METHOD FOR COUNTING LEUCOCYTES IN A SAMPLE

Description

DOMAINE TECHNIQUE

[0001] The invention is an optical method for analyzing particles contained in a sample. In particular, the invention enables the counting of cells in cerebrospinal fluid. One intended application is to aid in the diagnosis of meningitis. ART ANTERIEUR

[0002] Cytological analysis of cerebrospinal fluid is an important step in the diagnosis of meningitis. It is accepted that a concentration of at least 10 leukocytes per µl can be indicative of meningitis, particularly bacterial meningitis. Currently, such analysis is generally performed under a microscope, but the field of view is limited. Consequently, analyzing a sample is time-consuming, difficult to automate, and requires expensive analytical equipment as well as the intervention of a highly skilled operator. Furthermore, such analysis can be operator-dependent.

[0003] Furthermore, the observation of samples, and in particular biological samples, using lensless imaging has seen significant development over the last ten years. This technique allows a sample to be observed by placing it between a light source and an image sensor, without the need for an optical magnification lens between the sample and the image sensor. The image sensor thus collects an image of the light wave transmitted by the sample. This image, also called a hologram, is formed by interference patterns between the light wave emitted by the light source and transmitted by the sample, and diffraction waves resulting from the diffraction of the light wave emitted by the light source. These interference patterns are sometimes called diffraction patterns.

[0004] Document WO2008090330 describes a device for observing biological samples, specifically cells, using lensless imaging. The device associates each cell with an interference pattern whose morphology allows for cell type identification.

[0005] Lensless imaging thus appears as a simple and inexpensive alternative to a conventional microscope. Furthermore, it allows for the acquisition of images with a significantly larger field of view than that of a microscope.

[0006] US2014 / 0327944 describes a method for identifying particles, such as blood particles, by analyzing holograms acquired by an image sensor positioned in a lensless configuration. The analysis of a hologram generates data, which is then compared to a library of simulated holograms. However, this method faces the same limitations, namely, challenging implementation when particle density is high.

[0007] The hologram acquired by the image sensor can be processed using a holographic reconstruction algorithm to estimate optical properties of the sample, such as absorption or the phase shift of the incident light wave emitted by the light source. Such algorithms are well-established in the field of holographic reconstruction. Given the known distance between the sample and the image sensor, a propagation algorithm is applied, taking into account this distance as well as the wavelength of the incident light wave. An image of an optical property of the sample can then be reconstructed. The reconstructed image can, in particular, be a complex image of the light wave transmitted by the sample, containing information about the sample's absorption or phase shift properties.An example of a holographic reconstruction algorithm is described in the publication Ryle et al, “Digital in-line holography of biological specimens”, Proc. Of SPIE Vol.6311 (2006).

[0008] However, holographic reconstruction algorithms can introduce reconstruction noise into the resulting image, known as "twin image." This is primarily due to the fact that the image formed on the image sensor lacks information about the phase of the light wave reaching that sensor. Consequently, holographic reconstruction relies on incomplete optical information, based solely on the intensity of the light wave collected by the image sensor. Improving the quality of holographic reconstruction is the subject of numerous developments, often employing algorithms frequently called "phase retrieval" algorithms, which allow for an estimation of the phase of the light wave to which the image sensor is exposed. However, such algorithms may require multiple acquisitions, for example, by changing the position of the light source relative to the sample.While reconstruction performance is acceptable, the complexity of implementation limits these algorithms to laboratory use and makes them difficult to apply to everyday applications. An example of such an algorithm is described in Denis L. and AI, "Numerical suppression of the twin image in in-line holography of a volume of micro-objects," Meas. Sci. Technol. 19 (2008). This publication describes a noise suppression algorithm for reconstruction using a binary mask to delimit the reconstructed objects. The mask can be two-dimensional or three-dimensional. The publication also mentions so-called "phase retrieval" algorithms, which recover the phase of a light wave.

[0009] Publication WO2015 / 015023 describes a holographic reconstruction method that also reduces reconstruction noise. The method involves acquiring multiple images at different wavelengths using an image sensor. Each image is then propagated. The method includes calculating the Fourier transform of each propagated image.

[0010] Document US2012 / 0218379, for example, describes a method for reconstructing a complex image of a sample, this complex image containing amplitude and phase information. Such an image provides information that allows for cell identification. Document US2012 / 0148141 applies the method described in document US2012 / 0218379 to reconstruct a complex image of spermatozoa and to characterize their motility, orientation, or certain geometric parameters, such as flagellum size.

[0011] The publication "Autofocusing in digital holographic microscopy" by Langehanenberg P., published in 3D Research, vol. 2, no. 1, March 2011, describes a particle localization method based on the application of a digital focusing algorithm. Such an algorithm, known in itself, allows the position of particles to be determined from an image acquired by an image sensor positioned in a lensless imaging configuration. Multiple images are reconstructed using holographic reconstruction. Each particle is associated with a region of interest, and a sharpness indicator is calculated for each reconstructed image. This sharpness indicator is referred to as the "focus value." A distance is then determined between each particle and the image sensor: this is the distance at which the sharpness indicator reaches a minimum value.

[0012] The inventors sought a method for counting particles in a liquid sample, particularly leukocytes, that would allow for reliable observation, be automatable, and have a large field of view. Unlike the prior art, the method uses a complex image not to track the position of a particle, but to count and identify it. Furthermore, the method is simple, easily automated, and suitable for routine use.

[0013] The publication by Ning Wu et al., "Three-dimensional identification of microorganisms using a digital holographic microscope," Computational and Mathematical Methods in Medicine, vol. 2013, ID 162105, describes a three-dimensional analysis method applicable to holographic images digitally reconstructed from holographic microscopes. This method is applied to the two-class problem of distinguishing between different types of bacteria. EXPOSE DE L'INVENTION

[0014] An object of the invention is a method for determining a quantity of particles of interest contained in a sample, the sample extending along a plane, called the sample plane, the method comprising the following steps: a) illumination of the sample using a light source, the light source emitting an incident light wave propagating towards the sample along a propagation axis; b) acquisition, using an image sensor, of an image of the sample, formed in a detection plane, the sample being disposed between the light source and the image sensor, each image being representative of a light wave, called the exposure light wave, to which the image sensor is exposed under the effect of said illumination; the process being characterized in that it also comprises the following steps: c) from the image acquired in step b), application of a propagation operator to calculate a complex image, called the reference complex image, representative of the sample, in a reference plane; d) determination of a radial position of several particles in a plane parallel to the detection plane, each radial position being associated with a particle; e) calculation of at least one characteristic quantity of the exposure light wave, at each radial position determined in step d), and at a plurality of distances from the detection plane, called reconstruction distances; f) formation of a profile, representing an evolution of the characteristic quantity calculated in step e) as a function of the reconstruction distance, along an axis parallel to the propagation axis and passing through each radial position determined in step d), each profile being associated with a particle;g) based on the profile formed during step f), identification of particles of interest; h) determination of a quantity of particles of interest in the sample from a number of particles of interest identified during step g). ;

[0015] The particles of interest may, for example, be blood cells, such as leukocytes and / or erythrocytes.

[0016] The reference plane can be the sample plane.

[0017] According to one embodiment, in step e), the characteristic quantity comprises the modulus or phase of a complex expression of the exposure light wave. This may be the modulus, the phase, or a combination of the modulus and the phase, for example in the form of a ratio.

