Search

CN-116888643-B - Method for classifying an input image containing particles in a sample

CN116888643BCN 116888643 BCN116888643 BCN 116888643BCN-116888643-B

Abstract

The invention relates to a method for classifying at least one input image containing target particles (11 a-11 f) in a sample (12), characterized in that it comprises the steps of (b) extracting, via a data processing device (20) of a client (2), a vector of characteristics of the target particles (11 a-11 f), the characteristics being digital coefficients, each digital coefficient being associated with one of a set of elementary images, each elementary image representing a reference particle, such that a linear combination of the elementary images weighted by the coefficients approximates a representation of the target particles (11 a-11 f) in the input image, and (c) classifying the input image in dependence on the extracted vector of characteristics.

Inventors

  • PIERRE MAHE
  • Mariam El Azami
  • Elodi Degu Shalmet
  • Zukh Sedgat
  • Quentin Russo
  • ROL FABIAN

Assignees

  • 生物梅里埃公司
  • 拜尔阿斯特公司

Dates

Publication Date
20260505
Application Date
20211019
Priority Date
20201020

Claims (13)

  1. 1. A method for classifying at least one input image representing target particles (11 a-11 f) in a sample (12), the method being characterized in that it comprises the following steps performed by a data processing device (20) of a client (2): (b) Extracting feature vectors of features of the target particles (11 a-11 f), the features being digital coefficients, each digital coefficient being associated with one of a set of base images, each base image representing a reference particle, such that a linear combination of the base images weighted by the coefficients approximates a representation of the target particles (11 a-11 f) in the input image; (c) Classifying the input image in dependence on the extracted feature vector, the method comprising the step (b 0) of unsupervised learning the base images using a database of training images of particles in the sample, wherein the learned base images are those base images that allow a best approximation of the representation of particles in the training images by a linear combination of the base images.
  2. 2. Method according to claim 1, wherein particles (11 a-11 f) are represented in the input image and in each elementary image in a uniform manner, in particular centered and aligned along a predetermined direction.
  3. 3. A method according to claim 2, comprising the step (a) of extracting the input image from a global image of the sample in order to represent the target particles (11 a-11 f) in the unified manner.
  4. 4. A method according to claim 3, wherein step (a) comprises segmenting the overall image to detect the target particles (11 a-11 f) in the sample (12), and then re-cropping the input image according to the detected target particles (11 a-11 f).
  5. 5. The method according to one of claims 3 and 4, wherein step (a) comprises obtaining the overall image from an intensity image of the sample (12) acquired by a viewing device (10).
  6. 6. The method according to one of claims 1 to 5, wherein step (c) is implemented by means of a classifier, the method comprising the step (a 0) of training parameters of the classifier by the data processing device (3) of the server (1) using a training database of classified feature vectors/matrices of particles (11 a-11 f) in the sample (12).
  7. 7. The method of claim 6, wherein the classifier is selected from a support vector machine, a k-nearest neighbor algorithm, or a convolutional neural network.
  8. 8. The method according to one of claims 1 to 7, wherein step (c) comprises reducing the number of variables of the feature vector by means of a t-SNE algorithm.
  9. 9. The method according to one of claims 1 to 8, for classifying a sequence of input images representing the target particles (11 a-11 f) in a sample (12) over time, wherein step (b) comprises obtaining feature vectors of the target particles (11 a-11 f) by concatenating the extracted feature vectors of each input image of the sequence.
  10. 10. A system for classifying at least one input image representing target particles (11 a-11 f) in a sample (12), the system comprising at least one client (2), the client (2) comprising data processing means (20), characterized in that the data processing means (20) are configured to implement: -extracting feature vectors of features of the target particles (11 a-11 f), the features being digital coefficients, each digital coefficient being associated with one of a set of elementary images, each elementary image representing a reference particle, such that a linear combination of the elementary images weighted by the coefficients approximates a representation of the target particles (11 a-11 f) in the input image; classifying the input image in dependence of the extracted feature vector, Wherein the data processing device is further configured to perform unsupervised learning on the base images using a database of training images of particles (11 a-11 f) in the sample (12), wherein the learned reference images are those reference images that allow a best approximation of the representation of particles (11 a-11 f) in the training images by a linear combination of the base images.
  11. 11. The system according to claim 10, further comprising a device (10) for observing the target particles (11 a-11 f) in the sample (12).
  12. 12. Computer program product comprising code instructions for performing the method for classifying at least one input image representing target particles (11 a-11 f) in a sample (12) according to one of claims 1 to 9 when the program is executed on a computer.
  13. 13. A storage medium readable by a computer device, a computer program product thereon comprising code instructions for performing the method for classifying at least one input image representing target particles (11 a-11 f) in a sample (12) according to one of claims 1 to 9.

