Method for characterizing the trajectory of a moving particle in a sample

Abstract

Method for characterizing at least one mobile particle (10i) in a sample (10), the method comprising: acquiring at least one image (, ) of the sample during an acquisition period, using an image sensor (20), defining an observation field, the acquisition period comprising different acquisition times (); using the image or each image resulting from a), forming a trajectory image (I) representing the particles of the sample, in the observation field, at the different acquisition times; using the trajectory image resulting from a) as an input image for a detection algorithm, programmed to detect particles, and for a supervised learning artificial intelligence algorithm, programmed to calculate at least one average displacement parameter for different detected particles.

Description

Title of the invention: Method for characterizing the trajectory of a moving particle in a sample technical field

[0001] The technical field of the invention is the observation of mobile or motile microscopic particles in a sample, with a view to their characterization. One intended application is the characterization of spermatozoa. EARLIER ART

[0002] The observation of motile cellular particles, such as spermatozoa, in a sample is usually carried out using a microscope. The microscope comprises an objective lens defining an object plane, extending into the sample, and an image plane, coinciding with a detection plane of an image sensor. The microscope acquires images of the spermatozoa in a focused configuration. The choice of such a configuration involves a compromise between spatial resolution, the field of view, and depth of field.

[0003] US patent application 20240044771 describes a method for characterizing spermatozoa using an image acquisition device that trains input data into a neural network. The acquisition device may be of the defocused or lensless imaging type.

[0004] The publication Ershov D “TrackMate 7: integrating State of art segmentation algorithms into tracking pipelines”. Nat Methods 19, 829-832 (2002) describes an application for tracking moving particles, for example cells, and determining their morphological characteristics. Since the particles are moving, the practice involves acquiring a large number of images. In each image, the particles must be detected and located in order to follow their trajectory and determine their movements.

[0005] However, current characterization methods are computationally expensive, especially when the number of particles is high. It is estimated that the time to track each particle varies, depending on the number N of particles, in ;V3]ogGV)- And this applies to each image. The computation time can exceed several tens of seconds when N is several thousand. Added to this are the processing steps required to characterize the particles.

[0006] The inventor proposes a more computationally efficient method, thus reducing the analysis time. The method is particularly well-suited to the characterization of samples containing a large number of cells. Description of the invention

[0007] A first object of the invention is a method for characterizing at least one mobile particle in a sample, the method comprising: a. acquisition of at least one image of the sample during an acquisition period, using an image sensor, defining an observation field, the acquisition period comprising different acquisition times; b. using the image or each image resulting from a), formation of an image of trajectories representing the particles of the sample, in the field of observation, at the different times of acquisition; c. use of the image of trajectories resulting from a) as an input image of a detection algorithm, programmed to detect particles, and of a supervised learning artificial intelligence algorithm, programmed to calculate at least one average displacement parameter for different detected particles.

[0008] The supervised learning artificial intelligence algorithm can be a convolutional neural network.

[0009] According to one possibility, each image of the sample is acquired according to a defocused imaging modality or according to a lensless imaging modality, so that each particle forms, on each image, a diffraction pattern.

[0010] According to one possibility: - the sample extends according to a sample plan; - the image sensor extends along a detection plane; - an optical system extends between the sample and the image sensor, the optical system defining an object plane and an image plane; - the object plane is offset from the sample plane by an object defocus distance and / or the image plane is offset from the sample plane by an image defocus distance, so that during step a), each image of the sample is acquired according to a defocused imaging modality.

[0011] According to one possibility, no image-forming optics extend between the sample and the image sensor, so that in step a), each image of the sample is acquired according to a lensless imaging modality.

[0012] According to one possibility, each image of the sample is acquired according to an interference imaging modality.

[0013] According to one possibility: - step (a) involves the acquisition of several images; - in step (b), the image of trajectories is obtained by a combination of images acquired during step (a).

[0014] The combination may be or include a sum.

[0015] According to one possibility: - each acquired image and the image of trajectories being defined according to pixels; - the value of the trajectory image for a pixel, is the maximum value, at said pixel, of the set of acquired images.

