Polarimetric multispectral imaging

Abstract

Polarimetric multispectral imaging Multispectral polarimetric imaging system (10), in particular for colposcopy, comprising: an illumination system comprising at least one light source (50), this illumination system emitting in at least two, in particular at least three, spectral bands, a polarization state generator (PSG) arranged downstream of the light source and upstream of a target (T) to be imaged, a polarization state analyzer (PSA) arranged downstream of the target to be imaged, a multi-sensor camera (5) comprising at least two, better still at least three, sensors (601, 602, 603) for recording respectively at least two, better still at least three, images in said spectral bands Figure for the abstract: Fig. 2

Description

Title of the invention: Title Polarimetric multispectral imaging Technical field

[0001] The present invention relates to the field of polarimetric imaging, and more particularly but not exclusively to polarimetric imaging applied to colposcopy. State of the art

[0002] Colposcopy involves examining the cervix using a device called a colposcope. The colposcope allows for remote observation of the cervix without any contact between the optics or other components of the colposcope and the tissue to be examined.

[0003] The colposcope includes an illumination system allowing the uterine cervix to be illuminated through a speculum introduced into the vagina and the cervix to be observed from a distance.

[0004] Illumination is typically carried out using white light.

[0005] The optical system of the colposcope may be connected to a color camera that allows images to be recorded as well as the cervix to be observed on a computer screen. In some cases, the optical system of the colposcope is also connected to eyepieces that allow direct observation of the cervix by the user.

[0006] The colposcope may comprise a head mounted on an articulated arm, and may comprise two left and right eyepieces communicating with two respective light input ports present on the front face of the head, oriented towards the area to be examined, and next to these two input ports, a light output port serving to illuminate the area of ​​interest; the light is produced for example by a lamp arranged outside the head and conveyed to the latter by a light guide comprising a bundle of silica optical fibers.

[0007] Furthermore, Mueller polarimetric imaging consists of measuring the Mueller matrix of a target sample and can provide various useful information on the nature of the sample, in particular by allowing its optical anisotropy and its light scattering properties to be analyzed.

[0008] Mueller polarimetry imaging is of interest for the ex vivo and in vivo study of various biological tissues and in particular the uterine cervix.

[0009] Patent EP 1 738 682 describes means for implementing a polarimetric image in colposcopy.

[0010] Indeed, the knowledge provided by polarimetric information makes it possible to improve the quality of medical diagnosis, for different types of pathologies and in particular for the early detection of cervical cancer.

[0011] In particular, Mueller polarimetric imaging is a technique that allows the complete polarimetric characterization of a sample through the measurement of its Mueller matrix.

[0012] A Mueller polarimeter is generally composed of a light source, a polarization state generator (PSG), a polarization state analyzer (PSA) and a detector.

[0013] A Mueller polarimeter can provide a point measurement. In this case, a photodetector can be used as the detector. In the case of a point Mueller polarimeter, only one Mueller matrix is ​​measured.

[0014] But a Mueller polarimeter can advantageously work as an imager. In this case, a CCD or CMOS camera is used as a detector. Each pixel of the camera is the equivalent of a photodetector for a point Mueller polarimeter. In the case of an imaging Mueller polarimeter, a Mueller matrix is ​​measured for each pixel.

[0015] Mueller polarimetric imaging is an inherently slow technique because it requires the acquisition of multiple intensity images to measure the Mueller matrix of a sample.

[0016] In the case of an imaging Mueller polarimeter, the measurement of light intensity is done simultaneously for all pixels of the camera used. For each pixel, the light signal is transformed into photoelectrons. The modality with which the electrical signal produced by the photoelectrons is processed, as well as the number of pixels in the region of interest (ROI) of the selected sensor, determine the number of images acquired per second (“frames per second” or FPS).

[0017] If several wavelengths must be acquired to explore a biological tissue at different depths, Mueller polarimetric imaging becomes even slower if these wavelengths are acquired in succession, which is the case for most Mueller polarimeters currently in use.

[0018] However, multispectral analysis is crucial for exploring biological tissue at different depths. Indeed, the shorter wavelengths of the visible spectrum, corresponding for example to blue and green colors, are strongly absorbed by hemoglobin and mainly allow the surface of biological tissues to be explored. On the contrary, red / near infrared light, which is much less absorbed by hemoglobin, allows biological tissues to be explored more deeply. The penetration length of light in the red / near infrared part of the spectrum increases with wavelength.

[0019] Many different types of Mueller polarimeters exist in the literature. The most common are time-sequence Mueller polarimeters which perform intensity measurements in succession.

