Improved method for performing fluorescence measurements on a sample - Patent Application 20070122997

Abstract

A method for performing fluorescence measurements on a sample is disclosed, comprising: illuminating the sample using a light source; acquiring at least one fluorescence image of the illuminated sample; and processing the fluorescence image to determine a fluorescence intensity of the sample, wherein processing the fluorescence image to determine a fluorescence intensity of the sample comprises extracting an artifact-free region of interest (ROI) from the fluorescence image by application of a trained model; and determining the fluorescence intensity of the sample from the extracted ROI.

Description

[Technical Field]

[0001] The present disclosure relates to methods and systems for performing fluorescence measurements on a sample, which may be applied in particular to determining the presence of a biological analyte or quantifying said analyte. [Background technology]

[0002] Many in vitro diagnostic systems and other biological applications are known to detect the presence of an analyte or perform quantitative determinations of analyte concentrations via fluorescence measurements. Fluorescence is the ability of a substance to emit light of a specific wavelength after absorbing electromagnetic radiation. Thus, fluorescence measurements are performed by illuminating a sample contained in a reading cuvette with a selected excitation wavelength corresponding to the excitation wavelength of the analyte of interest, and detecting and measuring the excitation-induced fluorescence emission of the sample.

[0003] Detection and processing of the fluorescence signal must be performed accurately to obtain a reliable value for the concentration of the analyte of interest. However, measurements of fluorescence emission can be affected by artifacts such as dust, air bubbles, or even the walls of the cuvette, which may have their own fluorescence contribution.

[0004] From document WO 2014 / 102502 a method for analyzing a sample is known, which comprises detecting a first fluorescent signal from a cuvette before the introduction of a reaction medium into the cuvette and a second fluorescent signal from the cuvette after the introduction of the reaction medium, and comparing the two signals to obtain a signal corresponding to the reaction medium only.

[0005] This method allows removing the fluorescence contribution from the cuvette, but does not guarantee that the acquired signal is not affected by artifacts such as dust or air bubbles. Summary of the Invention

[0006] The present disclosure aims to improve upon the prior art.

[0007] In particular, it is an object of the present disclosure to improve the accuracy of the fluorescence signal acquired on a sample.

[0008] Another object of the present disclosure is to enable correction of fluorescence measurements from contributions caused by artifacts such as air bubbles or dust.

[0009] therefore, - illuminating the sample using a light source; - acquiring at least one fluorescence image of the illuminated sample; - processing the fluorescence image to determine the fluorescence intensity of the sample; Including, processing the fluorescence image to determine the fluorescence intensity of the sample; - Extracting artifact-free regions of interest from fluorescence images by applying a trained model; determining the fluorescence intensity of the sample from the extracted region of interest; characterized in that it comprises A method for performing fluorescence measurements on a sample is disclosed.

[0010] In an embodiment, the method further includes acquiring a plurality of fluorescence images of the sample using different combinations of illumination and exposure, selecting one of the plurality of fluorescence images, and processing the selected fluorescence image to determine a fluorescence intensity of the sample.

[0011] In embodiments, selecting one of the plurality of fluorescent images comprises: - determining the number of saturated pixels for each acquired fluorescence image; selecting fluorescent images in which the number of saturated pixels is below a predetermined threshold; Includes:

[0012] In an embodiment, multiple fluorescence images are acquired under different illumination intensities, and the selected image is the acquired fluorescence image with the highest illumination intensity for which the number of saturated pixels is below a predetermined threshold.

[0013] In embodiments, determining the fluorescence intensity of the sample comprises measuring the intensity of the fluorescent signal on the selected image and deriving the fluorescence intensity of the sample from said intensity and the illumination and exposure conditions of acquisition of the selected image.

[0014] In an embodiment, the method further comprises extracting at least one other region from the fluorescence image that corresponds to at least one category of artifact.

[0015] In an embodiment, determining the fluorescence intensity of the sample from the extracted region of interest includes calculating the intensity of the fluorescence signal in a region of the fluorescence image outside the region of interest by extrapolating the intensity of the fluorescence signal in the region of interest to said region.

[0016] In an embodiment, extracting a region of interest from the fluorescence image comprises performing semantic segmentation on the fluorescence image.

