METHOD FOR CHARACTERIZING THE EPIDERMIS AND DERMIS FROM THREE-DIMENSIONAL IN VIVO MULTIPHOTON IMAGES OF THE SKIN.

Abstract

The present invention relates to a method for observing a biological tissue, comprising the steps of: a) acquiring by multiphoton excitation at least one three-dimensional image for each of the 2PEF and SHG signals emitted by said tissue, and b) determining by image processing from these three-dimensional images a segmentation of said tissue into at least two regions.

Description

This segmentation is not trivial, because three-dimensional acquisitions are, by their nature, complex; not only are they large, typically comprising 511 x 511 x 70 voxels, but they can also be numerous, a study possibly requiring several hundred acquisitions. It is particularly advantageous to be able, thanks to the method according to the invention, to obtain varied measurements from a single multiphoton acquisition. This method is particularly interesting because it allows the estimation of scalar or vector parameters originating from the analysis of three-dimensional data and which summarize the 3D information contained in the images. These quantitative parameters make it possible to characterize a biological tissue, in particular the skin, and its different constituents in terms of morphology, quantity and organization. The method according to the invention can be implemented in vivo, but also on ex vivo and in vitro samples. This method can be used for the in vivo evaluation of the effectiveness of certain active ingredients in tissues, for example for anti-aging purposes or for action on pigmentation. The acquired images may correspond to 2PEF, SHG and / or other signals observed by multiphoton microscopy. The method can generate combined images, for example 2PEF / SHG. In step a), we can obtain a single image linked to the 2PEF signal and a single image linked to the SHG signal. Preferably, several successive images of the 2PEF and SHG signals are acquired. The successive three-dimensional images can be obtained in step a) by integrating the fluorescence signal and the second harmonic signal at regular intervals. The integration of the signals can in particular be carried out over a duration of between 1 and 3 ns which depends on the repetition rate of the laser, for example equal to 1; 1.5; 2; 2.5 or 3 ns. The acquisition of images with integration can be carried out by a method 10 such as that described in application FR 11 60172 for the 2PEF signal. For each 2PEF and SHG signal, a limited number of images can be acquired, for example between three and five, which offers a good compromise between acquisition time and precision. Four images can thus be obtained for each signal with integration over a regular interval of 2 ns. The simultaneous acquisition of the 2PEF 15 and SHG time signals is carried out with an acquisition time per pixel of approximately 0.28 ps in order to obtain a sufficient signal-to-noise ratio. The temporal acquisition of several 3D images is particularly advantageous for the calculation of very specific measurements, for example aiming to quantify the presence of melanin from the 2PEF signal. 20 The invention improves the speed of the acquisition method and the method of image processing. Thanks to the invention, 3D characterization of the tissue is possible. The biological tissue may contain melanin and / or elastin. The tissue may contain collagen fibers. The tissue may be made of human keratin materials. "Keratin materials" means hair, eyelashes, eyebrows, skin, nails, mucous membranes, scalp, among others. The biological tissue according to the invention, in particular keratin materials, may be natural or artificial, it may be human skin, reconstructed or artificial skin. i Segmentation and regions of interest 30 Segmentation can initially partition the observed volume into at least least two areas, especially three, corresponding to the main regions of interest. In the case of the skin, this can be the exterior of the skin, the epidermis and the dermis. Segmentation can define other regions of interest such as the boundaries between these areas, or different layers within these areas, parallel to these boundaries, of thickness chosen by the user. The different regions of interest can be chosen to create a 5 quantization parameter. Indeed, all the desired measurements can be calculated within a particular region of interest, that is, by only taking into account the information contained in this region of interest. A segmentation algorithm is used to obtain the main regions of interest by taking into account two 3D grayscale images corresponding to the two 2PEF 10 and SHG channels and using all the three-dimensional information. For each voxel, the grayscale indicates the number of detected photons. The segmentation can result from four steps: (i) optional pre-filtering, taking into account the different sampling steps of the 3D images, which can be achieved using a two-dimensional Gaussian filter; (ii) definition of the markers of the main regions from the intensity of the signals, their specificity and their position; (iii) calculation of a low-level segmentation using the watershed method and construction of the corresponding adjacency graph; and (iv) final segmentation using graph cut methods, from the markers and the adjacency graph. Segmentation can involve calculating a marker for each region, then propagating the markers to the entire observed volume. Markers can be calculated from thresholds of the initial signals. Once the markers are calculated, we proceed to their propagation using the same propagation method in the different regions structured in layers. To propagate the markers, a watershed algorithm can be applied. The so-called "watershed" method is fast, but is not entirely satisfactory in regions of the image where information is lacking, particularly near the edges. Alternatively, or in combination, one can use other known graph-based methods, which do not have this drawback, but are expensive in terms of computation time and memory. The preferred propagation method combines the advantages of both approaches. It is based in part on the work described by Li et al. in the article “Lazy snapping,” in SIGGRAPH 2004 Proceedings, 2004, vol. 23, pp. 303-308 and on that of Stawiaski and Decencière, presentations in “Region merging via graph cuts,” Image Analysis and Stereology, vol. 27, no. 1, pp. 39—46, 2008. The morphological gradient of the supremum of the 2PEF and SHG signal images is calculated, and then a watershed algorithm with 2D connectivity is applied, 5 to obtain a so-called low-level segmentation of the image. The choice of 2D connectivity for this step is motivated by the high anisotropy of the sampling steps described previously. A graph is associated with low-level segmentation. A node in the graph corresponds to each region. Two nodes corresponding to two neighboring regions (considering 6-connectivity in 3D) are connected by an edge. Two additional nodes are added, corresponding to the two markers. They are connected to the nodes whose associated regions touch one or the other marker. Then, a value or weight is assigned to each edge of the graph. This is a well-known essential step in the design of graph-cutting methods. 