[0020] Mueller polarimetric imaging is done in three main steps: 1. Measurement of the intensity matrix (B = AMW according to the established notation). This matrix can be acquired several times in order to increase the signal-to-noise ratio with an averaging process, thus improving the reliability of the measurements, which can be called the measurement step. 2. Obtaining the Mueller matrix M=A *BW *, which can be called the processing step. 3. Determination of the relevant polarimetric parameters from M, which can be called the post-processing step, using for example different algebraic methods, such as Mueller matrix decompositions.

[0021] For biomedical applications, step 1) must be completed in a maximum time of the order of a second to reduce the blurring effects due to involuntary movements of patients (breathing, heartbeat, etc.) during the measurement. Steps 2) and 3) must ideally be completed very quickly to restore the images useful to practitioners as soon as possible after the measurement, with a delay of a few seconds.

[0022] Step 2) is generally very quick and can be accomplished quite easily.

[0023] Step 3) can on the contrary be quite slow, in particular if it requires example the calculation of the eigenvalues ​​of the matrix M, which can often be the case for the decompositions of Mueller matrices or other algebraic treatments.

[0024] To determine the Mueller matrix of a sample, using a time-sequence Mueller polarimeter, it is necessary to carry out a succession of light intensity measurements, at least 16 in number for a 4x4 Mueller matrix, using a PSG and a PSA allowing respectively to modulate the polarization of the light sent to the sample and to analyze the polarization of the light received after interaction with the sample. Different combinations of the PSG and the PSA make it possible to acquire the 16 intensity images necessary to obtain the Mueller matrix of the sample.

[0025] For 4x4 Mueller polarimetric imaging, at least 16 intensity images must be acquired, which are grouped into the 16-component intensity matrix B.

[0026] The use of polarized cameras allows the acquisition of 12 intensity images which give access to a 3x4 Mueller matrix. For biological tissues, with very particular polarimetric properties, it is possible to go back to the 4x4 Mueller matrix with algebraic calculations starting from the measured 3x4 Mueller matrix.

[0027] However, several factors are limiting for the use of polarized cameras. Polarized cameras are generally monochromatic. Several cameras are therefore necessary to reconstruct a color image.

[0028] Additionally, with these cameras, each pixel is divided into 4 sub-pixels, each sub- pixel corresponding to a different polarization state. The light intensity is therefore divided by four. In addition, a further loss of light is due to the presence of a polarizer on each pixel. Pixels divided into four sub-pixels also cause a loss of image resolution. Finally, there can be crosstalk between different sub-pixels because their polarization states are not completely separated.

[0029] For biomedical applications, it is necessary that the matrix B be acquired within a time of about one second and that the relevant polarimetric parameters, calculated during the post-processing step, be restored within a few seconds.

[0030] Acquiring the intensity matrix B in about one second is also necessary to reduce blurring effects due to involuntary movements of the patient during the measurement. In one second, the matrix B can be acquired several times to increase the signal-to-noise ratio with an averaging process.

[0031] The restitution of Mueller polarimetric parameters in a few seconds, after post-processing using for example the decompositions of the measured Mueller matrices, is thus crucial to allow the practitioner to have useful information very quickly during the visit.

[0032] The ideal would be to be able to restore the measured Mueller matrix in real time, or even to restore the relevant polarimetric parameters (after post-processing of the Mueller matrix) in real time.

[0033] However, it is very difficult to obtain a stream of images returned to the user with a sufficient frequency for comfortable observation, substantially in real time by the practitioner during the clinical examination.

[0034] The first limitation is the speed of acquisition of the intensity images necessary to obtain the intensity matrix B and therefore the Mueller matrix M.

[0035] The second limitation is related to obtaining the main polarimetric parameters of M with a post-processing process. Most Mueller matrix decompositions for example require the calculation of the eigenvalues ​​of the Mueller matrix M, which represents a computational process that can hardly be parallelized and accomplished in real time.

[0036] Indeed, in a conventional white light imaging system, for smooth real-time tracking of an image, the flow should generally be at least 8 FPS if the target moves very slowly. Thus, in the case of Mueller polarimetric imaging, the number of images to be acquired and processed is at least 8*16 = 128 if 16 measurements are to be made between two displayed images, which can be a rather high frequency for high-resolution images.

[0037] Some methods allow the number of measurements to be reduced but do not allow the acquisition of a complete image in return. Others are based on relatively complex specialized sensors such as polarized cameras, which allow faster acquisition but have other significant limitations described above.

[0038] Real-time imaging becomes difficult to accomplish if a multispectral approach is required and if the different wavelengths are acquired in succession. One possibility would be to use several monochromatic CCD or CMOS cameras simultaneously, which would make the polarimetric system bulky and difficult to integrate into an existing imaging system or within a completely new system to be used in medical practice.

[0039] Several problems therefore arise in adapting Mueller polarimetry to colposcopy or other imaging techniques.