[0017] In an embodiment, extracting the region of interest comprises at least one class corresponding to the absence of artifacts, -The following artifacts: o Air bubbles, o shadow, oDust, o background and at least one other class that corresponds to one of the applying a trained classification model to the fluorescence image, the trained classification model being configured to classify pixels of the fluorescence image according to a plurality of classes, the trained classification model comprising: The region of interest is formed by pixels classified as corresponding to the absence of artifacts.

[0018] In an embodiment, the method further includes a preliminary step of training a classification model by supervised learning on a training database including, for each of a plurality of training fluorescent images, an identification of regions corresponding to artifacts, and each trained fluorescent image is rescaled by a randomly selected factor less than or equal to 1 and cropped to a fixed size.

[0019] In an embodiment, the trained model is a convolutional neural network.

[0020] In an embodiment, the trained model is a convolutional encoder; an intermediate module configured to process the output of the convolutional encoder at multiple different scales; -Decoder and A segmentation multi-scale attention network comprising:

[0021] According to another aspect, a system for performing fluorescence measurements on a sample is disclosed, the sample being contained in a cuvette, the system comprising: an illumination device configured to illuminate the cuvette with at least one determined wavelength; a detection device configured to acquire at least one fluorescence image comprising the fluorescence emission of the sample subsequent to its illumination by the light source; - a computing device configured to process the fluorescence image to determine the fluorescence intensity of the sample; and Equipped with The system is characterized in that it is configured to carry out the method according to the above description.

[0022] The claimed method allows for fluorescence measurements to be performed from regions of interest extracted from a fluorescence image, the regions of interest being free of regions containing artifacts. Fluorescence measurements may then be performed more reliably.

[0023] Extraction of the region of interest may be performed by application of a trained model configured to perform semantic segmentation of the image and classify each pixel or group of pixels of the image as belonging to an artifact-free region or corresponding to a given type of artifact. Artifact detection can then be performed automatically, even when the location of the artifact varies within the image.

[0024] Once a region of interest has been extracted and fluorescence measurements have been performed on said region, it is possible to extrapolate the fluorescence measurements to the portion of the fluorescence image that was excluded in order to obtain a complete fluorescence measurement and therefore to calculate the fluorescence intensity of the sample.

[0025] In an embodiment, the accuracy of the fluorescence intensity measurement is further improved by selecting a fluorescence image from which the region of interest is extracted from among multiple fluorescence images acquired for different illumination or acquisition conditions.

[0026] Other characteristics and advantages of the invention will become apparent from the following detailed description, given by way of non-limiting example with reference to the accompanying drawings, in which: [Brief explanation of the drawings]

[0027] [Figure 1] FIG. 1 illustrates an example of a system for performing fluorescence measurements on a sample. [Figure 2] FIG. 2 is a diagram that schematically illustrates the main steps of a method for performing fluorescence measurements, according to one embodiment. [Figure 3a] FIG. 1 shows a first example of semantic segmentation of a fluorescence image. [Figure 3b] FIG. 1 shows a first example of semantic segmentation of a fluorescence image. [Figure 4a] FIG. 10 illustrates a second example of semantic segmentation of a fluorescence image. [Figure 4b] FIG. 10 illustrates a second example of semantic segmentation of a fluorescence image. [Figure 5] FIG. 1 is a diagram illustrating a schematic representation of the structure of a neural network that can be used to extract regions of interest. [Figure 6a] FIG. 10 is a diagram illustrating an example of a fluorescence signal measured on a region of interest. [Figure 6b] FIG. 10 is a diagram illustrating an example of a fluorescence signal calculated based on a fluorescence signal measured over a region of an image excluded from the region of interest. [Figure 7a] FIG. 10 is a diagram showing a region of interest extracted from a fluorescence image. [Figure 7b] FIG. 10 shows an image reconstructed from fluorescence signals calculated over the entire image based on fluorescence signals measured over a region of interest. [Figure 8] 10A-10C represent comparative fluorescence signals obtained from a cuvette before and after the introduction of an air bubble and with and without application of a method according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0028] Referring to Figure 1, an example of a system for performing fluorescence measurements on a sample is shown. The sample may be of various origins, such as food, environmental, veterinary, clinical, pharmaceutical, or cosmetic origin.