[0018] According to one embodiment, step d) comprises the following substeps: d1) application of a propagation operator to the reference image, so as to obtain complex images, called secondary complex images, at different distances from the reference plane, or from the detection plane, along the propagation axis, the secondary complex images and the reference complex image forming a stack of complex images; d2) determination of a radial position of particles from the images of the stack of complex images obtained during substep d1).

[0019] In this case, the radial position of each particle is obtained by forming: a first image in which each point represents a maximum value, at said point and along the propagation axis, of a modulus or phase of the images in the complex image stack; a second image in which each point represents a minimum value, at said point and along the propagation axis, of a modulus or phase of the images in the complex image stack; a differential image representing a difference between the first image and the second image; the radial position ( xi , yi ) of each particle is then obtained by applying a threshold to the differential image.

[0020] Step d) can be implemented from the image acquired by the image sensor in step b), or from the complex reference image calculated in step c).

[0021] According to one embodiment, step e) comprises the following substeps: e1) application of a propagation operator to the reference image, so as to obtain complex images, called secondary complex images, at different distances from the reference plane along the propagation axis, the secondary complex images and the reference complex image forming a stack of complex images; e2) determination of the magnitude or phase of the wave to which the image sensor is exposed from the images of the stack of complex images obtained during substep e1).

[0022] Step (g) may involve classifying one or more profiles associated with each particle. This classification may include applying a threshold to each profile, such that a profile crossing a predetermined threshold is considered representative of a particle of interest. The profile may, for example, represent the phase evolution of the exposed light wave along the propagation axis.

[0023] Step (g) may include determining a reconstruction distance, along the propagation axis, of a maximum or minimum value of a profile, such that a profile for which the maximum or minimum value extends beyond a threshold distance from the sample plane, or the detection plane, is considered representative of a leukocyte. The profile may be a profile representing a change in the magnitude of the exposure light wave, along the propagation axis.

[0024] Step (g) may include calculating a surface extending between the profile and a straight line segment connecting a first notable point and a second notable point on the profile. The first notable point of the profile may be an extremum of the profile, for example, the minimum. The second notable point may be a point at which the profile takes on a predetermined value, at a reconstruction distance greater than that corresponding to the first point.

[0025] According to a preferred mode of implementation: Step a) involves illuminating the sample in two distinct spectral bands; step b) involves acquiring an image of the sample in each of the two spectral bands; step d) is implemented on the basis of the image acquired in step b) in the first spectral band; step e) is implemented on the basis of the image acquired in step b) in the second spectral band.

[0026] The first spectral band can be between 400 nm and 550 nm. The second spectral band can be between 500 nm and 600 nm.

[0027] According to one embodiment, the process comprises the following steps: (g bis) classification of each particle, whose radial position was identified during step d), as being, or not, an erythrocyte according to the profile formed during step f); (h bis) determination of a quantity of erythrocytes in the sample according to the classifications carried out during step g); and possibly: i) determination of a ratio between the quantity of leukocytes determined in step h) and the quantity of erythrocytes determined in step h bis ).

[0028] Preferably, there is no magnifying optics between the sample and the image sensor. Another object of the invention is a device for counting particles of interest, for example leukocytes, in a sample, the device comprising: a light source capable of emitting an incident light wave propagating towards the sample, along a propagation axis; a support, configured to hold the sample between said light source and an image sensor, the image sensor extending along a detection plane; a processor, configured to receive an image of the sample acquired by the image sensor and to implement steps c) to h) of the process as described in this description.

[0029] Preferably, no magnifying optics are placed between the sample and the image sensor when the sample is held on the support.

[0030] Other advantages and features will become clearer from the following description of particular embodiments of the invention, given by way of non-limiting examples, and represented in the figures listed below. FIGURES

[0031] There figure 1 represents an example of a device enabling the implementation of the invention. figure 2A represents the main steps of a process according to the invention. figures 2B à 2G show the results of certain steps described in connection with the figure 2A . There figure 2B shows a portion of an image acquired by the image sensor. figures 2C et 2D represent respectively the modulus and phase of a complex image, called the reference image, reconstructed on the basis of the figure 2A , in a reference plane. The figure 2E This diagram represents a stack of images obtained by propagating the complex reference image in different planes parallel to each other, on either side of the reference plane. figures 2F et 2G illustrate respectively the evolution of the modulus and phase of the wave to which the image sensor is exposed, the modulus and phase being determined from the images in the image stack shown on the figure 2E . There figure 2H represents a preferred variant of a method according to the invention. The figure 3A This illustrates the main steps in a process for obtaining a complex image of the sample, called a reference image, in a reference plane. figure 3B outlines the steps described in relation to the figure 3A . There figure 4A shows thumbnails representing the evolution of an area of ​​interest in the module of each image in the stack shown on the figure 2E , along the axis of light propagation. The area of ​​interest shown on each thumbnail corresponds to the same particle. The figure 4B shows thumbnails representing the evolution, along the axis of light propagation, of a region of interest in the phase of each image of the stack shown on the figure 2E . THE figures 5A et 5B These profiles respectively represent the evolution, along a direction parallel to the axis of light propagation, of the modulus and phase of the light wave to which the image sensor is exposed. The radial position of the profiles is centered on particles, specifically leukocytes or erythrocytes. These profiles were obtained by illuminating the sample with a red spectral band. figures 5C et 5D These profiles respectively represent the evolution, along a direction parallel to the axis of light propagation, of the modulus and phase of the light wave to which the image sensor is exposed, with the radial position of the profiles centered on detected leukocytes or erythrocytes. These profiles were created by illuminating the sample with a blue spectral band. figure 6A represents an area calculated between, on the one hand, the profile and, on the other hand, a straight line segment connecting a minimum point on the profile to a point on the profile at which the profile takes a predetermined value. The area shown in this figure is used to classify particles according to their profile. figure 6B represents the value of such an area on profiles corresponding to erythrocytes and leukocytes. figure 7 shows the results of a leukocyte and erythrocyte count on cerebrospinal fluid samples taken from human patients with various pathologies. EXPOSE DE MODES DE REALISATION PARTICULIERS

[0032] There figure 1 represents an example of a device according to the invention. A light source 11 is capable of emitting a light wave 12, called the incident light wave, propagating towards a sample 10, along a propagation axis Z. The light wave is emitted along a spectral band Δλ.

[0033] Sample 10 is a sample that we wish to characterize. It includes, in particular, a medium 10m in which particles 10a and 10b are suspended. In this example, particles 10a and 10b are blood cells, specifically red blood cells (erythrocytes) and white blood cells (leukocytes), respectively. The medium 10m, in which the particles are suspended, can be a bodily fluid, such as cerebrospinal fluid (CSF), possibly diluted. The sample may contain other particles. A particle is defined, for example, as a cell or a fragment thereof, a microorganism, a spore, or a microbead.

[0034] In this example, sample 10 is contained within a fluidic chamber 15. The fluidic chamber 15 is, for example, a Countess® type fluidic chamber with a thickness e = 100 µm. The thickness e of sample 10, along the propagation axis, typically varies between 10 µm and 1 cm, and is preferably between 20 µm and 500 µm. The sample extends along a plane P 10, called the sample plane, perpendicular to the propagation axis Z. It is held on a support 10s at a distance d from an image sensor 16.