[0040] The realization of Mueller polarimetric imaging in real time is a first challenge. It involves the restitution in real time of the Mueller matrix after the measurement and the processing step, and even better of the most relevant polarimetric parameters after the post-processing step of the Mueller matrix, in order to allow the practitioner to have access to information relating to the microstructure of the tissue at different depths.

[0041] A second challenge is the realization of step 1) above simultaneously for several wavelengths of the visible / near infrared spectrum.

[0042] A third challenge is to also provide real-time reference color imaging, which is necessary for selecting the area to be analyzed with polarimetry, thus allowing the practitioner to have well-known spatial references, which would not be possible by directly displaying the polarimetric images.

[0043] Another challenge is the possibility of superimposing pixel by pixel a color image, light intensity images at different wavelengths and polarimetric images at different wavelengths, which is crucial for the efficient analysis of the images using image processing algorithms or learning algorithms. This step can allow, among other things, to select the most relevant parameters. It can also allow to determine the most relevant combination of polarimetric and non-polarimetric parameters for the diagnosis. Finally, it can allow to merge together the polarimetric and non-polarimetric parameters of interest in a single image or in a limited number of images to provide practitioners with simplified and optimized information for the diagnosis which is not directly observable in conventional images or in the initial non-merged polarimetric images.

[0044] Another issue may be the restitution of the combined images substantially in real time.

[0045] Finally, a last challenge is to have a very compact and fa Mueller polarimeter easily adaptable, ergonomically, to different existing imaging systems, such as a colposcope or an endoscope or even a microscope (or exoscope) for neurosurgery, or easily usable for the creation of a new ergonomic and space-saving imaging system to be used in medical practice according to the intended application. Statement of the invention

[0046] Consequently, there is a need to benefit from high-performance polarimetric imaging systems, in particular polarimetric colposcopes, in order to allow more comfortable multispectral and polarimetric visualization of the area examined, in particular in real time.

[0047] More generally, there is a need for a solution to increase the image frequency for any imaging technique such as polarimetry, requiring a large number of measurements to be taken before an image can be generated, so as to make the viewing of the images by the practitioner more fluid during the examination, and thus facilitate decision-making and / or reduce the duration of the examination, among other things.

[0048] Furthermore, independently of the above, there is also an interest in: - facilitating the post-processing of the Mueller matrix in order to reduce the need for computing resources, and reducing the time required for calculating polarimetric images, - facilitate the production of an imaging system, in particular colposcopy, capable of both multispectral imaging and polarimetric imaging, and in particular of allowing the easy transformation of a conventional observation system such as a conventional colposcope into an observation system allowing multispectral and polarimetric imaging, - improve the illumination system of a colposcope, in particular with a view to improving the quality of the images, - generally improve the ergonomics of an observation system such as a colposcopy system, and / or - facilitate the development of the system, in particular to carry out calibration.

[0049] The invention aims to meet all or part of the needs identified above. Summary of the invention Multispectral polarimetric imaging

[0050] According to a first of its aspects, the invention relates to a multispectral polarimetric imaging system, in particular polarimetric colposcopy, comprising: - an illumination system comprising at least one light source, this illumination system emitting in at least two, better still at least three, spectral bands, - a polarization state generator (PSG) arranged downstream of the light source and upstream of a target to be imaged, - a polarization state analyzer (PSA) placed downstream of the target to be imaged, - a multi-sensor camera comprising at least two, better still at least three, sensors, for recording respectively at least two, better still at least three, images in said spectral bands.

[0051] The polarimetric imaging system may be macroscopic or microscopic, and work in transmission or reflection, in free space or not. Preferably, it is macroscopic and works in reflection, in particular in free space (such as a colposcope). The imaging system may also be used for other biomedical applications, such as brain surgery or endoscopy, and non-biomedical applications, for example in the field of cosmetics or microelectronics.

[0052] The use of a light source emitting in several spectral bands in combination with the use of a multi-sensor camera makes it possible to generate images in different wavelength ranges that are perfectly superimposable, since by construction, the different sensors of the camera produce perfectly superimposable images, that is to say that for all the pixels of the image, the same point on the image corresponds to the same pixel on each of the sensors. In particular, parallax can be avoided between the different types of images produced by the camera, in particular color and polarimetric images, since the same camera is used to produce the color and polarimetric images and the system produces images of the same size on each of the sensors and positioned identically on each of the sensors, thus allowing pixel-to-pixel superimposability.

[0053] Preferably, there are three spectral bands and they range, for example, from 445 nm to 475 nm for the first, from 510 nm to 550 nm for the second, and from 600 nm to 660 nm for the third, being, for example, centered respectively on 460 nm, 530 nm and 630 nm.