[0029] Among the samples of food origin, non-exhaustive mention can be made of samples of dairy products (yogurt, cheese, etc.), meat, fish, eggs, fruit, vegetables, water, beverages (milk, fruit juice, soda, etc.). Of course, these samples of food origin can also come from sources or more complex meals, or from raw or partially processed raw materials. Food samples can also be derived from animal feedstuffs, such as oil cakes, animal waste, etc.

[0030] As previously mentioned, the biological sample may be of environmental origin and may consist, for example, of a surface sample, a water sample, or the like.

[0031] The samples may also consist of biological samples of clinical, human or animal origin, which may correspond to specimens from biological fluids (urine, whole blood or derivatives such as serum, plasma, saliva, pus, cerebrospinal fluid, etc.), stool (e.g., cholera-induced diarrhea), nose, throat, skin, wounds, organs, tissues, or isolated cells. This list is obviously not exhaustive.

[0032] Generally, the term "sample" refers to a portion or amount, more particularly a small portion or amount, sampled from one or more entities for the purpose of analysis. The sample may have undergone pretreatment, including, for example, mixing, dilution, or even grinding steps, especially if the starting entity is in a solid state. The analyzed sample may contain or is suspected to contain at least one analyte indicative of the presence of a microorganism or disease being detected, characterized, or monitored.

[0033] For the purpose of performing a fluorescence measurement, a sample is contained in a cuvette 4. The cuvette comprises a bottom wall 40 and a side wall 41 extending from the bottom wall to an upper edge 43. The upper edge defines an opening 44 suitable for filling and emptying the cuvette.

[0034] In an embodiment, the bottom wall 40 may include a bottom portion 401 extending perpendicular to the side wall and a beveled wall 402 extending between the bottom portion 401 and the side wall 401, the beveled wall forming a 45° angle with the side wall. By way of non-limiting example, the bottom wall 40 may have a truncated cone shape. The cuvette may be formed from a material suitable for storing liquids and other materials necessary for performing biological analyses. The cuvette 4 may be formed, for example, from a plastic material. Moreover, the cuvette may be transparent to the wavelengths used to illuminate the sample, hereinafter referred to as illumination wavelengths. It may also be transparent to wavelengths emitted by the sample due to fluorescence. For example, the cuvette may be formed from polypropylene, glass, polymethyl methacrylate, polystyrene, polycarbonate, and other optical plastics, depending on the illumination and fluorescence wavelength ranges.

[0035] The system 1 for performing fluorescence measurements includes an illumination device 10 with a light source 11, such as a light-emitting diode (LED). The light source 11 may be any monochromatic light source corresponding to the wavelength of the excitation peak of an analyte, i.e., a chemical molecule used as a marker or sought in a sample. Alternatively, the light source 11 may be capable of generating light of multiple wavelengths, e.g., white light, and the illumination device may further include a filter 12 capable of selecting at least one wavelength of interest corresponding to the excitation peak of a desired molecule. The illumination device may further include an optical element 13 adapted to match the illumination light beam according to an appropriate shape. The optical element may include, for example, at least one optical lens, such as an aspherical lens.

[0036] The system 1 for performing fluorescence measurements also includes a detection device 20 configured to acquire at least one fluorescence image of the sample, the fluorescence image comprising a fluorescence signal emitted by the sample following its illumination by the light source. The detection device includes a detector 21, such as a camera or a highly sensitive CMOS or CCD or other 2D optical sensor. The detection device may further include an optical element 23 adapted to match a light beam emitted by the sample in response to illumination toward the detector. The optical element 23 may include an optical lens, such as an aspheric lens. Furthermore, the detection device may include an optical filter 22 adapted to limit detection to a wavelength of interest or a narrowband spectrum centered on the wavelength of interest, which may typically be the fluorescence wavelength emitted by the analyte.

[0037] By way of non-limiting example, the illuminator may be configured to illuminate the cuvette 4 holding the sample with a beam of incident light forming a 90° angle with the walls of the cuvette, in particular the slanted wall 402 of the cuvette which forms a 45° angle with the side wall 41 of the cuvette. The detector 20 may be configured to collect the light therefrom exiting the cuvette from the slanted wall 402 and form a 90° angle with said wall, such that the axes of the illuminator and the detector are arranged at 90° with respect to each other.