[0035] The distance D between the light source 11 and the sample 10 is preferably greater than 1 cm. It is preferably between 2 and 30 cm. Advantageously, the light source, as seen by the sample, is considered a point source. This means that its diameter (or diagonal) is preferably less than one-tenth, or better yet, one-hundredth, of the distance between the sample and the light source. On the figure 1 The light source is a light-emitting diode (LED). It is generally used in conjunction with a diaphragm 18, or spatial filter. The diaphragm aperture is typically between 5 µm and 1 mm, preferably between 50 µm and 500 µm. In this example, the diaphragm is supplied by Thorlabs under the reference P150S and has a diameter of 150 µm. The diaphragm can be replaced by an optical fiber, one end of which is placed facing the light source 11 and the other end of which is placed facing the sample 10. The device shown in the diagram... figure 1 It also includes a diffuser 17, positioned between the light source 11 and the diaphragm 18. The use of such a diffuser eliminates the need to center the light source 11 relative to the aperture of the diaphragm 18. The function of such a diffuser is to distribute the light beam, produced by an elementary light source 11i, (1≤i≤3) according to a cone with an angle α. Preferably, the diffusion angle α varies between 10° and 80°. The presence of such a diffuser is particularly useful when the light source comprises several elementary light sources, as described below.

[0036] Alternatively, the light source can be a laser source, such as a laser diode. In this case, it is not necessary to use a spatial filter or diffuser.

[0037] Preferably, the spectral emission band Δλ of the incident light wave 12 has a width less than 100 nm. By spectral bandwidth, we mean a full width at half maximum of said spectral band.

[0038] According to one embodiment, the light source 11 comprises several elementary light sources 11i, each capable of emitting an incident light wave 12i in a spectral band Δλi. Preferably, the spectral bands Δλi of the different light sources 11i are different from each other.

[0039] The sample 10 is positioned between the light source 11 and the image sensor 16 mentioned earlier. The latter preferably extends parallel, or substantially parallel, to the plane P10 along which the sample extends. The term substantially parallel means that the two elements may not be strictly parallel, an angular tolerance of a few degrees, less than 20° or 10° being allowed.

[0040] The 16 image sensor is capable of forming an image I 0 according to a detection plan P 0. In the example shown, this is an image sensor with a pixel array, either a CCD or a CMOS type. The detection plane P 0 preferably extends perpendicularly to the propagation axis Z of the incident light wave 12. The distance d between the sample 10 and the pixel matrix of the image sensor 16 is preferably between 50 µm and 2 cm, preferably between 100 µm and 2 mm.

[0041] We note the absence of magnification optics between image sensor 16 and sample 10. This does not preclude the possible presence of focusing microlenses at each pixel of image sensor 16, the latter not having a function of magnifying the image acquired by the image sensor.

[0042] According to the invention, under the effect of the incident light wave 12, the particles present in the sample generate a diffracted wave 13, which produces, at the level of the detection plane P 0, interference, with a portion of the incident light wave 12 transmitted by the sample. Furthermore, the sample can absorb a portion of the incident light wave 12. Thus, the light wave 14, transmitted by the sample, and to which the image sensor 20 is exposed, designated by the term "exposure wave", can comprise: a component 13 resulting from the diffraction of the incident light wave 12 by each particle of the sample; a component 12' resulting from the absorption of the incident light wave 12 by the sample.

[0043] These components form interference patterns in the detection plane. Therefore, the image acquired by the image sensor contains interference patterns (or diffraction patterns), each interference pattern being generated by a particle of the sample.

[0044] A processor, for example a microprocessor, is capable of processing each image I0 acquired by the image sensor 16. In particular, the processor is a microprocessor connected to a programmable memory 22 in which a sequence of instructions is stored to perform the image processing and calculation operations described in this description. The processor can be coupled to a screen 24 allowing the display of images acquired by the image sensor 16 or calculated by the processor 20.

[0045] An image IThe image acquired by the image sensor 16, also called a hologram, does not provide a sufficiently accurate representation of the observed sample. As described in relation to the prior art, a holographic propagation operator h can be applied to each image acquired by the image sensor in order to calculate a quantity representative of the light wave 14 transmitted by the sample 10 and to which the image sensor 16 is exposed. This process, known as holographic reconstruction, makes it possible, in particular, to reconstruct an image of the magnitude or phase of the exposed light wave 14 in a reconstruction plane parallel to the detection plane. P 0, and especially in the plan P 10 along which the sample extends. To do this, a convolution of the image is performed I0 acquired by the image sensor 16 by a propagation operator h. It is then possible to reconstruct a complex expression A of the light wave 14 at any point with coordinates ( x, y, z ) of space, and in particular in a reconstruction plan P z located at a distance | z | of the image sensor 16, called the reconstruction distance, this reconstruction plane preferably being the plane of the sample P 10, with: A ( x,y,z ) = I 0 ( x,y,z ) * h * denoting the convolution product operator.

[0046] In the remainder of this description, the coordinates ( x , y ) denote a radial position in a plane perpendicular to the propagation axis Z. The z coordinate denotes a coordinate along the propagation axis Z.

[0047] The complex expression Ais a complex quantity whose argument and modulus are respectively representative of the phase and intensity of the light wave 14 of exposure to the image sensor 16. The convolution product of the image I 0 by the propagation operator h allows for obtaining a complex image A z representing a spatial distribution of the complex expression A in a plane, called the reconstruction plane P z extending to a coordinate z of the detection plan P 0. In this example, the detection plan P 0 has the equation z = 0. The complex image A z corresponds to a complex image of the sample in the reconstruction plane P z It also represents a two-dimensional spatial distribution of the optical properties of the exposure wave 14.

[0048] The propagation operator hhas the function of describing the propagation of light between the image sensor 16 and a point with coordinates ( x, y, z ), located at a distance | z | of the image sensor. It is then possible to determine the module M ( x,y,z ) and / or the phase φ ( x,y, z ) the light wave 14, at this distance | z |, called reconstruction distance, with: M x y z = abs A x y z φ x y z = arg A x y z

[0049] The operators abs And arg respectively denote the module and the argument.

[0050] The propagation operator is, for example, the Fresnel-Helmholtz function, such that: h x y z = 1 jλz e j 2 π z λ exp jπ x 2 + y 2 λz .

[0051] In other words, the complex expression A of the light wave 14, at every point with coordinates ( x,y, z ) of space, is such that: A ( x,y, z) = M ( x,y, z ) e jφ ( x , y , z )< (3). It is possible to form images M z And φ z representing respectively the modulus or phase of the complex expression A in a plane P z located at a distance | z | of the detection plan P 0, with M z = mod ( A z ) And φ z = arg( A z ).

[0052] The inventors have developed a method for counting leukocytes in a sample, this method being described in connection with the figures 2A à 2G The sample may, in particular, be or include cerebrospinal fluid obtained by lumbar puncture, with the leukocyte count performed for the purpose of diagnosing meningitis. Indeed, as described in connection with the prior art, a concentration of leukocytes in the cerebrospinal fluid exceeding a certain threshold can be indicative of meningitis. It is accepted that a concentration of 10 leukocytes per microliter constitutes a threshold concentration beyond which meningitis should be suspected. The main steps of the counting procedure according to the invention are: the acquisition, by the image sensor, of an image I0 of the sample in one or more spectral bands; the determination of radial coordinates of particles detected on the acquired image, the term "radial" meaning in a plane parallel to the detection plane; from the acquired image, the calculation of a characteristic quantity, for example the modulus or the phase, of the light wave of exposure 14, at different distances from the sample, called reconstruction distances; the formation of a profile representing an evolution of the characteristic quantity as a function of the reconstruction distance, each profile being associated with a particle; from one or more profiles associated with each particle, the identification of leukocytes as particles of interest; the counting of the leukocytes thus identified, so as to estimate a quantity of leukocytes in the sample.