15 Edges connecting marker nodes to other nodes are assigned an infinite value. Consider an edge e connecting two neighboring nodes, corresponding to two regions A and B. Let N(A,B) be the set of pairs of voxels (p,q) such that p belongs to A, and q belongs to B, let I(p) and I(q) be the values ​​associated with these voxels, the weight w(e) associated with this edge is then: œ (e) (s(p,q) / (l+Li(I(p), I(q))) for (p, q) € N(A, B) where s(p,q) is the surface separating voxels p and q, and L](I(p), I(q)) is the distance associated with the norm Zj (for example, in the case where we work with two images corresponding to the AF and SHG signals, this distance is given by the sum, on the one hand, of the absolute difference between the values ​​of the pixels in the AF channel and, on the other hand, the absolute difference between the values ​​of the pixels in the SHG channel). The resulting graph is cut into two parts, each containing one of the nodes associated with the markers, using a flow maximization algorithm. for a segmentation comprising several levels of separation, this determination of valued edges can be carried out several times, at each level of separation of two zones, however the calculation of the low-level segmentation is carried out only once. The method may further comprise the step of: c) carry out a measurement in at least one of the regions resulting from the segmentation. The measure can be defined using a generic nomenclature of quantization parameters. Quantification measures or parameters The invention makes it possible to calculate scalar or vector quantization parameters originating from the analysis of three-dimensional data and capable of synthesizing 3D information. Each quantification parameter as defined in the context of the invention corresponds to a measurement itself. Many measurements are available. Some are classic (average gray levels, average thickness, etc.) or are already known to those skilled in the art, such as SAAID. The invention also makes it possible to make accessible measurements which have never been described, in particular, new measurements linked to the organization and orientation of collagen fiber or elastic fiber networks. The measurements quantifiable using a method according to the invention can be grouped into different families: • Classic geometric measurements, such as average thickness or volume of regions of interest; • Signal intensity measurements, such as the average in a region of interest; • Geodesic measurements which make it possible to understand the shape and size of structures such as fiber networks, based on the concept of geodesic diameter; • Orientation measurements which determine the preferred direction of a fiber network in relation to a given reference frame (reference linked to the image or to the dermo-epidermal junction); • Granulometries, giving a distribution of the sizes of the structures concerned; • Mixed measures capable of jointly taking into account 2PEF and SHG signals (SAAID, nesting, normalized nesting, for example); • Density measurements, such as cell density in the epidermis or fiber density in the dermis; • Distribution measures that vectorially describe a distribution as a function of depth (e.g., the distribution of melanin as a function of the normalized distance to the basal layer). Nomenclature The method may be based on a generic nomenclature that allows a large number of quantification parameters to be defined, and facilitates the software implementation of the method according to the invention. The expression "quantification parameter" links the measurement 5 itself to the image(s) to be considered, such as for example those acquired in step a) or derived images, and to the region of interest on which the measurement relates. The nomenclature provides great flexibility in the definition of the quantification parameters. The nomenclature can allow the introduction of pre-treatment methods, such as the fiber enhancement filtering method described below. The nomenclature described below constitutes an example of nomenclature which can be used by a method according to the invention. In a preferred variant, the nomenclature makes it possible to define, in a conventional manner, a quantification parameter, from four descriptors: 15 • the modality of the data taken into account, i.e. the category(ies) of signals considered, • the presence or absence of pretreatment and type of pretreatment used; • the region of interest on which the measurement is carried out (examples: epidermis, dermis, sub-parts of the dermis, dermo-epidermal junction); and 20 • the measurement itself (for example: average of the signal intensity, proportion of pixels whose value exceeds a given level, tortuosity of structures, volume; representation of the distribution of a constituent of the skin, etc.). This analytical approach to defining quantization parameters provides great flexibility. 25 For the skin, the number of possible combinations which are a priori of interest is now in the order of several hundred. These combinations make it possible to characterize the different structures of the skin in terms of morphology, quantity and organization of its constituents. The use of this nomenclature considerably simplifies the computer processing of quantification parameters as well as the development of the program itself. The user can define, prior to step c), the measurement to be carried out by selecting descriptors chosen from lists of predefined descriptors. The choices can be done using a graphical interface. The descriptors can correspond to the four fields defined in the nomenclature. The tables below give the lists of different values ​​for each of the descriptors, each value can be identified by a code of a few letters. The name of the 5 quantization parameter is then given by the concatenation of these four codes, separated by The theoretically possible combinations are numerous, and some are uninteresting. For example, grain sizes are advantageously calculated in a region of the image with a homogeneous texture (ED, DT and DU, essentially). The system can be arranged to prohibit certain combinations. The table below gives some example codes. Quantification parameter name Meaning AFJJFJoy Average (AVG) calculated on the values ​​of the pixels belonging to the fiber mask (MF) taken from the raw signal (B) of autofluorescence (AF) AFJJTTSign Proportion of raw pixels (B) of significant AF (SIGN) in the total dermis (DT) AFSHG B DT SAAID WM (SHG B DT SIGN - AF B DT SIGN) / (AF B DT SIGN + V ■» aM •— «WM / X ««WH WW SHGJ_DTSIGN) AFJJJt Standard deviation of the raw AF signal over the entire image Segm_B_JDE_PosMoy Average depth of the JDE Table 1: Examples of quantization parameters. 