliferation regulation and observed changes in intensity (when the absolute difference is between 2% and 3% intensity differences), leading to misreading of the inhibition area around a given antibiotic disk.

[0068] This disclosure allows for the collection of contrast in at least two ways: spatially and temporally. Spatial contrast, or local contrast, quantifies the difference in color or brightness between a given region (e.g., a pixel, a group of neighboring pixels) and its surroundings in a single image. Temporal contrast, or temporal contrast, quantifies the difference in color or brightness between a given region in one image and the same region in another image captured at a different time. The equation governing temporal contrast is similar to that governing spatial contrast, as follows:

number

[0069] To maximize spatial or temporal contrast with the target background, the system can capture images using different incident light sources against different backgrounds. For example, lighting from above, below, or the side can be used against a black or white background.

[0070] At a given time, multiple images can be acquired under multiple lighting conditions. Images can be acquired using various light sources that are spectrally different due to the illumination light level, illumination angle, and / or filters placed between the object and the sensor (e.g., red, green, and blue filters). Thus, image acquisition conditions can differ in terms of the position of the light source (e.g., from above, the side, or below), the background (e.g., black, white, any color, any intensity), and the light spectrum (e.g., red channel, green channel, blue channel). For example, the first image can be acquired using illumination from above and a black background, the second image can be acquired using illumination from the side and a black background, and the third image can be acquired using illumination from below and no background (i.e., a white background). Furthermore, a set of various image acquisition conditions can be created using specific algorithms to maximize the spatial contrast used. These or other algorithms may also be useful to maximize temporal contrast by changing the image acquisition conditions according to a given sequence and / or over a certain period of time. Several such algorithms are described in International Publication No. 2015 / 114121.

[0071] Contrast information between two images can be determined. Contrast information can be collected on a pixel-by-pixel basis. For example, the presence of temporal contrast can be determined by comparing a pixel of a second digital image with a corresponding pixel (of the same coordinates) of a first digital image. Furthermore, the presence of spatial contrast can be determined by comparing adjacent pixels of a second digital image with each other or with other pixels known to be background pixels. Changes in the color and / or brightness of pixels indicate contrast, and the magnitude of such changes per image or per pixel (or area of ​​pixels) can be measured, calculated, estimated, or otherwise determined. When determining both temporal and spatial contrast for a given image, the overall contrast of a given pixel in the image can be determined based on a combination of the spatial and temporal contrasts of that given pixel (e.g., mean, weighted mean).

[0072] The proliferation in the second digital image can be identified based on the calculated contrast information. Adjacent pixels in the second digital image with similar contrast information can be considered to belong to the same proliferation. For example, pixels can be considered to belong to the same proliferation target if the difference in brightness between adjacent pixels and their background, or between a pixel and its brightness in the first digital image, is approximately the same (e.g., within a given threshold). As an example, the system may assign a "1" to any pixel with significant contrast (e.g., exceeding a threshold), and then identify a group of adjacent pixels that are all assigned a "1" as proliferation targets. Targets can be given a specific label or mask so that pixels with the same label share certain characteristics. The label may be useful in subsequent processes for distinguishing the proliferation from other targets (e.g., disk) and / or background.

[0073] Identifying objects in digital images may involve segmenting or dividing the digital image into multiple regions (e.g., foreground and background). The goal of segmentation is to change the image to a display of multiple components in a way that facilitates the analysis of those components. Image segmentation is used to locate objects of interest within an image, such as an antibiotic disk.

[0074] The use of such automated processes can enable faster AST testing. Such testing in an automated process can begin immediately after the initial placement of the AST disk, allowing for faster results and reporting. In contrast, such testing in a manual process often requires extra time to complete before the data can be reviewed and reported. Therefore, the automated processes of this disclosure, assisted by the modeling and / or contrast processing described herein, can provide faster testing without adversely affecting the quality or accuracy of the test results.