[0053] There figure 2A represents a first example of a process implemented, the steps of which are described below.

[0054] Step 100: Image Acquisition I 0 of sample 10 by image sensor 16, this image forming a hologram. The figure 2B represents a portion of such an image. One of the advantages of the lensless configuration, shown on the figure 1 The field of view is the large observed area, allowing for the simultaneous examination of a large sample volume. This enables the simultaneous observation of multiple particles. The observed area depends on the size of the image sensor, being slightly smaller than its detection area due to the spacing between the sensor pixels and the sample. The observed area is generally greater than 10 mm², and is typically between 10 mm² and 50 mm², which is significantly larger than with a microscope.

[0055] Step 110: Formation of a complex image called the initial image A 0 k = 0 of sample 10 in the detection plane P0. During this step, an initial image is defined A 0 k = 0 from sample 10, from image I 0 acquired by image sensor 16. This step is an initialization of the iterative algorithm described later in connection with step 120, the exponent k designating the rank of each iteration. The module a 0 k = 0 of the initial image A 0 k = 0 perhaps obtained by applying the square root operator to the image I 0 acquired by the image sensor, in which case a 0 k = 0 = I 0 In this example, we perform a normalization of the acquired image. I 0 by a term representing the intensity of the light wave 12 incident on the sample 10. The latter can be, for example, the square root of a mean I 0 of the image I 0, in which case each pixel I 0 ( x, y ) of the acquired image is divided by said average, such that a 0 k = 0 = I 0 x y I 0 ¯ The phase φ 0 k = 0 of the initial image A 0 k = 0 is either considered zero at each radial coordinate (x,y), or predetermined according to an arbitrary value. Indeed, the initial image A 0 k = 0 results directly from the image I 0 acquired by the image sensor 16. However, the latter does not contain information relating to the phase of the light wave 14 transmitted by the sample 10, the image sensor 16 being sensitive only to the intensity of this light wave.

[0056] Step 120: Calculating a complex image A ref , said reference, of sample 10 in a reference plane P ref , This reference image is a complex image containing phase and amplitude information of the light wave 14 to which the image sensor 16 is exposed. The reference plane is advantageously a plane perpendicular to the propagation axis Z, and / or parallel to the detection plane. P 0. This is preferably the sample planP 10. Step 120 is performed by applying the propagation operator h, described previously, to the initial complex image A 0 k = 0 However, applying the propagation operator to the initial complex image can result in a reference image. A ref affected by significant reconstruction noise, frequently referred to as a "twin image". To obtain a usable complex reference image while minimizing reconstruction noise, iterative algorithms can be implemented. One such algorithm is described below.

[0057] The complex image A ref is designated as a reference image, as it serves as the basis for forming profiles on which the particles in the sample are characterized. figures 2C et 2D show respectively an image of the module M ref and the phase φ ref of the complex reference image A ref .

[0058] The coordinates z ref of the reference plan P ref is determined either a priori, particularly when the position of the sample is controlled relative to the image sensor 16, or by means of autofocus, based on a sharpness criterion of the reference image A ref , The image is sharper when the reference plane corresponds to the plane in which the particles are located. The sharpness criterion can be applied to the image of the module M ref or the phase of the image φ ref reference.

[0059] Step 130: Application of a propagation operator h to the complex reference image A ref in order to calculate complex images A ref,z , called secondary, along the propagation axis Z. During this step, the complex reference image A ref is propagated along a plurality of reconstruction distances z using a propagation operator has previously defined, so as to have a plurality of complex, or secondary, images, A ref,z reconstructed at different distances z of the reference plan P ref . Thus, this step involves determining a plurality of complex images A ref,z such as: A ref , z = A ref ∗ h z avec z min ≤ z ≤ z max .

[0060] The values z min And z max are the minimum and maximum coordinates, along the Z-axis, between which the reference complex image is propagated. Preferably, the complex images are reconstructed along a plurality of z-coordinates between the sample 14 and the image sensor 16. The inventors considered it preferable to obtain secondary complex images on either side of the reference plane. P ref , in such a way that z min ≤ z ref ≤ z max . Contrary to the acquired image I 0 by the image sensor 16, or to the initial complex image A 0 k = 0 The complex reference image accurately describes the exposure light wave 14, particularly its phase. Therefore, it is estimated that the secondary images A ref,z , obtained by propagation of the reference image, form a good descriptor of the propagation of the light wave of exposure 14 along the propagation axis Z.

[0061] Preferably, two adjacent reconstruction planes are spaced apart by a fine mesh, for example between 5 µm and 50 µm, and for example 25 µm. This is a local propagation, as it is carried out over a distance between 250 µm and 2 µm on either side of the reference plane. P ref , for example at ± 500 µm. Based on a reconstruction at a distance of 500 µm on either side of the reference plane P ref , and a distance of 25 µm between two adjacent planes, the complex reference image is propagated A ref according to forty reconstruction plans P ref,z , in order to form as many complex secondary images as possible, A ref,z . There figure 2E illustrates a stack of images, formed by the complex reference image A ref as well as various secondary images A ref,z obtained by local propagation of the complex reference image A ref . This stack of images can, in particular, be used to detect the presence of particles i in each radial coordinate ( x i , y i ) (step 140) and to obtain a representative profile of the modulus and / or phase of the light wave 14 along a propagation axis passing through each detected particle (step 150).

[0062] Step 140: Particle detection in the sample. This step aims to determine a radial position ( x i , y i ) of each particle i to be characterized, either by using the reference complex image A ref , either the image I0 acquired by the image sensor 16, either by using the image stack obtained during step 130, the latter option constituting the preferred embodiment.

[0063] To do this, at each radial coordinate ( x , y ) of the reference plane, the maximum and minimum values ​​of the modulus are determined in the stack of complex images formed during step 130. This forms an image of the maximum modulus M max , each of whose coordinates M max ( x , y ) gathers the maximum values ​​of the modulus, in the image stack, at the radial coordinate (x,y). An image of the minimum modulus is also formed. M min , each of whose coordinates M min ( x , y ) gathers the minimum values ​​of the modulus, in the image stack, at the radial coordinate ( x, y ) .From the image of the maximum modulus and the image of the minimum modulus, we establish a differential image of the modulus Δ M, with Δ M ( x, y ) = M max ( x, y ) - M min (x, y ) . Each radial coordinate ( x i , y i ) corresponds to the center of a particle i when the value of the differential image of the modulus is greater than a certain threshold, i.e. Δ M ( x i , y i ) ≥ M Th .

[0064] There figure 4A represents a series of thumbnails, each thumbnail corresponding to the modulus of an image in the image stack, within a region of interest. The thumbnails represent the same region of interest, centered on a particle. A significant variation in the modulus value is observed between the different thumbnails. Thus, in the presence of a particle i to a radial coordinate ( x i , y i ), the value of the differential image, at this coordinate, is high ΔM ( x i , y i ) . The thresholding described above allows us to identify the radial coordinates ( x i , y i ) corresponding to a particle, and in particular designating the center of each particle.