[0038] According to another example, the illuminator and detector may be located on the same side of the cuvette (i.e., the axes of the illumination and detection light form a null angle), or they may be located on opposite sides of the cuvette (i.e., the axes of the illumination and detection light form a 180° angle), since radiation emitted from the sample molecules at 360°, regardless of the illumination direction. In the latter case, the wall on which the illumination light is incident and through which fluorescence is measured may be side wall 41.

[0039] The system 1 for performing fluorescence measurements also includes a computing device 30, which includes at least a computer 31 and a memory 32. The computer 31 is adapted to control the operation of the illumination device and the detection device 20 and to receive fluorescence images acquired by the detection device 20 from the latter. To this end, the computer 31 is connected to the detection device 20 and the illumination device 10 using either a wired or wireless connection. The computer 31 is further adapted to process the acquired fluorescence images to calculate the fluorescence intensity emitted by the sample in response to illumination. To this end, the computer can execute code instructions stored in the memory 32 to perform the methods disclosed below. The computer may include one or more processors, such as a central processing unit (CPU) or a graphical processing unit (GPU). The memory 32 may include, for example, a magnetic hard disk, a solid-state disk, an optical disk, electronic memory, or any type of computer-readable storage medium. The memory further stores a trained model configured to extract a region of interest from the fluorescence image, as described in more detail below.

[0040] Referring to FIG. 2, the main steps of the method for performing fluorescence measurements on a sample will now be described.

[0041] The method includes illuminating 100 the sample with an illumination device 10. The illumination may be performed at a selected wavelength corresponding to the excitation wavelength, i.e., a wavelength suitable for inducing fluorescence, for the molecule of interest to be detected, quantified, or analyzed in the fluorescence measurement. The illumination may be continuous or may involve multiple flashes. The use of flashes may allow for reduced degradation of the sample. The illumination is also performed under defined conditions of illumination intensity. When the light source includes an LED, the excitation intensity may be controlled by driving the input current of the LED.

[0042] The method further includes acquiring 200 at least one fluorescence image of the illuminated sample by the detection device 20. Acquisition 200 may be performed while the sample is illuminated, since fluorescence begins nearly instantaneously. Image acquisition is performed with determined acquisition parameters, including detector gain and integration time, i.e., the time window over which the camera collects light for a single image.

[0043] The acquired fluorescence image includes a plurality of pixels, each pixel corresponding to a respective point on the cuvette, and each pixel is associated with an intensity, which may be expressed, for example, as a grayscale value or an RGB value, and the intensity corresponds to the intensity of the fluorescence signal emitted from the cuvette. The fluorescence image is then processed to determine the fluorescence intensity of the sample from the fluorescence signal in the fluorescence image.

[0044] In an embodiment, the method includes acquiring 200 a plurality of fluorescence images corresponding to different sets of illumination and acquisition parameters. The set of parameters that may be varied to change the exposure may include illumination intensity, sensor gain, and sensor integration time.

[0045] In an embodiment, the method includes acquiring 200 multiple fluorescence images corresponding to different operating points, each defined by an illumination intensity and an exposure, where the exposure is determined by the integration time of the image acquisition and the camera gain. The operating points may be established to provide images with brightness values ​​scaled by a known magnification. Indeed, the fluorescence in the cuvette can span a very wide range (more than four orders of magnitude) depending on the sample type and analyte concentration. As a result, the brightness, i.e., the level of light captured by a pixel during a reading performed at a given exposure and illumination intensity, may exceed the maximum readable value of the detector. Acquiring multiple fluorescence images under different exposures and illumination intensities may allow for the selection of an image with the most suitable brightness, and thus the most information. Therefore, 2 to 10 fluorescence images, preferably 2 to 5, may be acquired at different exposures and illumination intensities.

[0046] According to one embodiment, the number of images and operating points may be selected to cover the range of fluorescence described above. Thus, by way of non-limiting example, four images may be acquired with brightness values ​​scaled by a factor of 4. The inverse ratio between the brightness acquired at each operating point and the brightness acquired at the most sensitive operating point is called the "exposure factor."