15 Most measurements produce scalar values. Some, however, sleep vector values. In the case of vector measurements, in the output table we have, in addition to the data columns, a column indicating the length of the vector ("_n"), a second specifying the index of the first value ("_x0"), and a third giving the step of the indices 20 ("__dx"). The table below gives some examples of modality codes. Code Modality Definition AF Autofluorescence Signal obtained by adding the four time values ​​given by the microscope for each pixel for the AF channel SHG Second harmonic generation Signal obtained by adding the four time values ​​given by the microscope for each pixel for the SHG channel AFSHG Autofluorescence and second harmonic generation Both signals are taken into account jointly AFcorr Autofluorescence after attenuation correction Correction of the lateral and axial attenuation of the AF channel. Depth correction is done by fitting an exponential decay model to the average signal, frame by frame. Lateral correction, for a given channel, is done by calculating the average of all the images in a study for this channel, either by fitting a parabolic model to this average, or by using this average directly to correct the images.SHGcorr Second harmonic generation after attenuation correction Correction of the lateral and axial attenuation of the SHG channel. The correction is done in the same way as for the AF channel. AFSHGcorr Autofluorescence and second harmonic generation after attenuation correction Correction of the lateral and axial attenuation of both channels AFslopes Decay slope of the AF signal Slope of the linear regression of the logarithm of the temporal signals, pixel by pixel, in AF, Proportional to the inverse of the lifetime Segm Segmentation The measurements taken start directly from the segmentation of the skin. Table 2: Nomenclature - Modality codes. The preprocessing type indicates the processing applied to the initial data, identified by the modality code. The absence of preprocessing is indicated by the preprocessing code "B" for "raw data". The preprocessing method, allowing the enhancement of dermal fibers, is indicated by the preprocessing code "RF3D" for "10 fiber enhancement in 3D". The region of interest can be the entire image. When the region of interest is only part of the image, the calculations are done inside the region of interest and the pixels outside are not taken into account. The table below gives examples of region of interest codes for skin. Code Region of interest Definition 1 Entire image We take into account all the pixels of the image Surf Couplant / epidermis interface 2D junction surface between the couplant and the epidermis JDE Epidermis / dermis interface 2D junction surface between the epidermis and the dermis A Couplant Entire couplant, as given by the segmentation result ED Epidermis Entire epidermis, as given by the segmentation result EDsansSC Epiderm without stratum corneum Epiderm deprived of its superficial layer (stratum corneum disjunctum and stratum compactum). The thickness of this layer is given by the variable MEL MARGE DT Total dermis Entire dermis, as given by the segmentation result XXXTrl, XXXTr2 XXXTr3 First and second slice under the JDE of thickness XXX, then the rest of the region XXXTR1: set of pixels located under the JDE at a distance less than XXX pm from it.XXXTR2: set of pixels located under TR1 at a distance less than XXX pm from it. XXXTR3: set of pixels in the dermis not belonging to XXXTR1 or XXXTR2. XXX can take the values ​​10 and 20 microns. MFAF AF fiber mask Region corresponding to the elastic fibers in the dermis. To calculate it, we start from the enhanced AF image (RF3D), which we threshold. The threshold value is given by the average of the enhanced AF signal in the dermis, plus its standard deviation multiplied by the constant QUANTIF XSigma MF. MFSHG SHG fiber mask Region corresponding to the collagen fibers in the dermis. To calculate it, we start from the enhanced SHG image (RF3D), which we threshold. The threshold value is given by the average of the enhanced SHG signal in the dermis, plus its standard deviation multiplied by the constant QUANTIF XSigma MF. DUAF Useful dermis obtained from AF contrast Part of the dermis in AF containing elastic fibers that can be detected.It is obtained with a morphological closure of the AF fiber mask (MFAF) by a disk of diameter QUANTIF DIAM CAPIL. DUSHG Useful dermis obtained from the SHG contrast Part of the dermis in SHG containing collagen fibers that can be detected. It is obtained with a morphological closure of the SHG fiber mask (MFSHG) by a disk of diameter QUANTIF_DIAM_CAPIL. . DUinter Common useful dermis Intersection of DUAF and DUSHG MMs Mask of pixels attributed to melanin by thresholding This mask is made up of the pixels of EDsansSC, such that the AF value is greater than the MEL_SEUIL threshold. The processing is followed by a surface filter of size MEL FILTER AREA. MMp Mask of pixels attributed to melanin by the slope method This mask is made up of the pixels of EDsansSC, such that the AFpentes value is greater than the MEL_SEUIL_PENTES threshold. The treatment is followed by a detailed surface filter MEL FILTER AREA.xMMs Inverse mask of MMs Complementary in EDsansSC of MMs xMMp Inverse mask of MMp Complementary in EDsansSC of MMp . Table 3: Nomenclature - Codes of regions of interest. Measurements can be related to signal intensity, for example giving the mean of the signal intensity in the region of interest (Avg), the median of the signal intensity in the region of interest (Med), or the standard deviation of the signal intensity in the region of interest (St). The following tables focus more specifically on geometric measurements, measurements specific to epidermis quantification, and those specific to the dermis. Geometric measurements relate to the shape of the region of interest only: the 10 two fields “modality” and “preprocessing” have no influence on the measurement. Code Measure Definition EpMoy Average of the layer thickness enz(enpm) Average of the layer thicknesses in z (in pm) obtained at each point (x, y) of the image EpEt Standard deviation of the layer thickness in z (in pm) Standard deviation of the layer thicknesses in z (in pm) obtained at each point (x, y) of the image EpMin Minimum of the layer thickness in z (in pm) Minimum of the layer thicknesses in z (in pm) obtained at each point (x, y) of the image EpMax Maximum of the layer thickness in z (in pm) Maximum of the layer thicknesses in z (in pm) obtained at each point (x, y) of the image Vol3D Volume occupied by the layer (in pm3) Volume (in pm3) of the pixels of the layer PosMoy Average of the position of the interface (in pm) Average of the z position of the interface (in pm) relative to 0 (defined by the first xy image acquired) PosEt Standard deviation of the position of the interface (in.pm) Standard deviation of the interface z-position (in pm) from 0 (defined by the first acquired xy image) Amp Amplitude of variation of the interface position (in pm) (Max-Min) of the interface z-position (in pm) AreaN Normalized interface area Area of ​​the DEJ or skin surface, normalized by the area of ​​the image base surface (reference plane: (x,y)). This value is greater than 1, unless part of the DEJ disappears under the last image slice. AreaFront Area of ​​the border of the region of interest. Area, in pm2, of the border of the region of interest. Table 4: Nomenclature - Geometric measurement codes. Code Measure Definition MelOens . Melanin density - direct method Proportion of pixels considered to correspond to melanin in the ROI. Threshold: MEL_SEUIL MelDensPentes Melanin density - method based on temporal losses Proportion of pixels considered to correspond to melanin in the ROI considered, calculated from the temporal slopes. Only: MEt_SE«L_PE«ïES r / el'.ledDiGîr. Z Median of the distribution of melanin (direct method) as a function of normalized depth Evolution of the quantity of melanin as a function of depth, normalized in the epidermis.' In the output file, the median of this distribution is recorded. Number of classes in the distribution: MEt_MSOTBZ_CMSSES MetM-edDistribZ Slopes Median of the melanin distribution (slope method) as a function of normalized depth Evolution of the quantity of melanin as a function of depth, normalized in Itepideme.