[0065] Based on the phase of each image in the image stack, we can similarly form an image of the maximum phase φ max and an image of the minimum phase φ min . We then establish a differential image of the phase Δ φ, with Δ φ ( x, y ) = φ max ( x , y ) - φmin(x, y ) . Each radial coordinate ( x i , y i ) corresponds to the center of a particle i when the value of the differential phase image is greater than a certain threshold, that is Δφ(x i , y i ) ≥ φ Th . There figure 4B represents a series of thumbnails, each thumbnail representing the phase of an image from the image stack, in a region of interest corresponding to the same particle as on the figure 4A . In the same way as for the modulus, in the presence of a particle, a significant variation in the phase value is observed between the different vignettes.

[0066] Thus the radial position ( x i , y i The ) of each particle is determined by retaining the points of the differential image of the modulus Δ M or the differential image of the phase Δ φ whose values ​​are greater than the predetermined thresholds. A minimum distance criterion between two adjacent radial positions can also be applied to prevent two radial positions too close to each other from corresponding to the same particle. At the end of step 140, the radial position ( x i , y i ) of each particle as well as the number of particles detected.

[0067] According to one variant, the detection of the radial position ( x i , y i ) of each particle is performed by considering an image from the image stack, whether it is the reference image A ref obtained during step 120 or from an image A ref,z of the image stack. On the image of the modulus or phase of the image under consideration, each particle takes the form of a spot, as can be seen on the figures 2C et 2D The application of image processing techniques makes it possible to determine the centroid coordinates of each spot, these coordinates forming the different radial positions ( x i , y i ) to consider.

[0068] Step 150: Formation of a profile associated with each particle. From the complex reference image A ref and each secondary complex image A ref,z a characteristic quantity of the light wave 14 is estimated at each radial position ( x i , y i ) previously selected during step 140, and at a plurality of distances z reconstruction of the reference plan P ref (or the detection plan) P 0), then a profile is formed representing the evolution of the characteristic quantity as a function of z, along the propagation axis Z. The characteristic quantity can notably be established from the magnitude and phase of each particle. The figures 2F et 2G represent respectively a profile of the modulus M(z) and a profile of the phase φ ( z ), at the same radial position ( x i , y i ), each profile being obtained from the images in the image stack formed during step 130, by performing an interpolation between the coordinates of two adjacent images.

[0069] Step 160: Identification of each particle from the profiles formed during step 150. Preferably, a database of standard profiles formed during a training phase using known standard samples is available. Identification is then performed by comparing or classifying the profile associated with each particle, based on standard profiles. This identification allows us to determine whether each particle corresponds to a particle of interest, which we wish to count.

[0070] Step 170: Counting. The particles of interest identified in the previous step are counted. This allows us to estimate the quantity (number or concentration) of particles of interest in the sample.

[0071] The inventors found that it was preferable for step 140 to be implemented based on a reference image. A ref ( Δλ 1) reconstructed from an imageI The image sensor acquires a spectral band (Δλ1) in the first spectral band (Δλ1) between 400 nm and 550 nm, and preferably between 400 nm and 500 nm, which corresponds to the blue wavelengths of the visible spectrum. In such a spectral band, particles stand out more clearly from the background of the image, whether it is the modulus or phase image, allowing for better particle detection and more precise localization. This enables accurate definition of each radial position, especially when the particles are leukocytes or erythrocytes. They also found that step 150 can yield better results when based on a reference image. A ref (Δλ 2 ) reconstructed from an image I 0 (Δλ 2 ) acquired by the detector in a second spectral band (Δλ 2 ) between 600 nm and 700 nm, which corresponds to the red wavelengths of the visible spectrum.

[0072] Thus, according to a preferred embodiment, step 140 is performed using a complex reference image A ref (Δλ 1 ) calculated on the basis of an image I 0 (Δλ 1 ) acquired by the image sensor in a first spectral band Δλ 1 while step 150 is implemented using a complex reference image A ref (Δλ 2 ) calculated on the basis of an image I 0 (Δλ 2 ) acquired by the image sensor in a second spectral band Δλ 2, different from the first spectral band. In this example, the first spectral band Δλ 1 is between 400 and 550 nm, and the second spectral band Δλ 2 is between 600 nm and 700 nm. This combination of spectral bands has proven particularly suitable for observing blood cells such as leukocytes or erythrocytes. This embodiment is described in connection with the figure 2H .

[0073] One of the important points of this algorithm is step 120, which involves training the complex reference image. A ref from an image I 0 acquired by the image sensor 16. This complex image is formed from an iterative algorithm, each iteration involving the propagation of a complex image representative of the sample of the detection plane P 0 towards the reference plane, in this case the sample plane P 10, then an update of the phase value of each pixel of the complex image of the sample 10 in the detection plane P 0. Several algorithms are potentially usable. It is important that the algorithm exhibits good reconstruction performance, with the reconstructed image having a good signal-to-noise ratio, while being simple to implement, without requiring too many acquisitions, and without moving the light source relative to the sample.

[0074] It is possible to obtain a complex reference image by simply applying the propagation operator to the image. I 0 acquired by the image sensor. However, this results in a reconstruction affected by significant noise, as previously mentioned.

[0075] According to one possibility, the sample is illuminated successively or simultaneously in different spectral bands Δ λ and we acquire, in the detection plane P 0, an image I 0 (Δ λ ) representative of each spectral band. The algorithm makes it possible to obtain a complex image A ref (Δ λ ) of sample 10, in the reference plane, in each spectral band Δλ. The complex images thus obtained can be combined, for example by averaging, at each pixel, their magnitude and phase, which makes it possible to form the reference image A ref . Alternatively, the complex reference image is a complex image A ref (Δ λ ) in a spectral band Δ λ This algorithm was described in the publication "Wide-Field Lensfree Imaging of Tissue Slides" by Morel, A. Delon, P. Blandin, T. Bordy, O. Cioni, L. Hervé, C. Fromentin, J. Dinten, and C. Allier, in Advanced Microscopy Techniques IV; and Neurophotonics II, edited by E. Beaurepaire, P. So, F. Pavone, and E. Hillman, Vol. 9536 of SPIE Proceedings (Optical Society of America, 2015), as well as in patent application FR1554811 filed on May 28, 2015, and more specifically in steps 100 to 500 described in that application. It has been shown that using two or three different spectral bands allows for good reconstruction performance.