[0016] According to one possibility: - each acquired image and the image of trajectories being defined according to pixels; - the value of the trajectory image for a pixel, is the maximum value, at said pixel, of the set of acquired images.

[0017] According to one possibility: - step (a) involves the acquisition of several images; - a holographic reconstruction algorithm is applied to each acquired image, so as to form, from each acquired image, a reconstructed image, - In step (b), the image of the trajectories is obtained by a combination of the reconstructed images. The combination can be, or include, or be a sum.

[0018] According to one possibility: - during step (a), the image is acquired while the sample is subjected to several successive illuminations, each illumination being carried out at an acquisition time; - the image of trajectories corresponds to the image acquired during step (a).

[0019] According to one possibility, step c) involves, starting from the image of trajectories: • a determination of at least one average characteristic of the particle trajectories during the acquisition period; • and / or a calculation of an average particle speed from the trajectory.

[0020] According to one possibility, the particles are spermatozoa.

[0021] A second object of the invention is a device for observing a sample, the sample comprising mobile particles, the device comprising: - a light source, configured to illuminate the sample; - an image sensor, configured to form an image of the sample; - a holding structure, configured to hold the sample between the light source and the image sensor; - a processing unit, connected to the image sensor, and configured to implement steps b) and c) of a process according to the first object of the invention from at least one image acquired by the image sensor.

[0022] The invention will be better understood upon reading the description of the exemplary embodiments presented later in this description, in connection with the figures listed below. FIGURES

[0023] Fig. 1 represents a first embodiment of a device enabling implementation of the invention.

[0024] Fig. 2 represents a second embodiment of a device enabling implementation of the invention.

[0025] Fig. 3 schematically represents the trajectory of a spermatozoon.

[0026] Fig. 4 shows the main steps of a process for characterizing mobile particles in the sample.

[0027] Figure [5A] shows different images of a sample, acquired at different times

[0028] Figure [5B] shows a trajectory image obtained from images acquired at different times.

[0029] Figure 6 shows a comparison of measurements obtained by implementing the invention with real-world data (ground truth). Figure 6 is divided into 10 elementary figures, 6A to 6J, each addressing a measured characteristic. Figure 6A shows the number of spermatozoa counted in the field of observation. Figures 6B to 6F concern sperm trajectory characteristics. Figures 6G to 6J concern the proportions of spermatozoa classified by type in the different samples analyzed.

[0030] Fig. 7 is divided into 10 elementary figures 7A to 7J, each elementary figure being a Bland-Altman representation of the data respectively represented in figures 6A to 6J.

[0031] Figures 8A and 8B show images acquired for different concentrations of spermatozoa.

[0032] Figure 9 is divided into seven elementary figures 9A to 9G. Figures 9A, 9B, and 9C show images acquired under different conditions. Figures 9D, 9E, and 9F show integrated images obtained respectively from images shown in Figures 9A, 9B, and 9C. Figure 9G shows a trajectory image obtained by combining the maxima of elementary images. PRESENTATION OF SPECIFIC IMPLEMENTATION METHODS

[0033] Figure 1 shows a first embodiment of a device 1 enabling implementation of the invention. According to this first embodiment, The device allows the observation of a sample 10 interposed between a light source 11 and an image sensor 20. The light source 11 is configured to emit an incident light wave 12 propagating to the sample parallel to a propagation axis Z.

[0034] The device includes a sample holder 10s configured to receive the sample 10, such that the sample is held on the holder 10s. The sample thus held extends along a plane, called the sample plane P10. The sample plane corresponds, for example, to a mean plane around which the sample 10 extends. The sample holder can be a glass slide, for example, 20 µm thick. The thickness can be between 10 µm and 1 mm, and preferably between 10 µm and 500 µm or between 10 µm and 100 µm.

[0035] The sample includes, in particular, a liquid medium 10m in which mobile and possibly motile particles 10 are immersed. The medium 10m may be a biological fluid or a buffer fluid. It may, for example, include a bodily fluid, in its pure or diluted state. By bodily fluid, we mean a fluid generated by a living body. This may include, but is not limited to, blood, urine, cerebrospinal fluid, semen, and lymph.