[0054] Preferably, the camera is a bi-CCD, bi-CMOS, tri-CCD, tri-CMOS, 4-CMOS or 4-CCD camera, preferably tri-CCD, tri-CMOS, 4-CCD or 4-CMOS. Also preferably, the spectral bands are respectively in the red or near infrared, green and blue. For example, for a 2-CCD camera, the spectral bands are preferably visible and near infrared, for a 3-CCD camera, the spectral bands are blue, green, and red / near infrared, and for a 4-CCD camera, the spectral bands are blue, green, red and near infrared.

[0055] This allows the use of fast cameras, manufactured on a large scale. A camera multi-sensor provides an independent grayscale intensity image for each sensor, these grayscale images can be used on the one hand to perform measurements of the intensity matrix coefficients in the case of Mueller polarimetry and on the other hand they can be combined to reconstruct a color image of the observed area.

[0056] The camera may comprise at least two, better still three, dichroic prisms, to separate the wavelengths towards the different sensors. A tri-CCD camera for example comprises three dichroic prisms, the sensor for detecting blue being fixed on the first prism, the sensor for detecting red being fixed on the second prism, itself fixed on the first prism, and the sensor for detecting green being fixed to the third prism, itself fixed to the second prism, in a manner known per se. Other arrangements are possible, this type of camera being well known.

[0057] Preferably, the polarization state generator (PSG) comprises electrically controllable liquid crystal polarization modulators, which allows for relatively fast switching, good compactness, suitable bandwidth and does not disturb the superposability of the images. The liquid crystals also allow the implementation of full-field polarimetric imaging for the analysis of surfaces on a macroscopic scale as well as microscopic polarimetric imaging.

[0058] Alternatively, the polarization state generator comprises, for example, mechanically controllable optical components.

[0059] Similarly, the polarization state analyzer preferably comprises electrically controllable liquid crystal polarization modulators. Alternatively, the polarization state analyzer comprises mechanically controllable optical components. Preferably, the liquid crystal polarization modulators are ferroelectric liquid crystal modulators, which allows a relatively high switching frequency, with a control voltage of a few volts. Alternatively, the liquid crystal polarization modulators are nematic liquid crystal modulators.

[0060] The light source is preferably a white light source, in particular a xenon lamp. Such a lamp makes it possible to emit in a broad spectrum and to produce both good color and polarimetric images. Alternatively, it is a halogen type lamp, at least one LED or a source of any other type.

[0061] When using a xenon lamp, the illumination system may include a dual-band, or better, tri-band, dichroic filter downstream of the light source. This filter can block UV and IR, for example, and allow white light to pass through, in particular the red, green and blue bands of the visible spectrum. This filter can be removable, in particular mounted on a filter wheel, in order to allow for example acquisition in the IR or near IR, or even the near UV. Alternatively, depending on the type of source used, the illumination system can include a dual-band filter, better tri-band, or at least a single-band filter.

[0062] Alternatively, the illumination system comprises several light sources emitting respectively in each of said spectral bands, for example blue, green and red, or blue and yellow LEDs.

[0063] The imaging system may comprise a polarimetric system according to another aspect of the invention, as defined below.

[0064] The imaging system may include a filter wheel disposed directly upstream of the camera, and possibly a linear polarizing filter of the analyzer, as described later. Alternatively, the filter wheel is disposed directly downstream of the source.

[0065] The filter wheel may carry a tri-band filter, as mentioned above, to allow for example the acquisition of images at wavelengths in spectral ranges centered around 460nm, 530nm and 630nm by the camera, as well as one or more monochromatic dichroic filters, to allow for example the acquisition of images in spectral ranges around 650nm and 700nm. The sensor intended to image red may also allow near infrared imaging.

[0066] The filter wheel may allow the selection of different spectral ranges in the blue, green and red / near infrared by changing the tri-band filter, removing the filter, replacing the type of filter used, for example replacing a tri-band filter with at least one monochromatic filter. The filter wheel may allow certain wavelength ranges to be explored in more detail.

[0067] The imaging system may comprise a processor for generating a Mueller polarimetric image of the target in each of the spectral bands of interest. By Mueller polarimetric image is meant an image comprising at least one piece of information extracted from the Mueller matrix, for example a polarimetric property such as diattenuation, polarisance, birefringence, depolarisation, etc., as well as any information at least partially described from the knowledge of one of these properties. In the case of colposcopy for the analysis of the uterine cervix, knowledge of the birefringence and depolarisation properties are particularly useful for diagnosis.In particular, in the case of use in colposcopy, polarimetric parameters that are particularly useful for diagnosis include linear phase retardation, the degree of randomness of the azimuth of the slow (or fast) axis of the linear phase retardation, the combined image of the linear phase retardation and . the azimuth of the slow (or fast) axis of linear phase retardation, depolarization, linear correlation between depolarization and linear phase retardation, especially at 530 nm.