■ In the sort® file, we record the median of this ■ distribution. Number of classes in your distribution: MEL_DSTCBZ_CWSS£S MelDistribZ Distribution of melanin (direct method) as a function of normalized depth Evolution of the quantity of melanin as a function of depth, normalized in the epidermis. In the output file, we record the list of values ​​in this distribution. Number of downgraded distributions: MEL_DfSmBZ_CLASSES MelDistrfoZPen tes Distribution of melanin (slope method) as a function of normalized depth Evolution of the quantity of melanin as a function of depth, normalized in the epidermis. In the output file, we record your list of values. Number of classes in your distribution: UELJHSTRIBZ_CLASSES . Table 5: Nomenclature - Codes for quantitative measurements on the epidermis. Code Measure Definition Sign Proportion of significant pixels Ratio between significant pixels (i.e. pixels whose value exceeds the value of SIGN_SEUIL) and the total number of pixels contained in the region of interest. This measure is therefore unitless. SAAID "SHG-to-AF aging index of dermis" Ratio between, on the one hand, the difference between the proportions of SHG and AF pixels, and on the other hand, the sum of the proportions of AF and SHG pixels. As for the Sign measure, a threshold value must be defined, also given by SIGN_SEUIL. It follows from this definition that this measure can only be calculated on the two modalities taken together.SHGtoAF Ratio SHG / AF Ratio between the integrals of the SHG and AF signals, within the region of interest. It follows from this definition that this measure can only be calculated on the two modalities taken jointly. Imbr Imbrication between the two fiber networks AF and SHG Volume of the intersection of the two signals AF and SHG, divided by the volume of the union. The result is necessarily between 0 and 1 and is unitless. ImbrN Normalized imbrication between the two fiber networks AF and SHG Volume of the intersection of the two signals AF and SHG, divided by the volume of the union. - The AF and SHG signals are normalized so that their integral is equal.The result is necessarily between 0 and 1 and is unitless, AnisXYZ Anisotropy index In the xyz frame: Index indicating whether the orientation of the fibers is anisotropic or not: values ​​between 0 and 1: , 0 = isotropic (no preferred orientation) 1 = anisotropic (there is a preferred orientation) ThetaXYZ Preferred angle Theta (in degrees) In the xyz frame: preferred angle theta (azimuth) of the 1st eigenvector PhiXYZ Preferred angle Phi (in degrees) In the xyz frame: preferred angle Phi (elevation) of the 1st eigenvector, relative to the plane (xy).AnisJDE Anisotropy index In the local reference frame normal to the JDE: index indicating whether the fiber orientation is anisotropic or not: values ​​between 0 and 1: 0 = isotropic (no preferred orientation) 1 = anisotropic (there is a preferred orientation) PhiJDE Perpendicularity to the JDE In the local reference frame normal to the JDE: angle value between 0 (parallel to the JDE) and 90° (perpendicular to the JDE) Decr Signal decay in depth Slope of the line obtained by linear regression on the central ninth of the 3D volume (central part of the image) from the last part of the profile curve of the logarithm of the intensity. AED 2D Elongation Index Ratio between the volume of the disk corresponding to the geodesic diameter of each object, and the volume of the object The result is unitless and greater than 1. Ced is made 2d cut by 2d cut.The overall value is obtained by averaging the values, pixel by pixel, in the objects (the background is not taken into account). Code Measurement Definition AI3D 3D Elongation Index Ratio between the volume of the sphere corresponding to the geodesic diameter of each object, and the volume of the object. The result is unitless and greater than 1. The giobal ​​value is obtained by averaging the values, pixel by pixel, in the objects (the background is not taken into account). Tort2D 2D Tortuosity 2d tortuosity of the fibers. To binarize the signal, we use a single value: Mean + standard deviation * QUANTIFJ <Sigma~Tort2d Tort3D Tortuosité 3D Tortuosité 3D des fibres. Pour binariser le signal, on utilise un seuil de valeur : Moyenne(z) + écart-type(z) * QUANTiF_XSigma_Tari3d DiamG2D Diamètre géodésique 2D Diamètre géodésique des fibres en 2d. Pour binariser le signal, on utilise un seuil de valeur : Moyenne(z) + écart-type(z) QUANTIF_XSigma_DiamG2D DiamG3D Diamètre géodésique 3D Diamètre géodésique moyen des fibres en 3d.To binarize the signal, we use a threshold value: Mean(z) + standard deviation(z) * QUANTIF_XSigma_DiamG3D SkelBranch Number of skeleton branches Number of branches after disconnection (removal of triple points) from the 3D fiber skeleton. To binarize the signal, we use a threshold value: Mean(z) + standard deviation(z) * QUANTIF_XSigma_SkelBranch SkelLongBr Number of "long" skeleton branches. Number of "long" branches after disconnection of branches (removal of triple points) from the 3D fiber skeleton. A branch is considered long if its length is greater than the value of the QUANTIF_THRESH_SKEL_LENGTH variable. To binarize the signal, we use a value threshold: Average(z) + standard deviation(z) * QUANTIF_XSigma_SkelLongBr SkelLongMoy • Average branch length Average length of the skeleton branches.To binarize the signal, we use a threshold value: Mpy(z) + standard deviation(z) * QUANTIF_XSigma_SkelLongMoy At aGVoi[i] i=2,3,4 Values ​​of the anti-granulometric curve Three values ​​(at GRANU_1, GRANU_2 and GRANU_3, in microns) of Vantigranulumetry of the fibers. The maximum value of the curve is GRANU_MAX. Calculated in 3D from the fiber mask. Table 6: Nomenclature - Codes for quantitative measurements on the dermis Here are some examples of settings among the many combinations possible: • Example 1: Modality: SEGM Pretreatment: none 10 Region of interest: epidermis Measurement: average thickness Code according to the nomenclature: SEGMJB_ED_EpMoy This parameter calculates the average thickness of the epidermis, from the 3D segmentation. • Example 2: 5 Modality: SEGM Pretreatment: none Region of interest: IDE Measurement: standardized area Code according to the nomenclature: SEGM_B_JDE AireN 10 This parameter calculates the normalized area (3D generalization of the interdigitation index) of the dermo-epidermal junction from 3D segmentation. • Example 3: Modality: SHG Pre-treatment: 3D fiber enhancement 15 Region of interest: dermis Measurement: geodesic diameter Code according to the nomenclature: SHG_RF3D_DT_DiamG3D This parameter allows the calculation of the geodesic diameter of collagen fibers (SHG signal) in the dermis, after 3D fiber enhancement. 20 • Example 4: , Modality: AF Pre-treatment: 3D fiber enhancement Region of interest: first layer of dermis 10 pm below the DEJ Measurement: angle of fibers relative to the DEJ 25 Code according to the nomenclature: AF_RF3D_10TRl_PhiJDE This parameter calculates the orientation of elastic fibers (AF) relative to the DEJ (measurement of an angle in a rotating frame) from the fibers enhanced in the first 10 pm slice of dermis below the DEJ. • Example 5: 30 Modality: AFpentes Pretreatment: none Region of interest: epidermis Measurement: Z-distribution of melanin estimated by the Slopes method Code according to the nomenclature: AFpentes_B_ED_MelDistribZPentes. This parameter returns a vector representing the distribution of melanin as a function of the distance z (from the basal layer to the surface of the skin) taking as the criterion for detecting melanin the decrease in the autofluorescence signal analyzed by the slope method. 