[0076] Another preferred approach is to reconstruct a complex reference image based on an image acquired from the sample when it is illuminated in a single spectral band Δλ. The complex reference image can be obtained using an algorithm as described in patent application FR1652500 filed on March 23, 2016. According to this second algorithm, the main steps in the formation of the complex reference image are described below, in connection with the figures 3A et 3B This algorithm is based on a single acquisition, which simplifies its implementation. It is an iterative algorithm, with steps 120 to 125 described below being repeated, where k denotes the rank of an iteration. Step 121: Propagation of the detection plane to the reference plane

[0077] During this step, an image is formed in the detection plane. P 0. During the first iteration, this is the initial image A 0 k = 0 , described in connection with step 110. In the other iterations, it is the image A 0 k − 1 resulting from the previous iteration. The image formed in the detection plane P 0 is propagated in a reference plane P ref , by applying a propagation operator h as previously described, in order to obtain a complex image A ref k , representative of sample 10, in the reference plane P ref . Propagation is achieved by convolution of the image A 0 k − 1 by the propagation operator h_ zref , such that: A ref k = A 0 k − 1 ∗ h − zref ,

[0078] The index -zref represents the fact that propagation occurs in a direction opposite to the propagation axis Z. This is called backpropagation. Step 122: Calculating a multi-pixel indicator

[0079] During this step, a quantity is calculated ε k< ( x, y ) associated with each pixel of a plurality of pixels ( x, y ) of the complex image A ref k , and preferably in each of its pixels. This magnitude depends on the value A ref k x y of the image A ref k , or its module, at the pixel level ( x , y ) to which it is calculated. It can also depend on a dimensional derivative of the image at that pixel, for example, the modulus of a dimensional derivative of that image. In this example, the quantity ε k< ( x, y) associated with each pixel is a modulus of a difference in the image A ref k , in each pixel, and the value 1. Such a quantity can be obtained according to the expression: ε k x y = A ref k x y − 1 A ref k x y − 1 * = A ref k x y − 1

[0080] Step 123: Establishing a noise indicator associated with the image A ref k .

[0081] In step 122, quantities are calculated ε k< ( x, y ) in several pixels of the complex image A z k These quantities can form a vector. E k< whose terms are the quantities ε k< ( x , y ) associated with each pixel ( x , y In this step, an indicator, called a noise indicator, is calculated from a norm of the vector E k< In general, a norm is associated with an order, such that the norm || x || p order p of a vector x dimension n coordinates ( x 1 ,x 2, ... . x n ,) is such that: x p = ∑ i = 1 n x i p 1 / p , with p ≥ 0. In this case, we use a first-order norm, in other words p = 1. During this step, the magnitude ε k< ( x, y ) calculated from the complex image A z k , at each pixel ( x , y ) of the latter, is summed in such a way as to constitute a noise indicator ε k< associated with the complex image A ref k .

[0082] So, ε k = ∑ x y ε k x y

[0083] An important aspect of this step is to determine, in the detection plan P 0, phase values φ 0 k x y of each pixel of the image A 0 k in the sample plane, allowing a reconstructed image to be obtained during a subsequent iteration A ref k + 1 including the noise indicator ε k +1< is less than the noise indicator ε k< .

[0084] During the first iteration, we only have relevant information about the intensity of the light wave 14, but not about its phase. The first reconstructed image A ref k = 1 in the reconstruction plan P ref is therefore affected by significant reconstruction noise, due to the lack of relevant information regarding the phase of the light wave 14 in the detection plane P 0. Therefore, the indicator ε k =1< is high. During subsequent iterations, the algorithm progressively adjusts the phase φ 0 k x y in the detection plan P 0, so as to gradually minimize the indicator ε k< .

[0085] The image A 0 k in the detection plane is representative of the light wave 14 in the detection plane P 0, both in terms of its intensity and its phase. Steps 120 to 160 aim to establish, iteratively, the value of the phase φ 0 k x y of each pixel of the image A 0 k minimizing the indicator ε k< , the latter being obtained from the image A ref k obtained by image propagation A 0 k − 1 in the reference plane P ref .

[0086] The minimization algorithm can be a gradient descent algorithm, or a conjugate gradient descent algorithm, the latter being described below.

[0087] Step 124: Adjusting the phase value in the detection plane.

[0088] Step 124 aims to determine a value for the phase φ 0 k x y of each pixel of the complex image A 0 k in order to minimize the indicator ε k +1< resulting from a propagation of the complex image A 0 k in the reference plane P ref , during the next iteration k +1. For this, a phase vector φ 0 k is established, of which each term is the phase φ 0 k x y of a pixel ( x , y ) of the complex image A 0 k The dimension of this vector is (Npix, 1), where Npix represents the number of pixels considered. This vector is updated during each iteration, using the following update expression: φ 0 k x y = φ 0 k − 1 x y + α k p k x y Or : α k< is an integer, designated by the term "step", and representing a distance; p k< is a direction vector of dimension (N pix, 1), of which each term p ( x, y ) forms a direction of the gradient ∇ ε k< of the indicator ε k< .

[0089] This equation can be expressed in vector form, as follows: φ 0 k = φ 0 k − 1 + α k p k

[0090] It can be shown that: p k = − ∇ ε k + β k p k − 1 Or : ∇ ε k< is a gradient vector, of dimension (N pix, 1), where each term represents a variation of the indicator ε k< as a function of each of the degrees of freedom of the unknowns of the problem, that is to say the terms of the vector φ 0 k . ; p k -1< is a direction vector established during the previous iteration; β k< is a scaling factor applied to the direction vector p k -1< .

[0091] Each term ∇ε k< ( x, y ) of the gradient vector ∇ ε is such that ∇ ε k r ′ = ∂ ε k ∂ φ 0 k r ′ = Im A 0 k * r ′ ⋅ A ref k − 1 A ref k − 1 ∗ h z r ′ where Im represents the imaginary part operator and r' represents a coordinate ( x, y) in the detection plan.

[0092] The scale factor β k< can be expressed in such a way that: β k = ∇ ε k . ∇ ε k ∇ ε k − 1 . ∇ ε k − 1

[0093] The step α k< may vary depending on the iterations, for example between 0.03 during the first iterations and 0.0005 during the last iterations.

[0094] The update equation allows us to obtain a fit to the vector φ 0 k , which leads to an iterative update of the phase φ 0 k x y in each pixel of the complex image A 0 k This complex image A 0 k , in the detection plane, is then updated by these new values ​​of the phase associated with each pixel. Step 125: Reiteration or exit of the algorithm.

[0095] Until a convergence criterion is met, step 125 consists of reiterating the algorithm, by a new iteration of steps 121 to 125 based on the complex image. A 0 k updated in step 124. The convergence criterion can be a predetermined number K of iterations, or a minimum value of the gradient ∇ ε k< of the indicator, or a difference considered negligible between two phase vectors φ 0 k − 1 , φ 0 k consecutive. When the convergence criterion is met, we have an estimate considered correct of a complex image of the sample in the detection plane P 0 and / or in the reference plane P ref . Step 126: Obtaining the complex reference image.

[0096] At the end of the last iteration, the process may include propagation of the complex image A 0 k resulting from the last iteration in the reference plane P ref , in order to obtain a complex reference image A ref = A ref k Alternatively, the complex reference image A ref is the complex image A 0 k resulting from the last iteration, in the detection plan P 0.

[0097] According to one embodiment, each profile is obtained by propagating an image I 0 acquired by the image sensor at different reconstruction distances z, resulting in as many reference images A ref than reconstructed distances. This yields a stack of complex reference images. However, such an implementation requires a precise estimation of the complex amplitude of the exposure light wave 14 at the different reconstruction distances z. The inventors considered it preferable to rigorously reconstruct a reference image A ref in a reference plan P ref . Starting from this complex reference image A ref , The profiles are obtained by applying a simple numerical propagation, considering different reconstruction distances, to the reference complex image, in order to obtain the so-called secondary complex images. A ref,z . Experimental trials.