[0036] The sample 10 is preferably contained in a fluidic chamber 10c. The fluidic chamber is, for example, a fluidic chamber with a thickness between 20 pm and 100 pm. The thickness of the fluidic chamber, and therefore of the sample 10, along the propagation axis Z, typically varies between 10 pm and 200 pm, and is preferably between 20 pm and 50 pm.

[0037] One of the objectives of the invention is the characterization of moving particles in the sample. In the described embodiment, the mobile particles are spermatozoa. In this case, the sample contains sperm, possibly diluted. In this case, the fluidic chamber 10c can be a counting chamber dedicated to the analysis of cell motility or concentration. For example, it could be a counting chamber marketed by Leja, with a thickness between 20 µm and 100 µm.

[0038] According to other applications, the sample comprises mobile particles, for example microorganisms, for example microalgae or plankton, or cells, for example cells undergoing sedimentation.

[0039] 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, for example 5 cm. Advantageously, the light source 11, as seen by the sample, is considered to be a point source. This means that its diameter (or diagonal) is preferably less than one tenth, better yet one hundredth of the distance between the sample and the light source.

[0040] The light source 11 is, for example, a light-emitting diode. In the example described, the light is emitted at a wavelength of 450 nm. It is preferably associated with a diaphragm 14, or spatial filter. The diaphragm opening is typically between 5 pm and 1 mm, preferably between 50 pm and 1 mm. In this example, the diaphragm has a diameter of 400 pm. According to another configuration, the diaphragm can be replaced by an optical fiber, one end of which is placed facing the light source and the other end of which is placed opposite the sample 10. The device can also include a diffuser 13, arranged between the light source 13 and the diaphragm 14. The use of a diffuser / diaphragm assembly is, for example, described in US10418399.

[0041] The image sensor 20 is configured to form an image of the sample according to a detection plane P20. In the example shown, the image sensor 20 comprises a pixel array, of the CCD or CMOS type. The detection plane P20 preferably extends perpendicularly to the propagation axis Z. Preferably, the image sensor has a large sensitive area, typically greater than 10 mm². In this example, the image sensor is an IDS-UL3160CP-M-GL sensor with 4.8 x 4.8 pm² pixels, the sensitive area being 9.2 mm x 5.76 mm, or 53 mm².

[0042] In the example shown in [Fig. 1], the image sensor 20 is optically coupled to the sample 10 by an optical system 15. In the example shown, the optical system comprises a lens 15i and a tube lens 152. The latter is intended to project a formed image onto the sensitive surface of the image sensor 20 (surface area of ​​53 mm²). The image acquisition rate is, for example, 60 frames per second, with an exposure time of 2 ms per frame.

[0043] In this example: - the 15i lens is a Motic CCIS EF-N Plan Achromat lOx lens, with a numerical aperture of 0.25; - Lens 152 is a Thorlabs LBF254-075-A lens - focal length 75 mm.

[0044] Such an assembly provides a field of view of 3 mm², with a spatial resolution of 1 pm.

[0045] The optical system 15 defines an object plane Po and an image plane P20. In the embodiment shown in [Fig. 1], the image sensor 20 is configured to acquire an image in a defocused configuration. The image plane P20 coincides with the detection plane P20, while the object plane Po is offset by an object focusing distance θ of between 10 pm and 500 pm, relative to the sample. The focusing distance is preferably between 50 pm and 200 pm, for example, 100 pm. The object plane Po extends outside the sample. Alternatively, the object plane extends within the sample, while the image plane is offset from the detection plane by an image defocus distance. The image focusing distance is preferably between 50 pm and 200 pm, for example, 100 pm. Alternatively, the object plane Po and the image plane P are both offset from the sample plane and the detection plane, respectively. Regardless of the configuration chosen, the defocus distance is preferably greater than 10 pm and less than 1 mm, or even 500 pm, and preferably between 50 pm and 150 pm. The observation of a cell sample in a defocused configuration was described in US patent 10545329.