5 • Example 6: Modality: AFSHG Pre-treatment: 3D fiber enhancement Region of interest: sub-region of the dermis where the fibers are clearly visible Measurement: interweaving of the two fiber networks 10 Code according to the nomenclature: AFSHG_RF3D_DUinter_ImbrN_SD This parameter provides an idea of ​​the interweaving of the two fiber networks between them, calculated from the fibers enhanced in the region of the dermis called DUinter. According to another of its aspects, the invention relates to a method for evaluating the action of a stimulus and / or a treatment, in particular anti-aging, antioxidant, photo-protective, anti-pigmenting, pro-pigmenting or depigmenting, on a biological tissue, comprising the steps of: carry out a measurement following step c) of a method as described previously on a first part of the tissue which has not undergone the action of the stimulus and / or the treatment and obtain a first result, 20 - carry out a measurement following step c) of a method as described previously on a second part of the tissue having undergone an action of the stimulus and / or the treatment and obtain a second result, compare the first and second results, and evaluate the action of the stimulus and / or treatment from at least this 25 comparison. The first and second measurements can be made on identical or different parts. The method according to the invention can be used to evaluate the effects on pigmentation of certain products, for example dermocorticoids. 30 The treatment may be non-therapeutic, particularly cosmetic. Treatment may involve taking food supplements and / or medication. The treatment can be chosen from: application, injection, ingestion, inhalation of a product, in particular a cosmetic product. By “cosmetic product” is meant a product as defined in Directive 93 / 35 / EEC of 14 June 1993, amending Directive 76 / 768 / EEC. The product may have an anti-aging, antioxidant, photo-protective, depigmenting, anti-pigmenting, or pro-pigmenting effect. According to another of its aspects, the invention relates to a method for evaluating the action of a stimulus and / or a treatment, in particular anti-aging, antioxidant, photoprotective, anti-pigmenting, pro-pigmenting or depigmenting, on a biological tissue^ in which at least two regions of the tissue are exposed differently to the stimulus and / or treated differently, and in which said regions are compared, using a method as previously described, before and after exposure to said stimulus or to said treatment. To test the action of a product, we can compare an evaluation made after treatment with a placebo of the product and an evaluation made after treatment with this product. The placebo is, for example, the same cosmetic medium as the product used for the treatment but without the corresponding active ingredient(s). The treatment may correspond to the application of a product, particularly a cosmetic product. The product may, for example, be in the form of a cream, a lotion, an ointment, an oil, a powder, this list not being exhaustive. The product may also be contained in a carrier for application to biological tissues, including human, reconstructed or artificial skin, for example a patch, dressing, bandage or mask. The product may be intended for example for makeup, hydration, protection of biological tissues, in particular sun protection, or repair of biological tissues. The product may contain an active ingredient, for example an anti-aging, antioxidant, photo-protective, depigmenting, anti-pigmenting, or pro-pigmenting active ingredient. The active ingredient may be for therapeutic purposes, such as hydroquinone or retinoids, and / or cosmetic purposes. The invention may be better understood by reading the following description of non-limiting examples of its implementation, and by examining the figures in the appended drawing, in which: Figure 1 represents, schematically and partially, an example of a multiphoton device which can be implemented in a method according to the invention, Figure 2 illustrates different steps of a method according to the invention, Figures 3a to 3b represent in three-dimensional fashion the observation of the AF signals, respectively SHG, Figures 4a to 4d illustrate an example of acquisition of an image according to the invention, Figures 5a and 5b show the raw 2PEF and SHG data and the corresponding three-dimensional segmentation for a multiphoton image stack acquired in vivo on human skin, respectively. Figures 6a and 6b schematically represent the segmentation of the skin (A - coupling, S - coupling-epidermis surface, B - epidermis, J - dermo-epidermal junction, C - dermis) Figures 7a to 7d represent images of the skin acquired at different depths relative to the surface, Figures 8a to 8f illustrate at a given depth different modalities defined by the nomenclature according to the invention, Figures 9a to 9c illustrate an example of a process with filtering, and Figures 10 to 13 illustrate measurement results obtained by a method according to the invention. To implement a method according to the invention, any known multiphoton microscopy system, combined or not with a FLIM module for the specific detection of melanin, can be used. The excitation wavelength used can be between 700 and 1000 nm, preferably around 760 nm. This wavelength range allows satisfactory imaging for 2PEF and SHG signals of most tissues, particularly skin. Each acquisition produces two images of 511 x 511 x 70 voxels with a resolution of 0.255 pm in the (x,y) plane perpendicular to the optical axis and 2.346 pm in the third direction (z). Figure 1 shows, schematically and partially, an example of a multiphoton device 100 that can be implemented in a method according to the invention. The multiphoton 100 device is, for example, of the Dermalnspect® type marketed by the company Jenlab. The device 100 comprises a femtosecond laser 10, for example a Titanium-Sapphire (Ti:Sa) laser, tunable in a range of infrared wavelengths and capable of providing pulses of the order of 100 femtoseconds at a repetition rate of the order of 80 MHz. The laser 10 emits an infrared beam 11, which is directed towards a scanning device 12 of the laser beam in the form of an “XY scanner”. A scan of the laser beam can be obtained by angularly moving two galvanometric mirrors of the scanning device 12. The beam is then reflected by a dichroic mirror 14 and it is focused on the keratin materials 13 using the objective 15. Thus, the two galvanometric mirrors allow a scan of the focal point in the plane (x, y) perpendicular to the direction of propagation of the beam (z axis). A piezoelectric device makes it possible to translate the objective 15 and thus change the focusing plane in the keratin materials 13. In this way, it is possible to reconstruct the three-dimensional distribution of melanin in the keratin materials by superimposing the acquired images. To do this, the signals created at the focal point can then be detected