[0098] THE figures 5A à 5D These represent experimental profiles obtained from leukocytes 10b and erythrocytes 10a bathed in cerebrospinal fluid. The experimental conditions are as follows: Sample 10: Cerebrospinal fluid contained in a Countess® fluidic chamber, the volume examined reaching 3 mm³, or 3 µl. Light source 11: Cree MC-E Color LED, comprising three LEDs that can be activated simultaneously or successively, each LED emitting respectively in the following spectral bands Δλ: 450 nm - 465 nm; 520 nm - 535 nm; 620 nm - 630 nm; Image sensor: 3840 x 2748 pixel monochrome CMOS sensor, each pixel measuring 1.67 µm on a side, the detection area extending over approximately 30 mm²; Distance D between the light source 11 and the sample 10: 5 cm; Distance d between the sample 10 and the image sensor 16: 2000 µm; thickness e of the fluidic chamber 15: 100 µm; diameter of the opening of the spatial filter 18: 150 µm;

[0099] Since the sensor is monochrome, images of the sample are acquired by activating one of the light-emitting diodes composing the light source 11, so as to acquire an image representative of the spectral band Δλ of the activated diode.

[0100] The process described in connection with the figure 2A , the complex reference image being obtained as described in connection with the figures 3A et 3B .

[0101] THE figures 5A et 5B represent respectively the modulus and phase profiles obtained along generators parallel to the Z-axis of propagation and passing through leukocytes (10b) or erythrocytes (10a), the sample being illuminated in the 620nm - 630nm spectral band, which corresponds to red illumination. In this example, as shown on the figure 2E the reference plan P ref corresponds to the sample plan P 10, located at a distance of 2000 µm from the detection plane P0. We note that the presence of an erythrocyte results in variations in phase and modulus on either side of the sample plane, the latter corresponding to the coordinate z = 0. For most particles, whether erythrocyte or leukocyte, the modulus takes a minimum value at a distance between 0 and 100 µm from the sample plane. P 10. However, for some leukocytes, this minimum value is reached at a distance greater than 100 µm from the plane of the sample.

[0102] THE figures 5C et 5D are similar to figures 5A et 5B The sample is illuminated in the spectral band 450 nm - 465 nm, which corresponds to blue. As previously stated, the inventors considered that illumination in a red spectral band, typically between 500 nm and 600 nm, allows for better discrimination between leukocytes and erythrocytes.

[0103] The modulus or phase profile of the exposed light wave 14, along the propagation axis Z, allows for the identification of erythrocytes 10a and leukocytes 10b, each particle exhibiting a typical profile in both phase and modulus. This property has been applied to the enumeration of leukocytes 10b in cerebrospinal fluid to aid in the diagnosis of meningitis. As indicated in relation to the prior art, it is known that a leukocyte concentration greater than 10 cells per µl should raise suspicion of bacterial meningitis.

[0104] The inventors analyzed, using the device and method described above, 215 samples of cerebrospinal fluid, each sample taken from a different person. Of these 215 samples: 15 correspond to a case of meningitis; 12 correspond to cases of cancer (glyoma, medulloblastoma or carcinoma); 25 correspond to hemorrhages; 13 correspond to an autoimmune disease; 150 are negative samples, with no particular pathology associated with them, including 17 with a particularly high concentration of erythrocytes, due to trauma occurring during lumbar puncture.

[0105] On each sample, the process described in connection with the figure 2H , the complex reference image being obtained according to the process described in connection with the figures 3A et 3B The sample was illuminated in the spectral band [450nm - 465 nm] to determine the radial position ( x i , y i ) from the center of each cell, then along the spectral band [620 nm - 630 nm] to acquire an image of the sample from which the reference complex image is obtained. From the reference complex image, modulus and phase profiles were obtained within a spatial interval of 500 µm on either side of the sample plane. Forty complex images were reconstructed at z-coordinates spaced 25 µm apart. From these images, the profile and phase values ​​were obtained at each reconstruction height and then interpolated to produce continuous profiles.

[0106] We first applied an initial threshold, so that profiles not exceeding this initial threshold are not subsequently considered. This allows us to exclude irrelevant particles, such as dust. The initial threshold value can be 0.1 when the profile is a phase profile.

[0107] In order to discriminate between leukocytes and erythrocytes, a high threshold was applied. Th h and a low threshold Th l to the profiles representing the evolution of the phase. These thresholds are represented on the figure 5B The phase profiles of each particle were determined, and those crossing the upper threshold Th h and / or the lower threshold Th l were considered representative of a leukocyte. The upper and lower threshold values ​​were +1.38 rad and -1.38 rad, respectively. Thus, an initial identification of leukocytes was based on thresholding the phase profile associated with each particle.

[0108] A second classification was based on the position of the minimum modulus value. When this value is located at a distance greater than 100 µm from the plane of the sample P10, the particle associated with the profile is considered to be a leukocyte. The 100 µm distance is indicated by a double arrow on the figure 5A This second classification was applied to profiles associated with particles not considered to be leukocytes following the first classification.

[0109] The classifications indicated above are correct, but other types of classifications, based on profiles obtained from the module or the phase, or combining the module and the phase, are conceivable.

[0110] One possible classification of profiles is based on calculating the area of ​​a surface extending between the profile and a line connecting two notable points on the profile. A first notable point P1 is, for example, the point at which the profile reaches its minimum value. A second notable point P2 is a point at which the profile takes on a predetermined value at a reconstruction distance z greater than that for which the minimum value is obtained. figure 6A represents the area between a profile and a line extending between the first notable point P1 and the second notable point P2. figure 6B shows values ​​for such an area for profiles established on erythrocytes 10a and on leukocytes 10b. To establish these areas, the following were considered: for the first notable point P1, the minimum value of each profile, for the second notable point P2, the point reaching a percentage of the minimum value of the profile, at a reconstruction distance z greater than that corresponding to the first notable point P1.

[0111] We observe that the calculated area is significantly larger for leukocytes than for erythrocytes. Such an area constitutes a reliable metric for the classification of detected particles.

[0112] Furthermore, the identification of particles of interest can be achieved by applying classical classification methods, such as Principal Component Analysis, to the profile associated with each particle. These classification methods rely on typical profiles obtained through calibrations using calibration samples containing known particles.

[0113] There figure 7 represents the experimental results obtained using a classification of profiles illustrated in relation to the figures 5A et 5B The x-axis and y-axis represent the concentrations of erythrocytes and leukocytes determined in each sample, respectively. The horizontal line L 1, equation y The value of 10 represents the threshold concentration of 10 leukocytes per µl. Confirmed cases of meningitis appear as gray discs, while other cases appear as dots (hemorrhages, cancer, trauma, autoimmune disease) or circles (negative cases). Using this identification method, 57 positive cases were identified, including the 15 cases of meningitis (true positives) symbolized by the discs, and 42 false positives. It is noted that this classification does not generate false negatives, indicating high sensitivity. The key to this figure lists the bacteria responsible for meningitis.

[0114] The number of false positives can be reduced by applying another threshold, based on a ratio between leukocyte and erythrocyte concentrations. When this ratio exceeds a certain threshold, for example 1 / 200, the sample is considered representative of a hemorrhage. This second threshold is shown on the figure 7 , by the line L 2 whose equation is y = x 200 Taking this second threshold into account eliminates 8 false positives, increasing the specificity of the process without impacting its sensitivity. The two lines L 1 and L 2 delimit a half-space, corresponding to a concentration of leukocytes and erythrocytes for which meningitis may be suspected.