[0068] The processor may be configured to at least partially superimpose, on an image corresponding to the non-polarimetric observation by the camera (for example the basic color image), an image containing at least one piece of polarimetric information. This polarimetric information may be displayed in false color.

[0069] The system can be arranged to carry out, in particular with each of the sensors of the camera and in parallel for all the different sensors, successive real series of measurements on a given sample corresponding to the acquisition of at least 9, in particular 12, better 16 coefficients of the intensity matrix B,

[0070] the processor being arranged to: - generate series of n additional measurements, also called series “artificial”, from a grouping of measurements from a series M of n measurements of a given rank, and at least one series of a different rank, in particular the following rank, and a permutation P(p) of the measurements within each artificial series so as to respect the predefined order of the n measurements within each series, then - generate a stream of polarimetric images at a frequency higher than that which would be possible with said real series of measurements without the artificial series of measurements, from the polarimetric images produced from the real series of measurements and from the polarimetric images generated from the artificial series interposed between the real series.

[0071] For a 4*4 Mueller matrix, we have

[0072] B=AMW,

[0073] where A=[SA1,SA2,SAj,Sa4]T

[0074] eL\y = [ Sw(, Swÿ Sw,, SwJ

[0075] the measurements being obtained using the PSG polarization state generator which produces four independent states described by four Stokes vectors Sw (i = 1,2, 3,4) and the PSA polarization state analyzer which generates analysis states described by four Stokes vectors Sa (i = 1,2, 3,4),

[0076] This increases the image frequency, which improves the comfort of viewing the images, and allows real-time polarimetric imaging, for example with a flow of at least 8 images per second of polarimetric images. Thus, when the practitioner changes the observed area, he quickly benefits from the corresponding polarimetric image. Increasing the image frequency makes it possible to artificially increase

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] The number of frames per second can be increased without changing the exposure time required for acquisition, and therefore offers the possibility of obtaining images in real time. The frequency of 8 frames per second is sufficient if the object being observed is static or moving slowly. If the object is moving more quickly, the required frame rate may be higher. The processor can be arranged to, before performing the permutation P, perform a spatial registration of the images corresponding to the measurements in order to take into account possible movements of the areas observed in the field of vision of the camera. This limits the risk of motion blur in polarimetric images and increases the precision of the polarimetric information. The processor can be arranged to perform an additive decomposition of the Mueller matrix M in the form M = qMnd+pMd where Mnd is the non-depolarizing component and Md is the depolarizing component. The parameters q and p are the weights of the two components of the Mueller matrix. In particular Mnd is the Mueller matrix of a given linear phase retarder by : Mnd=Mje,ô) = laugh o cos223 + sin220cosô sin23cos23( 1 - cosô) sm2Bcos23( 1 - cosô) sin223 + cos223cos 0 - sin23sinô cos23smô cosô .0 sin23sinô - cos23sinÔ and Md is the Mueller matrix of a pure depolarizer given by: '1 0 0 .0 0 0 0' 0 0 0 0 0 0 0 0 0. 6 is the azimuthal orientation of the fast (or slow) axis in degrees and <5 the retardance of linear phase in degrees (ranging from 0° to 180°) with 3- |atan(^) - dsr = |atan(^|f ) \. kx? / —~-—r - atan( ) sin(2fl)^ / V ««5 / d = ^-l = ^-l=l + dl M the coefficients mÿ (i,j=1,2,3,4) being those of the Mueller matrix M, mu being the unnormalized and unpolarized intensity coefficient of M, the other coefficients m^ (i,j = 1,2,3,4) being normalized with respect to mu. The processor may be arranged to calculate the depolarization by performing the following operation:

[0091] Depolarization = 1 - ™

[0092] The system can be arranged to generate the parallel display of at least one non-polarimetric image in at least one of the spectral bands, in particular the color image of the observed area, and of at least one polarimetric image.

[0093] The invention also relates to a method for learning an artificial intelligence system, for example comprising at least one convolutional neural network, in which the system receives non-polarimetric images and polarimetric images as input.

[0094] The fact that these images come from a multi-sensor camera avoids spatial image registration operations, and reduces the calculation time, which leaves more resources available for learning as such. Increased frame rate

[0095] According to another of its aspects, preferably taken in combination with the above, but which can be applied generally to other types of imaging, the invention relates to a real-time imaging method, comprising the steps consisting of: - Acquiring, per given time interval, original real series 4 of n measurements n— .. j succeeding each other in a predefined order within the series, each series of n measurements allowing by processing these measurements to generate at least one original result X= G6uy at a given frequency f, G being a function giving the result from the measurements, - generate artificial series from a grouping of n measurements from a real series M of a given rank, and from at least one real series of a different rank, in particular the following rank, and from a permutation of the measurements within each artificial series so as to respect the predefined order of the n measurements within each series, - generate a stream of results at a frequency greater than f from the original results and the results generated from the artificial series interspersed between the original real series.