[0046] One advantage of defocused imaging is that it allows observation of translucent or transparent particles, with satisfactory contrast.

[0047] According to one possibility, each acquired image can be subjected to a digital reconstruction algorithm to improve spatial resolution. It is known that the use of digital reconstruction algorithms makes it possible to obtain sharp images of particles. Such algorithms are described, for example, in US10564602, US20190101484, or US20200124586. In this type of algorithm, from a hologram acquired in a detection plane, an image of the sample is reconstructed in a reconstruction plane distant from the detection plane. It is common for the reconstruction plane to extend through the sample. However, this type of algorithm can require a relatively long computation time. The invention has proven effective in the case of sperm characterization from images acquired by the image sensor (holograms), without implementing the reconstruction algorithm.

[0048] It is known that lensless imaging, coupled with holographic reconstruction algorithms, allows the observation of transparent or translucent cells while maintaining a large field of view and a great depth of field. For example, US patents 9588037 and 8842901 describe the use of lensless imaging for observing spermatozoa. US patents 10481076 and 10379027 also describe the use of lensless imaging, coupled with reconstruction algorithms, for characterizing cells.

[0049] Figure 2 represents a second embodiment of device 1' suitable for implementing the invention. Device 1' comprises a light source 11, a diffuser 13, a diaphragm 14, an image sensor 20, a holding structure 17, and a processing unit 30, as described in connection with the first embodiment. The holding structure 17 is configured to define a fixed distance between the sample and the image sensor. According to this embodiment, the device does not include an image-forming lens between the image sensor 20 and the sample 10. The image sensor 20 is preferably positioned close to the sample, with the distance between the image sensor 20 and the sample 10 typically being between 100 µm and 3 mm. According to this embodiment, the image sensor acquires images using a lensless imaging modality. The sample is preferably contained in a fluidic chamber 10c, for example, a "Leja" chamber as described in connection with the first embodiment. The advantage of such an embodiment is that it does not require precise positioning of an optical system 15 relative to the sample 10, and it provides a large field of view. The disadvantage is the acquisition of lower-quality images, although these remain usable.

[0050] Other diffraction or interferometric imaging systems can be implemented. These may, for example, be phase contrast imaging systems.

[0051] The imaging methods described above are suitable for transparent or translucent particles. When the particles are sufficiently opaque, a conventional imager focused on the sample can be used.

[0052] Figure 3 illustrates a trajectory of a spermatozoon, between an initial instant and a final instant h. Each point illustrates a position of a spermatozoon, at an instant f between and tf.

[0053] From the trajectory, it is possible to characterize the movement of a spermatozoon by determining various characteristics, for example: - a velocity of the linear trajectory VSL, usually referred to as "velocity straightline path", which corresponds to the velocity calculated on the basis of a straight-line distance between the first and last points of the trajectory, corresponding respectively to the initial and final moments of the acquisition; - a velocity of the curvilinear trajectory VCL, usually referred to as "velocity curvilinear path": this is a velocity established by summing the distances traveled between two successive instants, and multiplying by the acquisition frequency; - a velocity of the average trajectory VAP, usually referred to as "velocity average path": this is a velocity established after smoothing the trajectory of a particle: the distance traveled along the smoothed trajectory (or average trajectory) is divided by the time between the initial and final instants. In [Fig.3], the average trajectory has been represented by dashed lines; - Based on the calculated velocities, it is possible to define indicators that characterize sperm motility, which are familiar to those skilled in the art. These include, for example, indicators such as: - STR straightness indicator, obtained by a ratio between VSL and VAP, usually referred to as "straightness". This indicator is closer to 1 the more the sperm moves in a straight line; - linearity indicator LIN, obtained by a ratio between VSL and VCL, usually referred to as "linearity". This indicator is also closer to 1 the more the sperm moves in a straight line.