either in epicollection through the excitation objective 15. The first dichroic mirror 14 makes it possible to select the multiphoton signals created, in particular the autofluorescence (AF) coming from the keratin materials 13. Then, a second dichroic mirror 16 makes it possible to separate the autofluorescence (AF) signals from the other multiphoton signals, corresponding for example to the second harmonic generation (SHG). In all cases, the signals pass through spectral filters 17 and arrive at standard detectors or at photon counting detectors (TCSPC0: Time-correlated single photon counting) for FLIM imaging 18, 19 which generate signals 18a and 19a which are transmitted to an electronic device 20 for acquiring and processing the signals. With the photon counting detectors the device comprises a FLIM imaging module for the specific detection of melanin. In this way, combined images, for example AF / SHG and 7 or FLIM images, can be obtained and these images processed to obtain the desired measurements. Step 1 of a method as illustrated in Figure 2 may consist of acquiring by multiphoton microscopy three-dimensional images comprising series of two-dimensional and / or temporal images at different depths of a sample of biological tissue. In the case of FLIM imaging, each image can be obtained by integrating the 2PEF and SHG signals over a time interval, for example using a TCSPC (Time-correlated single photon counting) photon counting device. A three-dimensional or three-dimensional temporal image can be formed from a mosaic of two-dimensional or two-dimensional temporal images at different depths. Each 3D image can contain several dozen images, for example between 50 and 100 images, especially between 65 and 80 images. Thus, Figures 3a and 3b represent a 3D image in the form of a mosaic of 2D images of the 2PEF signal integrated over time, respectively of the SHG signal integrated over time, at different depths. Multiphoton images acquired at depth contain acquisition-related noise: on the one hand, the scattering of excitation and multiphoton signals by the tissue leads to an attenuation of the signal at depth; and on the other hand, the increase in excitation power with depth increases the noise, because the excitation intensity at the surface of the sample becomes intense enough to create a multiphoton signal. This can lead to image blurring and thus disrupt subsequent measurements made from these images. This noise can be reduced using specific filtering, particularly in the case of the dermis. 3D Image Preprocessing - 3D Fiber Enhancement As shown in Figure 2, the method may include a step 2 corresponding to filtering. The filtering described below, and possibly used in the dermis, called 3D fiber enhancement, is particularly useful for improving the characterization of collagen fiber and elastic fiber networks. This enhancement is carried out in three dimensions, despite the difficulty that can arise from a significant difference in sampling steps between the different dimensions of the images, the sampling step in the plane parallel to the skin being able to be ten times smaller than that of the direction orthogonal to this plane. 25 It is a linear, orientable and highly anisotropic filter for enhancing fibers and eliminate all other structures. Let h: R3 —R be the impulse response of the filter and H its frequency response, H; Rs—* c, related to h by the Fourier transform H The filter is designed by adapting its impulse response to the shape of the structures to be enhanced. For fibers, the shape of h will be an elongated ellipsoid, rotationally symmetrical around the main axis z. The two secondary axes are denoted x and y. Let hx, hy and hz be the centered sections of h along the x, y and z axes. Thanks to the rotational symmetry of h around 2, we have hx = hy. The section of h perpendicular to the z axis and centered forms a Laplacian profile, obtained by subtraction of two Gaussians, hx = hy = G^-G» , with g the standard deviation. This parameter o is adjusted to the radius of the fibers, a is for example fixed at 0.765 pm. The section centered along the principal z axis forms a Gaussian profile Gc, with a fixed at 2.5 pm. The fibers have limited curvature, and to a first approximation, they can be considered linear over at least a length of 5 pm. The filter gain is then normalized to 1, by setting H = H / xs.:i H . We use the approach proposed by WT Freeman, EH Adelson in “The Design and Use of Steerable Filters”, IEEE Trans. On Pattern Analysis and Machine Intelligence, vol. 13, No. 9, September 1991 This method is based on the following principle. To enhance the fibers oriented at angle y, h is taken as a copy of h rotated by y, \| / =(0, ¢), where 0 and ¢ denote the azimuth and elevation. Then, it is discretized in Z3 with the same sampling steps as the image, dx=dy=Q,255208 pm and dz=2,346 pm. To enhance all fibers regardless of their orientation, we proceed by angular sampling. Let W = . be a set of discrete orientations, and let the filter bank oriented along all orientations be \|A. The angular discretization d\| / is fixed with a step of d0 = 30° and d0 = 45°. The filtering of the image f by the filter bank is carried out as follows: Wj s with F-1, the inverse Fourier transform. The filter gives a maximum response for an object with geometric properties similar to h and oriented along y. Thus, the filtered image, / ', contains the enhanced fibers; all other structures have been attenuated. In step 3, the observed volume is segmented. In the case of skin separated into three zones, the low values ​​of 2PEF are used for the coupling medium; the high values ​​of 2PEF for the epidermis; and the high values ​​of SHG for the dermis. Only the connected components of the binary images thus obtained which touch reference planes are kept, chosen for example such that z=0 for the coupling medium; z=15 for the epidermis, and such that the integral of the SHG signal is maximal for the dermis. Conflicts (i.e., a non-empty intersection) between epidermis and dermis markers may occur. In this case, priority is given to the dermis marker, because SHG is a more discriminating signal than 2PEF. This means, in practice, that if a voxel belongs to both dermis and epidermis markers, it is then removed from the epidermis marker. First, the boundary between the coupling medium and the epidermis is calculated, then the boundary between the epidermis and the dermis, using the previously defined preferential algorithm. It should be noted, however, that the low-level segmentation calculation is only performed once. Step 4 of Figure 2 corresponds to the actual measurement by image processing based on a nomenclature as defined previously. Figures 4a to 4d illustrate a time sequence of three-dimensional images (0-2ns; 2-4ns; 4-6ns; 6-8ns) of which only four two-dimensional time-lapse images are shown acquired in vivo on the forearm skin of a healthy volunteer, near the basal layer of the epidermis at a depth of 50 um from the skin surface. In each pixel representing each signal, we have four values, acquired consecutively over time. For each time frame, the fluorescence signal was integrated over a period of 2 ns. This information is used for the specific detection of melanin. Segmentation allows the definition of specific regions of interest within the epidermis or dermis. These regions of interest can correspond to skin