[0075] Although the present invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the present invention. Therefore, it should be understood that numerous modifications can be made to the exemplary embodiments, and other configurations can be devised without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for testing antibiotic susceptibility in processors, The preparation of a culture plate inoculated with a biological sample, wherein the culture plate comprises a culture medium and at least one antibiotic disc placed thereon, The method involves receiving first and second image data of the culture plate generated by an image sensor, wherein the first image data serves as a pre-growth reference, and the first and second image data each represent first and second acquisition images of the culture plate having a plurality of antibiotic disks, and the first and second acquisition images are captured at different points in time. Identifying non-growth regions adjacent to the antibiotic disk and growth regions further away from the antibiotic disk relative to the non-growth regions, From the comparison of the first and second acquired images, a polar coordinate transformation of the contrast information is obtained, the image of the antibiotic disk is transformed into a first bright band having a first intensity, the image of the non-growth region is transformed into a second bright band having a second intensity, and the image of the growth region is transformed into a third bright band having a third intensity, wherein the second intensity is lower than the first intensity, and the third intensity is higher than the second intensity and lower than the first intensity. Forming a distance map from the first, second, and third bright zones, Determining the minimum stopping distance from the aforementioned distance map, Methods that include...

2. The method according to claim 1, further comprising detecting the edge between the first bright band and the second bright band, or detecting the edge between the second bright band and the third bright band, or detecting the edge between the first bright band and the second bright band and the edge between the second bright band and the third bright band.

3. The method according to claim 2, wherein the first, second, and third bright bands include pixels, and each pixel is associated with an intensity value.

4. The method according to claim 3, wherein the distance map is formed from pixel intensity values ​​as a function of distance from the antibiotic disk.

5. The method according to claim 1, wherein the first, second, and third bright bands include pixels, and each pixel is associated with at least one of a color value, a gray value, an opacity value, and a blur value.

6. The method according to claim 5, wherein the distance map is formed from at least one of a color value, a grayscale value, an opacity value, and a blur value as a function of the distance from the antibiotic disk.

7. The method according to claim 1, further comprising associating the distance map with information about the antibiotic disk and storing the distance map associated with the antibiotic disk information in a database.

8. The method according to claim 7, wherein the antibiotic disk information is either the type of antibiotic or the concentration of the antibiotic.

9. A system for antibiotic susceptibility testing, An image sensor configured to capture an image of a culture plate on which a culture medium is placed with multiple antibiotic discs placed on top, wherein when the culture plate is within the field of view of the image sensor, a biological sample is inoculated onto the culture plate, the image sensor generates first image data and second image data, the first image data and the second image data each represent first and second captured images of the culture plate containing the multiple antibiotic discs, the first and second captured images are taken at different points in time, and the first image data serves as a pre-growth reference. A processor and memory, wherein the processor is configured to receive the first image data and the second image data and access the memory, and the memory stores a distance map. The processor includes, Identifying non-growth regions adjacent to the antibiotic disk and growth regions further away from the antibiotic disk relative to the non-growth regions, From the comparison of the first and second acquired images, a polar coordinate transformation of the contrast information is obtained, the image of the antibiotic disk is transformed into a first bright band having a first intensity, the image of the non-growth region is transformed into a second bright band having a second intensity, and the image of the growth region is transformed into a third bright band having a third intensity, wherein the second intensity is lower than the first intensity, and the third intensity is higher than the second intensity and lower than the first intensity. Forming a distance map from the first, second, and third bright zones, Determining the minimum stopping distance from the aforementioned distance map, A system configured to perform the following actions.

10. The system according to claim 9, wherein the processor is further configured to detect an edge between the first bright band and the second bright band, or an edge between the second bright band and the third bright band, or an edge between the first bright band and the second bright band and an edge between the second bright band and the third bright band.

11. The system according to claim 10, wherein the first, second, and third bright bands include pixels, and each pixel is associated with an intensity value.

12. The system according to claim 11, wherein the distance map is formed from pixel intensity values ​​as a function of distance from the antibiotic disk.

Images