[0054] Quantifying the velocities or parameters listed above allows spermatozoa to be categorized according to their motility. For example, a spermatozoon is considered to be: - motile if the length of the trajectory is greater than a first threshold, for example 10 pixels, and its displacement along the average trajectory (VCLx At, At being the acquisition period) is greater than a predefined length, corresponding for example to the length of a sperm head; - progressive if the length of the trajectory is greater than the first threshold, and if the straightness STR and the speed of the average trajectory VAP are respectively greater than two threshold values ​​STRth and VAPth i; - slow if the length of the trajectory is greater than the first threshold and if the speed of the linear trajectory VSL and the speed of the average trajectory VAP are respectively less than two threshold values ​​VSLth 2 and VAPth 2; - static if the length of the trajectory is greater than the first threshold and if the speed of the linear trajectory VSL and the speed of the average trajectory VAP are respectively less than two threshold values ​​VSLth 3 and VAPth 3. - uncategorized if the trajectory length is less than the first threshold.

[0055] Figure 4 schematically illustrates the main steps of a method for processing several images acquired by an image sensor according to the defocused imaging modality. The method is described in relation to the observation of spermatozoa, it being understood that it can be applied to the observation of other types of motile particles.

[0056] Step 100: Acquisition of a series of In images.

[0057] During this step, images of the sample are acquired at different acquisition times, with each time corresponding to an acquired image. The acquisition times extend between an initial time and a final time. A stack of acquired images is obtained, as shown in [Fig. 5A].

[0058] Step 110: Holographic reconstruction (optional step)

[0059] During this step, a holographic reconstruction algorithm is applied to each acquired image In, so as to form an image Ir,n of the sample at each acquisition time. This step is not essential.

[0060] Step 1 2 0: Formation of a trajectory image

[0061] During this step, a trajectory image I is formed. The term trajectory image refers to the fact that the image allows the respective trajectories of several particles in the sample to be estimated. Given the particle density, the trajectory image shows the successive positions of several dozen, or even several hundred or thousand, particles between the initial and final times. The trajectory image can be formed by a combination of the images acquired In during step 100 or the reconstructed images Ir,n during step 110.

[0062] Each acquired image In is defined according to pixels r. r designates a coordinate in the detection plane defined by the image sensor.

[0063] One way of obtaining the image of trajectories I is to take, in each pixel (xy), an extremal value of the set of images acquired In or reconstructed Jr,n.

[0064] To establish the image of trajectories, one can take into account the maximum value of each acquired image (or of each reconstructed image).

[0065] The value of each pixel in the trajectory image is then

[0066] / (xy) - maxZjx, y) (1) •

[0067] Obtaining a trajectory image is an important aspect of the invention. The trajectory image shows the position of the particles in the field of view of the image sensor at different acquisition times. Thus, the trajectory image allows visualization of the trajectory of the sample particles in the field of view at different acquisition times. Figure 5B shows an example of a trajectory image obtained by combining 30 acquired images.

[0068] Unlike the prior art, multiple images of the same particle are not formed at different acquisition times. The invention differs from the prior art in that it combines, on a single image, the different images of the sample at different acquisition times.

[0069] Step 1 3 0: Detection and characterization of the movement of each particle.

[0070] During this step, the trajectory image is used as input data for a detection algorithm, configured to detect particles and optionally count them. This may be a particle tracking algorithm, such as Trackmate 7, as described in the prior art.

[0071] The image of the trajectories is also used as input data for a characterization algorithm configured to calculate metrics enabling to determine average characteristics of the movements of the detected particles. These include, for example, average velocity or motility characteristics as previously described.

[0072] The characterization algorithm may be a supervised artificial intelligence algorithm, for example a convolutional neural network (CNN). It may, for example, be the neural network described in Tan M. et al., "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks," available at arxiv.org / pdf / 1905.11946.11, which is an algorithm configured to perform a specific task in imaging.

[0073] The algorithm was trained to determine the average characteristics VSL, VCL, VAP, STR, and LIN in the sample. The algorithm was then tested on different samples. A particular feature of the test was that the sperm concentrations in the different samples varied.