[0047] Once the plurality of images have been acquired in steps 100 and 200, the method further comprises step 300 of selecting a fluorescence image from the plurality of acquired images. The selection of the fluorescence image may be based on a condition regarding the number of saturated pixels in the image. A pixel is saturated if its intensity level is the maximum intensity value that the sensor can acquire. Thus, the selection of the fluorescence image may be based on the condition regarding the number of saturated pixels in the image. calculating the number of saturated pixels for each acquired image; selecting the brightest image in which the number of saturated pixels is below a determined threshold; May include:

[0048] A lower threshold may be considered for calculating the number of saturated pixels, for example, a pixel may be considered saturated if its intensity value is above 90% of the maximum intensity value that the sensor can acquire.

[0049] The method then includes processing 400 the fluorescence image to determine the fluorescence intensity of the sample, which may refer to the fluorescence image acquired when one image was acquired, or the fluorescence image selected when multiple images were acquired and one was selected in step 300.

[0050] Processing 400 the fluorescence image includes extracting 410 a region of interest ROI from the fluorescence image that is free from any artifacts, and determining 420 the fluorescence intensity of the sample from the extracted region of interest.

[0051] "Artifact" means a signal that does not correspond to the fluorescence of the analyte being measured. In the context of this disclosure, artifact is meant to include any of the following: - Air bubbles present in the cuvette, which may be caused by the presence of other types of artifacts - Dust particles that may contain particles or flakes or other inhomogeneities typical of the application - Shadows, for example the shadow of an air bubble in a cuvette background, i.e. the areas of the image that do not correspond to the illuminated areas of the cuvette, in particular the parts of the image that correspond to the walls of the cuvette (extending perpendicular to the orientation of the detection device and away from the walls that intersect with the fluorescence emitted by the sample), and the parts of the image that surround the walls of the cuvette.

[0052] Extracting 410 the artifact-free region of interest is performed by application of a trained model. In an embodiment, the trained model is configured to perform semantic segmentation of the image, i.e., to label each pixel of the image with a corresponding class. Here, the trained model: - one class corresponding to the artifact-free area, and -One class corresponding to the artifact The method is configured to label pixels of the image with corresponding classes among at least two classes, including

[0053] In a preferred embodiment, the model may be configured to label pixels of the image with a corresponding class among multiple classes, including one class corresponding to artifact-free regions and multiple other classes corresponding to respective artifacts.

[0054] For example, the class at least one class corresponding to an artifact-free area, -One class for bubbles -intersects with fluorescent light) -One class for dust particles -One class for shadows -One class for the background (everything not included in the other used classes) May include:

[0055] Defining one class per type of artifact allows for greater precision in defining regions of interest.

[0056] 3a and 3b, a first example of semantic segmentation of an image is shown. According to this example, the classes are defined as follows: Two classes are defined for areas of the cuvette that are free of artifacts but are placed at different positions, and one class is defined for bubbles, -One class is defined for shadows, -For "background" one class is defined for all parts not included in the previous classes (including walls).

[0057] A fluorescence image acquired by the detector is shown in Figure 3a. Figure 3b shows the segmentation map output by the model trained on this image. The segmentation map identifies an air bubble located at the bottom of the cuvette and the shadow cast by the bubble in the direction of the light source. Only regions classified as artifact-free are retained to form the region of interest (ROI).

[0058] A second example of semantic segmentation of a fluorescence image is shown in Figures 4a and 4b. In this example, only two classes are defined: one class corresponds to artifacts and the other class corresponds to artifact-free areas. In Figure 4a, the acquired image displays the cuvette wall. In Figure 4b, the region of interest is located in the center of the image, excluding the cuvette wall.

[0059] Before running the model on the fluorescence images, the model is trained by supervised learning on a training database of fluorescence images 90, where each image in the database is annotated according to the desired definition of classes, i.e., artifacts and regions of interest are so indicated for each image. The fluorescence images in the training database may all have the same dimensions. Annotation may be performed manually by an operator. Model training may be performed by the system's computing device 30 or by a separate computing device (not shown).

[0060] Moreover, the model or its training may be designed to process images at different scales.