[0115] During this trial, samples were analyzed using a standard microscopic method. Two positive samples are circled by a dotted line on the figure 7The results were not considered positive. These initial tests suggest that the method's sensitivity could be superior to conventional microscopic measurement.

[0116] The procedure described above allows for the simultaneous analysis, from a single image, of several microliters of cerebrospinal fluid, establishing leukocyte and even erythrocyte concentrations, and performing diagnostic analysis on samples with high sensitivity for a specific pathology, in this case, meningitis. It relies on the use of inexpensive, simple, and easy-to-implement equipment. In particular, it is automatable, making it possible to perform analyses near the point of collection using point-of-care devices. Furthermore, the procedure provides results in just a few minutes.

[0117] The procedure described above could be used for cell detection as an aid in the diagnosis of other pathologies. The bodily fluid can be urine, lymph, or blood, particularly dilute blood or synovial fluid.

Claims

1. Method for determining the quantity of leukocytes (10b) contained in a sample (10), the sample comprising particles (10a, 10b) and extending along a plane, referred to as the sample plane (P10 ), the method comprising the following steps: a) illuminating the sample using a light source (11), the light source emitting an incident light wave (12) propagating towards the sample (10) along a propagation axis (Z); b) acquisition, using an image sensor (16), of an image (I0 ) of the sample (10), formed in a detection plane (P0 ), the sample being disposed between the light source (11) and the image sensor (16), the image being representative of a light wave (14) known as the exposure wave, to which the image sensor (16) is exposed under the effect of illumination, the acquired image being a hologram formed from interference patterns between a portion of the incident light wave transmitted through the sample and a light wave resulting from the diffraction of the incident light wave by the sample; the method being characterised in that it also comprises the following steps: c) from the image acquired (I0 ) in step b), applying a propagation operator (h ) so as to calculate a complex image (Aref) , known as the reference image, representative of the sample, in a reference plane (Pref ); d) determination of radial positions (xi, yi ) of several particles in a plane parallel to the detection plane (P0 ), each radial position being associated with a particle; e) calculation of at least one characteristic quantity (M, φ ) of the exposure light wave (14), at each radial position (xi, yi ), and at a plurality of distances (z ) from the detection plane (P0 ); f) formation of a profile(M(z), φ(z)) representing a change in the characteristic quantity calculated in step e) along an axis parallel to the propagation axis (Z) and passing through each radial position (xi, yi) determined in step d), each profile being associated with a particle; g) based on each profile formed in step f), identification of the leukocytes; h) determination of a quantity of leukocytes in the sample based on a number of leukocytes identified in step g).

2. Method according to claim 1, wherein in step c), the reference plane (Pref ) is the plane of the sample (P10 ).

3. Method according to any of the preceding claims, wherein in step e), the characteristic quantity comprises the modulus (M ) or the phase (φ ) of a complex expression (A ) of the exposure light wave (14).

4. Method according to any of the preceding claims, wherein step d) comprises the following sub-steps: d1) applying a propagation operator (h ) to the reference image (Aref), so as to obtain complex images, known as secondary complex images(Aref,z), at different distances (z) from the reference plane (Pref) along the propagation axis (Z), the secondary complex images and the reference image forming a stack of complex images; d2) determining a radial position (xi, yi) of particles from the images in the stack of complex images obtained in sub-step d1).

5. Method according to claim 4, wherein the radial position (xi, yi) of particles is obtained by forming: - a first image (Mmax,φmax) , each point of which represents a maximum value, at said point and along the propagation axis (Z), of a modulus or phase of the images in the stack of complex images; - a second image (Mmin ,φmin), each point of which represents a minimum value, at said point and along the propagation axis (Z), of a modulus or phase of the images in the stack of complex images; - a differential image (ΔM , Δφ) representing a difference between the first image and the second image; the radial position (xi, yi) of each particle being obtained by applying a threshold to the differential image.

6. Method according to any of the preceding claims, wherein step e) comprises the following sub-steps: e1) applying a propagation operator (h ) to the reference image(Aref), so as to obtain complex images, known as secondary complex images(Aref,z), at different distances (z) from the reference plane (Pref ) along the propagation axis (Z), the secondary complex images and the complex reference image forming a stack of complex images; e2) determining the modulus or phase of the exposure light wave (14) to which the image sensor (16) is exposed from the images in the stack of complex images obtained in sub-step e1).

7. Method according to any of claims 1 to 5, wherein step c) comprises applying a propagation operator (h) to the image (I0) acquired in step b) at different reconstruction distances (z ) so as to obtain as many reference images as there are reconstruction distances, and wherein in step f), the profiles are obtained from said reference images.

8. Method according to any of the preceding claims, wherein step g) comprises applying a threshold to each profile (M(z), φ(z)), a profile crossing a predetermined threshold (Thh, Thl) being considered representative of a leukocyte.

9. Method according to claim 8, wherein each profile represents the evolution of the phase of the exposure light wave (14) along the axis of propagation.

10. Method according to any of the preceding claims, in which step g) comprises determining a position, along the propagation axis (Z), of a maximum or minimum value of profiles, such that a profile for which said maximum or minimum value extends beyond a threshold distance from the sample plane (P10 ) is considered representative of a leukocyte.

11. Method according to any of the preceding claims, wherein step g) comprises: - selecting, on the profile, a first notable point (P1) and a second notable point (P2); - calculating an area extending between the profile and a straight line segment joining the two notable points.

12. Method according to claim 11, in which the first notable point is the point at which the profile takes a minimum or maximum value, the second notable point being a point at which the profile takes a predetermined value.

13. Method according to any of the preceding claims, wherein: - step a) comprises illuminating the sample according to two spectral bands (Δλ1, Δλ2)) that are distinct from each other; - step b) comprises acquiring an image of the sample (I0(Δλ1), I0(Δλ2)) in each of the two spectral bands; - step d) is performed on the basis of the image acquired (I0(Δλ1)) in step b) in the first spectral band (Δλ1) ; - step e) is performed on the basis of the image acquired (I0(Δλ2)) in step b) in the second spectral band (Δλ2) .

14. Method according to claim 13, wherein the first spectral band(Δλ1) is between 400 nm and 550 nm, and wherein the second spectral band(Δλ2) is between 500 nm and 600 nm.

15. Method according to any of the preceding claims, comprising the steps: gbis) classifying each particle, whose radial position was identified in step d), as being, or not being, an erythrocyte (10a) based on each profile formed in step f); hbis) determining a quantity of erythrocytes (10a) in the sample based on the classifications made in step gbis); i) determining a ratio between the quantity of leukocytes (10b) determined in step h) and the quantity of erythrocytes (10a) determined in step hbis).

16. Method according to any of the preceding claims, in which there is no magnifying optics between the sample (10) and the image sensor (16).

17. A Device for counting leukocytes (10b) contained in a sample (10), the device comprising: s - a light source (11) capable of emitting an incident light wave (12) propagating towards the sample (10) along a propagation axis (Z); - an image sensor (16) extending along a detection plane (P0 ); - a support (10s) configured to hold the sample (10) between said light source (11) and the image sensor (16); a processor (20) configured to receive an image of the sample acquired by the image sensor (16) and to implement steps c) to h) of the method according to any of claims 1 to 16.

18. Device according to claim 17, wherein no magnifying optics are arranged between the sample (10) and the image sensor (16) when the sample (10) is held on the holder (10s).