constituents such as dermal fibers like elastic fibers or collagen fibers, epidermal cells, or melanin. The particular measurements carried out using the invention may relate to a region. Figure 5a illustrates three-dimensionally the 2PEF and SHG signals in three main regions (couplant, epidermis and dermis) and Figure 5b represents the same regions after segmentation. In this example, the 3D image is segmented into 3 areas (couplant, epidermis, dermis) coded with three values: a value of 6 is assigned to the couplant, a value of 9 to the epidermis and 12 to the dermis. Figures 6a and 6b give different examples of skin segmentation into several regions of interest. The coupling medium A is here water, but can change into depending on the objectives used (dry, oil immersion or other medium). While Figure 6a shows that the method according to the invention makes it possible to define slices in the epidermis B, for example three slices equidistant between the surface of the skin S and the DEJ J, the user also has the possibility of defining in the dermis C slices whose boundaries are normal to the DEJ J and of thickness defined as illustrated in Figure 6b. The slices t1 and t2 shown are of fixed thickness, for example 10 or 20 μm. Figures 7a to 7d represent multiphoton images of dimensions 130x130 pm2 of human skin observed in vivo. The skin observed here is normal and does not present any pathology. In the observed images, the 2PEF signal reveals the comeocytes in the stratum comeum observed in Figure 7a, the keratinocytes in the stratum granulosum shown in Figure 7b, and the stratum basale shown in Figure 7c. In the dermis, as shown in Figure 7d, the 2PEF signal reveals the elastic fibers while the SHG signal is characteristic of the distribution of collagen fibers. Raw images as illustrated for example in figures 3a, 3b, 4a to 4d are processed in order to create other images called calculated images which constitute as many modalities which can serve as a starting point for the determination of the regions of interest, the possible pre-processing and the calculation of the measurements. The method manipulates 3D images, but the examples given in Figures 8a to 8f are illustrated in 2D for clarity. Figures 8a to 8f correspond to two-dimensional images of human skin extracted at a depth of 50 pm from a 3D image. Figures 8a and 8b show two images, represented in gray levels, obtained by integration, in each pixel, of the four temporal values ​​and corresponding to the two 2PEF and SHG channels Figures 8c and 8d illustrate two images, represented in gray levels, called corrected, corresponding to the two 2PEF and SHG signals, in which the attenuation of the signal, both due to the depth and to the distance from the center of the field of view (vignetting phenomenon), has been compensated. Figure 8e shows a parametric image, obtained from the 2PEF image, and which represents, in each pixel, the decay rate of the fluorescence signal, obtained thanks to the four temporal values ​​previously mentioned, linked to the fluorescence lifetime; and Figure 8f shows a segmentation of the image into three regions, corresponding to the couplant, the epidermis and the dermis. Figures 9a to 9c illustrate the effect of specific “fiber enhancement” filtering. Figure 9a represents the SHG signal in the dermis by a 2D slice extracted from the 3D representation, Figure 9b represents the corresponding slice after filtering (enhanced image). Figure 9c gives a 3D image of the fiber volume after enhancement. TESTS CARRIED OUT This method has been used to analyze several databases. A. Photoaging Study The “Photo Aging” database is made up of data from 15 young subjects (18-25 years old) and 15 elderly subjects (70-75 years old); For each subject, the ventral and dorsal sides of the forearm were studied. 484 quantification parameters were defined from the analytical nomenclature according to the invention, among which the analysis of the data obtained made it possible to identify 62 parameters highlighting differences either according to age, or according to the area, or both. For example, a first parameter defined by: Code according to the nomenclature: AF_RF3D_DT_SkelLongMoy Modality: AF Pre-treatment: 3D fiber enhancement Region of interest: Total dermis Measurement: SkelLongMoy- average length of the branches of the elastic fiber skeleton The quantification parameter “AF_RF3D__DT_SkelLongMoy”, characterizing the average length of the branches of the elastic fiber skeleton, makes it possible to highlight an age effect on the elastic fibers as shown in Figure 10. Furthermore, the parameter defined by: Code according to the nomenclature: SHG_B_DT_Sign_SD Modality: SHG Pretreatment: no Region of interest: Total dermis Measure: Sign - thresholded pixel density allows us to highlight a difference in the quantity of collagen fibers between the dorsal D and ventral V faces of the forearm as illustrated in Figure 11. A final example of a quantization parameter defined as below: Code according to the nomenclature: AFSHG_RF3D_DUinter_ImbrN_SD (Corresponding to example 6) Modality: AFSHG Pre-treatment: 3D fiber enhancement 5 Region of interest: sub-region of the dermis where the fibers are clearly visible Measurement: ImbrN - interweaving of the two fiber networks This parameter highlights a double effect, age and area on elastic fibers and collagen fibers as shown in Figure 12. 10 B. Constitutive pigmentation study Code according to the nomenclature: AFpentes_B_ED_MeIDistribZPentes (Corresponding to example 5): Modality: AFpentes - slope images of the AF signal Pretreatment: no 15 Region of interest: epidermis Measure: MelDistribZSlopes - normalized melanin distribution, obtained from the AF signal decay slopes Another database, derived from the "constitutive pigmentation" study, consists of images from subjects belonging to 20 skin color typological groups ranging from I (light skin) to VI (very dark skin). Figure 13 shows, for two subjects with different phototypes, the variation in melanin density expressed in % as a function of the normalized distance from the basal layer to the surface expressed in number of normalized slices defined within the epidermis. 25 Curve (a) thus obtained corresponds to group I (very light skin) while curve (b) corresponds to group III (intermediate skin). We observe that the measured parameter clearly highlights the difference in phototype, the distinction is particularly marked at the level of the basal layer of the epidermis (abscissa 1 on the graph). 30 The invention is not limited to the illustrated examples. Considering the speed of the calculations, a method according to the invention can be used in a screening method. A method according to the invention can be used for the characterization of reconstructed skin models. A method according to the invention can be used to study any effect impacting the skin (pharmacological, physical, pathological, etc.) A method according to the invention can be used to analyze any type of tissue containing melanin, cells, collagen or elastin. In particular, the method 5 according to the invention is not limited to the skin, it can be applied to any type of tissue whose constituents are imaged using endogenous signal sources or exogenous sources. The expression "comprising a" should be understood as being synonymous with "comprising at least one", unless otherwise specified. 10