[0074] Figures 6A to 6F show, respectively for different characteristics, the actual characteristics (ground truth – x-axis) as a function of the estimates implemented by steps 100 to 140 described previously (y-axis). Figure 6A shows the number of spermatozoa in the field of view for different samples (y-axis: estimate – x-axis: ground truth). It can be observed that the number of spermatozoa observed in a single image varies from a few hundred to approximately 6,000. It is estimated that an image showing 3,000 spermatozoa corresponds to a sperm concentration of 60 million per mL. In Figure 6A, each point corresponds to a sample.

[0075] The features shown in Figures 6B to 6F are respectively the mean values ​​of sperm trajectory characteristics in the different samples. These are the mean values ​​of VAP, VCL, VSL, STR, and LIN for the sperm in the different samples. Each point represents a sample. The shade of gray of each point depends on the number of sperm in the field of view for the sample to which the analyzed sperm belongs. A linear regression model, represented by a solid line, was established for each feature. A linear correlation coefficient S and a linear coefficient of determination R² were also calculated for each feature.

[0076] On each of the figures 6B to 6F, the x-axis corresponds to the ground truth, and the y-axis corresponds to an estimate, for each sample, obtained by implementing the invention.

[0077] From the determined characteristics, a percentage of static (Figure 6G), motile (Figure 6H), progressive (Figure 61), and spermatozoa was determined. slow (figure 6J), and this in each sample. Figures 6G to 6J show the estimated percentages for each category of spermatozoa (ordinate axis) as well as the "ground truth" percentages obtained by implementing a reference method, with implementation of a tracking algorithm, and estimation of all the motility characteristics for the different spermatozoa detected.

[0078] Table 1 represents, for each characteristic studied, the linear correlation coefficient S and a linear coefficient of determination R2. parameter figure S R2 number 6A 0.93 0.84 VAP 6B 1.02 0.9 VCL 6C 1.03 0.9 VSL 6D 1.03 0.96 STR 6E 1.02 0.93 LIN 6F 1.00 0.90 % static 6G 0.93 0.91 % motile 6H 1.03 0.95 % progressive 6J 0.95 0.94

[0079] [Table 1]

[0080] Table 1 shows that the linearity and correlation coefficients are close to 1, which confirms the relevance of the estimation of each characteristic by the algorithm based on a convolutional neural network.

[0081] Figures 7A to 7J are Bland-Altman plots corresponding respectively to the features discussed in connection with Figures 6A to 6J. On each of these plots: - the y-axis shows a difference between each estimated value and each "ground truth" value; - the x-axis shows an average between each estimated value and each "ground truth" value.

[0082] Observation of figures 7A to 7J allows us to conclude that there is no systematic error.

[0083] Figures 8A and 8B show examples of image portions of samples containing 3000 and 6000 spermatozoa, respectively, corresponding to spermatozoa concentrations of 60 M / ml and 120 M / ml, respectively. It is estimated that the process described above is feasible in the case of spermatozoa, up to a concentration of 80 M / ml. Above this concentration, the number of spermatozoa in the image sensor's field of view is too high. The maximum concentration depends on the instrumentation used and the type of particles. The maximum concentration can be determined beforehand, based on simulations or experimental tests. Variants

[0084] According to one possibility, the image of trajectories is formed not from the maxima of each acquired image In, according to (1), but by a sum of each acquired (or reconstructed) image In.

[0085] Thus, / - V / (2)

[0086] Figure 9A shows an example of a reconstructed image. Figure 9D shows a sum of 30 images.

[0087] In the case where the trajectory image is formed by image summation, it is preferable that each summed image be thresholded (see Figure 9B) or have a LUT applied (for example, a gamma LUT: see Figure 9C). This allows for the obtaining of more usable integrated images: see Figure 9E (summation of thresholded images) or Figure 9F (summation of images corrected with a gamma LUT). The integrated images shown in Figures 9E and 9F are more usable.

[0088] Thus, when forming a trajectory image as described in (2), it is preferable that each summed image has undergone prior processing. The trajectory image is then comparable to an image obtained according to (1). Figure 9G shows an image obtained according to (1), from images such as the one shown in Figure 9A. The processing can be performed at the camera level or by software.