[0061] According to one example, during training, each image may be randomly rescaled by a factor between 0 and 1 to better train the model to handle the presence of objects and elements of different scales. For example, each training image may be rescaled by a factor randomly selected from the following: 1, 0.8, or 0.6. From the rescaled image, a random crop of constant dimensions is extracted and used to train the model. Thus, the images used to train the model are of constant size, but the size of artifacts may vary.

[0062] Regarding the trained model, it is preferably a convolutional neural network.

[0063] In an embodiment, to process images at different scales, the convolutional neural network: a convolutional encoder section; a multi-scale module, i.e. an intermediate module configured to process the output of the convolutional encoder at multiple different scales; -Convolutional decoder section and It may be equipped with.

[0064] In an embodiment, the neural network may be a segmentation multi-scale attention network (SMANet), the structure of which is shown schematically in FIG. 5 and disclosed in detail in the paper by Simone Bonechi et al., "Weak supervision for generating pixel-level annotations in scene text segmentation," Pattern Recognition Letters, Vol. 138, 2020, pp. 1-7, ISSN 0167-8655. The convolutional encoder portion of this network corresponds to a ResNet neural network, e.g., ResNet50, in which convolutions are replaced by dilation (i.e., dilated) convolutions to expand the receptive field of the neural network. The multi-scale module comprises a pyramid of dilated convolutions with different dilation rates, each followed by a spatial pyramid pool, i.e., multiple parallel pools of different patch sizes. The decoder section comprises two decoder stages with skip connections between the encoder and decoder sections. This neural network developed for scene text segmentation is suitable for applications processing fluorescence images, as images containing text present similar challenges to identifying the size and location of characters, in this case challenges with artifacts.

[0065] Once the region of interest is extracted in step 410, the method includes determining 420 an intensity signal of the sample from said region of interest.

[0066] Determining the intensity signal of the sample from the region of interest comprises extracting intensity values ​​of pixels of the extracted region of interest, said values ​​corresponding to the intensity of the fluorescent signal at each pixel, and inferring from said values ​​by extrapolation or interpolation the intensity of the fluorescent signal across the entire image from which the region of interest was extracted. According to one embodiment, this is performed by fitting a polynomial function to the intensity values ​​of the pixels of the region of interest and estimating from the polynomial function approximations of the intensity values ​​of the pixels removed from the image.

[0067] Referring to Figures 6a and 6b, an example of a polynomial fit performed on a region of interest is shown in three dimensions. The z-axis corresponds to the measured intensity of a pixel whose location on the image is defined by the coordinates (x, y). The intensity of the fluorescence signal in the region of interest is represented in Figure 6a, and the polynomial function fitted to this intensity is represented by a continuous surface in Figure 6b. The dots in this figure represent the original intensity values ​​of the pixels in the region of interest.

[0068] Another example is shown below in the two-dimensional representations of Figures 7a and 7b, where Figure 7a represents a region of interest extracted from the actual fluorescence image (i.e., with the artifacts removed), and Figure 7b represents a reconstructed fluorescence image in which the region corresponding to the artifact has been replaced by pixels whose values ​​are determined according to a matched polynomial function.

[0069] A specific experiment was designed to evaluate the effectiveness of the proposed method. In detail, a set of fixed dilution cuvettes was collected, in which the same cuvette was acquired before and after the introduction of artificial bubbles. The results are shown in Figure 8, where the horizontal axis represents each measured sample and the vertical axis represents the intensity measured for the sample. It is possible to observe how the method allows for a significant improvement in measurements by reducing the influence of artifacts. Four different measurement strategies are shown and compared in this figure. Line A corresponds to a measurement performed in a cuvette without bubbles. Line B corresponds to a measurement performed after an air bubble was inserted into the cuvette, without implementing the method described above. As can be seen, the presence of defects leads to a high distance between the two measurements. Line C corresponds to the measurement of the fluorescence intensity of the sample derived only from the region of interest, i.e., without extrapolating values ​​from the removed image area. Line D corresponds to the measurement of the fluorescence intensity of the sample, in which the values ​​for the defect-containing area are then derived from the values ​​for the region of interest. It can be observed that lines A and D are at substantially the same level and therefore the method disclosed above allows for accurate compensation of errors induced by the presence of defects in the cuvette.