[0096] This imaging method is advantageously applied to polarimetric colposcopy, and better, to multispectral polarimetric colposcopy as defined above.

[0097] The result is then a Mueller matrix or a physical property calculated from this matrix. The series of measurements are then polarimetric measurements.

[0098] The invention makes it possible, according to this aspect, to artificially increase the flow of images and to enable, or improve, real-time imaging despite the number of measurements to be carried out and the calculations necessary to generate the desired images.

[0099] The invention, according to this aspect, makes it possible in other words to artificially increase the number of images per second, while maintaining a fixed number of acquisitions.

[0100] By "real time" is meant a delay between acquisition and visualization which is relatively small, for example 2 or 3 seconds, and compatible with visualization of the information generated by the practitioner during the examination; for example, the delay between the time when the measurements are taken and the time when the information resulting from these measurements is displayed is less than or equal to 2s, better than 1s. The flow of results generated can be greater than 8 per second.

[0101] The imaging to which this method is applied is advantageously polarimetric imaging and the series of n measurements can then correspond to the measurements of the coefficients of the intensity matrix B used to calculate the Mueller matrix.

[0102] In particular, the series of measurements can correspond to the acquisition of at least 9, in particular 12, better 16, coefficients of the intensity matrix B.

[0103] As indicated above, for a 4*4 Mueller matrix, with

[0104] B=AMW,

[0105] where A= [Sap SA,. SA;, SaJT

[0106] andW=[Sw Sw,SWîSwJ

[0107] the measurements are obtained using a PSG polarization state generator which produces four independent states described by four Stokes vectors Sw (i = 1,2, 3,4) and a polarization state analyzer PSA which generates analysis states described by four Stokes vectors Sa (i = 1,2, 3,4).

[0108] Preferably, as mentioned above, before performing the permutation, a spatial registration of the images corresponding to the measurements is carried out in order to take into account possible displacements of the areas observed in the field of vision of the camera. Post-processing of the Mueller matrix

[0109] To accelerate the calculation of the Mueller matrix, one can implement, for example, a particular post-processing which avoids proceeding via a so-called Lu-Chipman decomposition but nevertheless makes it possible to obtain a similar result.

[0110] The invention thus also relates, according to another of its aspects, independently or in combination with the above, to a polarimetric imaging method in which intensity images are acquired using a polarimetric imaging system (preferably the one defined above) comprising an illumination system, a polarization state generator (PSG) placed on the light path between the illumination system and the area to be observed, a polarization state analyzer (PSA) placed on the light path between the area to be observed and at least one image acquisition system, then a decomposition is carried out additive of the Mueller matrix M in the form M = qMnd+pMd where Mnd is the non-depolarizing component and Md is the depolarizing component. In particular M nd is the Mueller matrix of a linear phase retarder given by: [YES]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125] '10 0 0 . . 0 cos228+sin^Ocosô 5zn20ct«20(l-cosÔ) -sin2QsmÔ ïnd=M (e,ô) = >and 0 sin28cos28(l-cosS) sin226 + cos226cos cos20sin5 .0 sin28sm8 -cos28sinô cosÔ and Md is the Mueller matrix of a pure depolarizer given by: '1 0 0 01 0 0 0 0 8 is the azimuthal orientation of the fast axis in degrees and 5 is the phase of the retardance in degrees (ranging from 0 to 180°) with ( -xmlVsï / iô \ \ W / (x / siït2&xmd X -:-¾— ) = atan = atan( -¾ ) sm(20)m44 / \ sin(.2(9)^ / \ cosà / M The parameters q and p are the weights of the two components of the Mueller matrix. The coefficients (i,j=1,2,3,4) are those of the Mueller matrix M, mu being the unnormalized and unpolarized intensity coefficient of M, the other coefficients (i,j = 1,2,3,4) being normalized with respect to mu. Such an additive decomposition of the Mueller matrix allows to parallelize the calculations and to extract useful parameters more quickly. Such a method avoids calculating the eigenvalues ​​of the Mueller matrix, which is computationally expensive, and allows extracting the desired parameters with much shorter calculations. This method is based on the assumption that in some tissues, such as the cervix, the observed effects are mainly depolarization and birefringence effects, and assumes that birefringence effects are related to the tissue surface area while depolarization effects are related to the tissue volume. Dual-port observation system equipped with a polarimetry system According to another of its aspects, independently or in combination with the above, the invention also relates to an observation system with two input ports, in particular a binocular system, for example for colposcopy, comprising a head comprising an optical system having a light output port for illuminating an area to be examined and left and right input ports directed towards the area to be observed, the observation system further comprising a polarimetric system comprising a polarization state generator arranged in front of the output port and a polarization state analyzer of which at least part of the optical elements is arranged in front of one of the input ports, this analyzer comprising at least one optical element held by at least one support applied to only part of the outline of the optical element.