[0089] According to another embodiment, the image of trajectories can be an image acquired with an exposure time corresponding to the acquisition period. The sample is then illuminated stroboscopically, by successive light pulses, each light pulse corresponding to an acquisition instant.

[0090] The invention enables the characterization of mobile particles in a sample by implementing a rapid processing method. This essentially involves counting and / or determining characteristics related to the trajectory of the particles in the sample. In the case of spermatozoa, certain trajectories are specific to a particular morphological feature. It is therefore possible to obtain information about sperm morphology by characterizing their trajectories. Thus, the algorithm can indirectly provide morphological information through trajectory analysis.

Claims

Demands

1. A method for characterizing at least one mobile particle (10i) in a sample (10), the method comprising: a. acquiring at least one image (I, Itl) of the sample during an acquisition period, using an image sensor (20), defining an observation field, the acquisition period comprising different acquisition times (¾ b. using the image or each image resulting from a), forming a trajectory image (I) representing the particles of the sample, in the observation field, at the different acquisition times; c. using the trajectory image resulting from a) as the input image of a detection algorithm, programmed to detect the particles, and of a supervised learning artificial intelligence algorithm, programmed to calculate at least one average displacement parameter for different detected particles.

2. A method according to claim 1, wherein the supervised learning artificial intelligence algorithm is a convolutional neural network.

3. A method according to any one of the preceding claims, wherein each image of the sample is acquired according to a defocused imaging modality or according to a lensless imaging modality, so that each particle forms, on each image, a diffraction pattern.

4. A method according to claim 3, wherein: - the sample extends along a sample plane (Pio); - the image sensor extends along a detection plane (P20); - an optical system (15) extends between the sample and the image sensor, the optical system defining an object plane (Po) and an image plane (P10); - the object plane is offset from the sample plane by an object defocus distance and / or the image plane is offset from the sample plane by an image defocus distance, such that during step a), each image of the sample is acquired according to a defocused imaging modality.

5. A method according to claim 3, wherein no image-forming optics extend between the sample and the image sensor, so that in step a), each image of the sample is acquired in a lensless imaging modality.

6. A method according to claim 1, wherein each image of the sample is acquired according to an interference imaging modality.

7. A method according to any one of the preceding claims, wherein: - step (a) involves the acquisition of several images; - in step (b), the image of trajectories is obtained by a combination of the images acquired during step (a).

8. Method according to claim 7, wherein the combination is a sum.

9. A method according to claim 7, wherein - each acquired image and the trajectory image being defined according to pixels; - the value of the trajectory image for a pixel, is the maximum value, said pixel, of the set of acquired images.

10. A method according to any one of the preceding claims, wherein - each acquired image and the trajectory image being defined according to pixels; - the value of the trajectory image for a pixel, is the maximum value, said pixel, of the set of acquired images.

11. A method according to any one of the preceding claims, wherein: - step (a) comprises acquiring several images; - a holographic reconstruction algorithm is applied to each acquired image, so as to form, from each acquired image, a reconstructed image, - in step (b), the image of trajectories is obtained by a combination of the reconstructed images.

12. A method according to any one of the preceding claims, wherein - during step (a), the image is acquired while the sample is subjected to several successive illuminations, each illumination being carried out at an acquisition time; - the image of trajectories corresponds to the image acquired during step (a).

13. A method according to any one of the preceding claims, wherein - the particles are spermatozoa; - step c) comprises, from the image of trajectories: • a determination of at least one average characteristic of the trajectories of the spermatozoa during the acquisition period; • and / or a calculation of an average velocity of the spermatozoa from the trajectory.

14. Device (1, 1') for observing a sample, the sample comprising mobile particles, the device comprising: - a light source (11), configured to illuminate the sample; - an image sensor (20), configured to form an image of the sample; - a holding structure (16, 17), configured to hold the sample between the light source and the image sensor; - a processing unit (30), connected to the image sensor, and configured to carry out steps b) and c) of a method according to any one of the preceding claims from at least one image acquired by the image sensor.

Images