[0070] When a polynomial function is fitted to the intensity values ​​of the region of interest, the fluorescence intensity of the sample is given by - the original intensity values ​​of the pixels in the region of interest, and - Intensity values ​​of pixels inside the artifact approximated by fitting a polynomial function It can be calculated from

[0071] In an embodiment, the fluorescence intensity of the sample is calculated as the average of the intensity values ​​of those pixels that fall within the aforementioned regions (the region originally recognized and the region approximated by the polynomial fit).

[0072] After extraction of the resulting brightness value, the value is multiplied by the exposure factor defined above corresponding to the acquisition of the image to obtain the actual value of the fluorescence intensity. If multiple images are acquired and one image is selected in step 300, the value is multiplied by the exposure factor of the selected image.

Claims

1. 1. A method for performing fluorescence measurements on a sample, comprising: illuminating the sample using a light source (100); acquiring (200) at least one fluorescence image of the illuminated sample; processing the fluorescence image (400) to determine the fluorescence intensity of the sample; Including, processing the fluorescence image to determine a fluorescence intensity of the sample; Extracting 410 artifact-free regions of interest from the fluorescence image by applying the trained model; determining (420) the fluorescence intensity of the sample from the extracted region of interest; characterized in that it comprises method.

2. 10. The method of claim 1, comprising acquiring a plurality of fluorescence images of the sample using different combinations of illumination and exposure, selecting one of the plurality of fluorescence images, and processing the selected fluorescence image to determine a fluorescence intensity of the sample.

3. selecting one of the plurality of fluorescent images; determining the number of saturated pixels for each acquired fluorescence image; selecting fluorescent images in which the number of saturated pixels is below a predetermined threshold; The method of claim 2 , comprising:

4. 4. The method of claim 3, wherein the plurality of fluorescence images are acquired under different illumination intensities, and the selected image is the acquired fluorescence image having the highest illumination intensity for which the number of saturated pixels is below the predetermined threshold.

5. 5. The method of claim 2, wherein determining the fluorescence intensity of the sample comprises measuring an intensity of a fluorescent signal on the selected image and deriving the fluorescence intensity of the sample from the intensity and illumination and exposure conditions of acquisition of the selected image.

6. The method of claim 1 , further comprising extracting at least one other region from the fluorescence image corresponding to at least one category of artifacts.

7. 7. The method of claim 1, wherein determining the fluorescence intensity of the sample from the extracted region of interest comprises calculating an intensity of the fluorescence signal of the region of the fluorescence image outside the region of interest by extrapolating the intensity of the fluorescence signal of the region of interest to the region.

8. The method of claim 1 , wherein extracting the region of interest from the fluorescence image comprises performing semantic segmentation on the fluorescence image.

9. Extracting the region of interest comprises at least one class corresponding to the absence of said artifact; The following artifacts: o bubbles, o shadow, o Dust, o background and at least one other class corresponding to one of applying a trained classification model to the fluorescence image, the trained classification model being configured to classify pixels of the fluorescence image according to a plurality of classes, the trained classification model comprising: the region of interest is formed by the pixels classified as corresponding to the absence of the artifact; 9. The method according to any one of claims 1 to 8.

10. 10. The method of claim 1, further comprising a preliminary step of training the classification model by supervised learning on a training database including, for each of a plurality of training fluorescence images, an identification of the regions corresponding to artifacts, wherein each trained fluorescence image is rescaled by a randomly selected factor less than or equal to 1 and cropped to a fixed size.

11. The method of claim 1 , wherein the trained model is a convolutional neural network.

12. The trained model is a convolutional encoder; an intermediate module configured to process the output of the convolutional encoder at a plurality of different scales; Decoder and A segmentation multi-scale attention network comprising:

12. The method according to any one of claims 1 to 11.

13. A system (1) for performing fluorescence measurements on a sample, the sample being contained in a cuvette (4), the system comprising: an illumination device (10) configured to illuminate the cuvette with at least one determined wavelength; a detection device (20) configured to acquire at least one fluorescence image comprising the fluorescence emission of said sample subsequent to its illumination by said light source; a computing device (30) configured to process the fluorescence image to determine the fluorescence intensity of the sample; Equipped with 13. The system is configured to perform the method according to any one of claims 1 to 12. System (1).

Images