[0126] This can make it possible to provide a free edge for the optical element, this free edge partially overlapping the other input port. The optical element can only partially overlap this other input port, without having this input port hidden by the support which would hinder observation. The optical element is for example a liquid crystal polarization modulator.

[0127] This aspect of the invention takes advantage of the fact that the free edge of the optical element, which is preferably of circular outline, although partially superimposed on the input port, remains substantially transparent in the eyepiece associated with this input port due to the focusing distance, which is typically of the order of 20 to 40 cm, therefore much greater than the distance separating the input port from the optical element, which is at most a few cm. In addition, since the optical elements of the analyzer are relatively transparent, except possibly for the linear polarizer which can be placed in front of the camera downstream of the prism for reflecting towards the eyepiece, these elements do not cause any significant loss of brightness for the observer.

[0128] It is thus possible to use a conventional colposcope head or a head of another observation system with two input ports, in particular a binocular observation system, by attaching all or part of the polarimetric system to it on the front face, which limits the manufacturing costs of the system and makes it possible to easily add a polarimetric imaging function to the observation system.

[0129] Each support of an optical element of the analyzer can extend over an angular extent of between 180° and 300° in contact with the optical element, preferably having a general shape of a C open in a downward oblique direction, substantially at 45°.

[0130] The polarimetric system may further comprise at least one other support for holding at least one optical element of the polarization state generator placed in front of the light output port of the colposcope head.

[0131] This other support may have a general shape of an upwardly open C, which limits the size of the state generator in the vertical direction and makes it possible not to encroach on the input ports present near the output port on the front face of the head of the colposcope.

[0132] Each support may comprise a succession of support pieces holding optical elements of the polarization state generator (or of the analyzer) between them, and two supports serving to hold two consecutive optical elements may share an intermediate support piece.

[0133] The polarimetric system can thus comprise at least two consecutive support pieces assembled against each other, each of these support pieces having on one face a housing for receiving a respective optical element and one of the support pieces serves to retain the optical element of the other piece. It is thus possible to have a relatively compact polarimetric system, which does not unduly hinder the manipulation of the head of the observation system by the practitioner.

[0134] The polarization state generator and the analyzer preferably each comprise similar optical elements which are placed in reverse orders relative to the direction of propagation of the light.

[0135] The generator thus comprises, for example, in the direction of propagation of the light, a linear polarizing filter, a quarter-wave liquid crystal polarization modulator QFLC, a half-wave plate QWP, and a half-wave liquid crystal polarization modulator HFLC.

[0136] The analyzer can then comprise, in the direction of propagation of the light, a half-wave liquid crystal polarization modulator HFLC, a half-wave plate QWP, and a quarter-wave liquid crystal polarization modulator QFLC.

[0137] The polarizer associated with the analyzer is preferably arranged after returning the light from the input port to the associated eyepiece, upstream of the camera used to collect the intensity images used to generate the Mueller matrix, which makes it possible to limit the difference in brightness between the images delivered to the observer by the left and right eyepieces, as mentioned above.

[0138] The housings receiving the electrically controllable optical elements may comprise at least one passage for an electrical cable. This passage may comprise a channel which matches the shape of the contour of the element, in particular a semi-circular channel.

[0139] The C-shaped supports of the optical elements allow, before tightening, manual rotation of the optical elements around their axis, in order to carry out orientation adjustments.

[0140] The optical elements of the generator are preferably oriented perpendicular to the direction of propagation of the light exiting the head of the observation system.

[0141] When this light is emitted by a prism in a direction making an angle with the direction of observation of the input ports, the mounting of the optical elements of the generator is preferably carried out at an angle relative to those of the analyzer, so as to orient them perpendicular to the axis of propagation of the light. leaving the output port. This helps limit stray reflections.

[0142] Preferably, the polarimetric system comprises a fan configured to blow air onto the optical element of the polarization state generator closest to the light output port of the head of the observation system. This ensures better temperature stabilization, and limits the corresponding drifts.

[0143] The polarimetric system may comprise a movable filter holder in front of the analyzer, allowing the system to be calibrated. This filter holder is for example fixed on one of the support parts used to hold the optical elements. The filter holder may be movable in a horizontal direction, generally perpendicular to the observation direction. The filter holder may for example take four positions, three of which are for placing a predefined optical element in front of the analyzer, and the last one to free the field. The movement of the filter holder is preferably motorized, and controlled by the aforementioned controller.

[0144] The system may include a housing protecting the supports and the filter holder.