Description

Method for classifying an input image containing particles in a sample Technical Field The present invention relates to the field of optical collection of biological particles. The biological particle may be a microorganism such as, for example, a bacterium, fungus, or yeast. This may also be a problem with cells, multicellular organisms, or any other type of particles such as contaminants or dust. The invention is particularly advantageously applicable to analysing the status of biological particles, for example with a view to determining the metabolic status of bacteria after the application of antibiotics. The invention makes it possible, for example, to perform an antibacterial spectrum on bacteria. Background Antibacterial spectrum (antibiogram) is a laboratory technique aimed at testing the phenotype of bacterial strains against one or more antibiotics. Conventionally, the antimicrobial spectrum is performed by culturing samples containing bacteria and antibiotics. European patent application number 2 603 601 describes a method of performing an antimicrobial spectrum, which involves visualizing the bacterial status after an incubation period in the presence of an antibiotic. To visualize bacteria, the bacteria are labeled with a fluorescent marker to reveal their structure. Measuring the fluorescence of the marker then makes it possible to determine whether the antibiotic is acting effectively on the bacteria. Conventional procedures for determining antibiotics effective against a given bacterial strain include (e.g., from a patient, animal, food batch, etc.) taking a sample containing the strain, and then sending the sample to an analysis center. When the analysis center receives the sample, the bacterial strain is first cultivated to obtain at least one colony thereof (colony), which takes 24 to 72 hours. Several samples containing different antibiotics and/or different concentrations of antibiotics were then prepared from the colonies, and these samples were then incubated again. After a new incubation period (also 24 to 72 hours), each sample was manually analyzed to determine if the antibiotic was effective. The results are then returned to the physician so that he can apply the most effective antibiotic and/or antibiotic concentration. However, the labelling process is particularly long and complex to perform, and these chemical markers have a cytotoxic effect on bacteria. Thus, this visualization method does not allow for multiple observations of the bacteria during their cultivation, as a result of which the bacteria must be cultivated for a sufficient time, about 24 to 72 hours, to ensure the reliability of the measurement. Other methods of visualizing biological particles use microscopy, allowing non-destructive measurements to be made on the sample. Digital holographic microscopy or DHM is an imaging technique that can overcome the depth of field limitations of conventional optical microscopes. Illustratively, it comprises recording a hologram formed by interference between a light wave diffracted by an object under observation and a spatially coherent reference wave. This technique is described in Myung k.kim's review article entitled "principle and technique of digital holographic microscopy (PRINCIPLES AND techniques of digital holography microscopy)", which published SPIE review volume 1, phase 1, month 1, 2010. More recently, the use of digital holographic microscopy to identify microorganisms in an automated manner has been proposed. Thus, international application WO2017/207184 describes a method for acquisition of particles that correlates simple defocused acquisition with digital focused reconstruction, thereby making it possible to observe biological particles while limiting acquisition time. Typically, this solution allows structural modification of bacteria to be detected after only about 10 minutes of incubation in the presence of antibiotics, and sensitivity (detection of the presence or absence of or pattern of division) to be detected after two hours, unlike the conventional methods described above which may take days. In particular, because the measurement is non-destructive, the analysis can be performed very early in the incubation process without risk of damaging the sample and thus extending the analysis time. The particles can even be tracked over multiple successive images to form a film that represents the progress of the particles over time (since the particles are not destroyed after the first analysis) in order to visualize their behavior such as their speed of movement or their process of cell division. It will therefore be appreciated that this method of visualization gives excellent results. The difficulty is in interpretation of these images or of the film itself, for example, whether it is desirable to conclude about the susceptibility of the bacteria to antibiotics present in the sample (susceptability). Various techniques have been proposed ranging from sim