Claims

CLAIMS 1. Method for observing a biological tissue, comprising the steps of: a) acquiring by multiphoton excitation at least one three-dimensional image for each of the 2PEF and SHG signals emitted by said tissue, and determine by image processing from these three-dimensional images a segmentation of said tissue into at least three regions.

2. Method according to claim 1, the tissue comprising melanin.

3. Method for observing a biological tissue, comprising the steps of: a) acquiring by multiphoton excitation at least one three-dimensional image for each of the 2PEF and SHG signals emitted by said tissue, and b) determining by image processing from these three-dimensional images a segmentation of said tissue into at least two regions, the tissue comprising melanin.

4. Method according to any one of the preceding claims, the fabric comprising elastin.

5. Method for observing a biological tissue, comprising the steps of: a) acquiring by multiphoton excitation at least one three-dimensional image for each of the 2PEF and SHG signals emitted by said tissue, and b) determining by image processing from these three-dimensional images a segmentation of said tissue into at least two regions, the tissue comprising elastin.

6. Method according to any one of the preceding claims, the tissue comprising collagen.

7. Method according to any one of the preceding claims, step b) comprising the calculation of markers corresponding to each region and the propagation of the markers.

8. Method for observing a biological tissue, comprising the steps of: a) acquiring by multiphoton excitation at least one three-dimensional image for each of the 2PEF and SHG signals emitted by said tissue, and b) determining by image processing from these three-dimensional images a segmentation of said tissue into at least two regions, step b) comprising the calculation of markers corresponding to each region and the propagation of the markers.

9. Method according to one of the two preceding claims, the propagation of the markers comprising the application of a watershed algorithm.

10. Method according to one of the three preceding claims, the propagation of the markers comprising the generation of a graph comprising nodes connected by edges, and to each edge (e) connecting two neighboring nodes corresponding to two regions A and B is assigned a weight œ (e) such that: œ (e) = S (s(p,q) / (l+Li(l(p), l(q))) for (p, q) e N(A, B) where s(p,q) is the surface separating voxels p and q, and Li(I(p), I(q)) is the distance associated with the Li norm.

11. Method according to any one of the preceding claims, the fabric being skin, comprising a step of filtering by raising fibers.

12. Method according to any one of the preceding claims, the three-dimensional images being obtained by integration of the fluorescence signal and the second harmonic generation signal carried out at different successive time intervals, in particular of duration between 1 and 3 ns.

13. Method according to the preceding claim, step a) comprising the acquisition of between three and five three-dimensional temporal images for each of the 2PEF and SHG signals.

14. A method according to any preceding claim, further comprising the step of: c) carry out a measurement in at least one of the regions resulting from the segmentation.

15. Method according to the preceding claim attached to claim 12 or 13 for detecting, quantifying and / or visualizing a fluorescent constituent, within a biological tissue, step c) comprising: determining, by performing a linear regression of the logarithm of the fluorescence signals, the spatial distribution in the sample of the slopes of said linear regression, and the generation of at least one piece of information on the presence and / or the volume or surface density of said constituent within the tissue at least from this spatial distribution.

16. Method according to claim 14, according to which prior to step c) the user defines the measurement to be carried out by selecting descriptors chosen from lists of predefined descriptors.

17. Method according to the preceding claim, the descriptors respectively indicating: • the category(ies) of signals considered, • the presence or absence of pretreatment and the type of pretreatment used; • the region of interest, on which the measurement is carried out; and • the nature of the measure itself.

18. Method according to one of the two preceding claims, a code representative of the measurement being generated from values chosen by the user.

19. Method according to any one of the preceding claims, being carried out in vivo on a biological tissue.

20. Method for observing a biological tissue, implemented in vivo on the biological tissue comprising the steps of: a) acquiring by multiphoton excitation at least one three-dimensional image for each of the 2PEF and SHG signals emitted by said tissue, and b) determining by image processing from these three-dimensional images a segmentation of said tissue into at least two regions 21. A method according to any preceding claim, the tissue being human skin, reconstructed or artificial skin, or cultured cells.

Images