[0145] The invention also relates to a polarimetry system intended to be attached to a head of an observation system, in particular a head of a colposcope, comprising an optical system having a light output port making it possible to illuminate an area to be examined and left and right input ports directed towards the area to be observed, the polarimetric system comprising a polarization state generator placed in front of the output port and a polarization state analyzer of which at least some of the optical elements are placed in front of one of the input ports, this analyzer comprising at least one optical element held by at least one support applied to only part of the contour of the optical element.

[0146] This may make it possible to provide a free edge for this optical element, this free edge being arranged so as to partially overlap the other input port.

[0147] The polarimetry system may comprise means of attachment to the head of the observation system.

[0148] These fixing means may comprise screws, which are screwed into the frame of the head, for example under the exit window. Alternatively, the fixing is carried out differently, for example by tightening.

[0149] The polarimetry system may have all or part of the characteristics given above. Compact and modular polarimetry system

[0150] The invention also relates to a polarimetry system intended to be attached to a head of an observation system, in particular a colposcope head, comprising an optical system having a light output port making it possible to illuminate an area to be examined and at least one input port, the polarimetric system comprising a polarization state generator (PSG) placed in front of the output port and a polarization state analyzer (PSA) of which at least some of the optical elements are placed in front of the input port, at least one of the analyzer and the generator comprising optical elements held by at least one support, each support comprising a succession of parts holding between them optical elements of the polarization state generator or of the analyzer, at least two supports serving to hold two consecutive optical elements sharing an intermediate support part, each of these parts preferably having on one face a housing for receiving a respective optical element and one of the parts serving to retain the optical element of the other part.

[0151] Such a polarimetry system is particularly compact, since it comprises support parts which serve as a receiving housing for an optical component and as a closing cover for the next support part.

[0152] Such a system is also modular, since it is easy to replace support parts with others.

[0153] It may have any of the characteristics of the observation system, in particular of the colposcope, described elsewhere. Colposcope illumination system

[0154] According to another of its aspects, independently or in combination with the above, the invention also relates to a colposcopy system comprising: - an illumination system comprising • at least one light source, • a liquid light guide, to be connected at one end to the light source light, this light guide having a core diameter less than or equal to 5mm, - a colposcope head, comprising an optical system having an input port to be connected to the other end of the liquid light guide, and a light output port for illuminating an area to be observed, the optical system preferably comprising between the input port and the output port an aspherical lens and a deflecting prism.

[0155] This configuration of the illumination system makes it possible to obtain a well-collimated light beam at a working distance of approximately 30 cm.

[0156] The choice for the light guide of a liquid guide and for the core diameter of a diameter less than or equal to 5 mm, makes it possible to reduce the divergence of the illumination beam and its size, and to increase the intensity of the light at the center of the area to be observed, that is to say the cervix, and thus to reduce the parasitic reflections on the surrounding surfaces (vaginal wall, speculum, etc.), and consequently to significantly improve the quality of the acquired images. The light guide with liquid allows illumination to be obtained with satisfactory uniformity, notably a uniformity superior to that obtained with the silica fiber bundles generally used in colposcope illumination systems.

[0157] Preferably, the inner diameter of the liquid optical guide is between 2.5 and 3.5 mm.

[0158] The illumination system may include an adapter configured to receive the optical guide and mount on the source with adjustment in the three directions X, Y and Z relative to the source. Ergonomic colposcopy system

[0159] According to another of its aspects, independently or in combination with the above, the invention also relates to a colposcopy system comprising: - a rolling base, - a mast carried by the rolling base, - at least one reference reflector carried by the mast, used for calibrating a polarimetry system, - a colposcope head carried by an articulated arm connected to the base.

[0160] The colposcopy system may also comprise: - a workstation comprising a computer, carried by the rolling base, - a keyboard carried by an articulated arm connected to the mast, and - a screen carried by an articulated arm connected to the mast at a height greater than that of the arm supporting the keyboard.

[0161] Such an arrangement makes it possible to keep the screen and keyboard close to the user, who can thus easily control the computer himself for image acquisition.

[0162] The mobility of the system is also improved, since the whole unit can be easily moved in one piece on the ground.

[0163] Preferably, the system comprises a housing housing the light source(s), placed on the workstation, as well as a controller which makes it possible to control the various electro-optical and electronic components of the polarimetric system such as the controllable liquid crystal filters, the filter wheel, and to process the images coming from the camera, if applicable, or even to generate the polarimetric images, etc.

[0164] The reference reflector may be arranged in a light shield at the top of the mast. This reflector is, for example, articulated around a vertical axis, and comprises, for example, on one face a frosted metal wall, for example made of aluminum, and on the opposite side a reference surface having known spectral properties. The rotating mounting of the reflector makes it possible to easily replace one of the faces with the other without moving the head of the colposcope, which facilitates calibration operations. When these, the head is for example placed approximately 30cm from the reflector, by maneuvering the articulated arm carrying the head.

[0165] The colposcopy system advantageously comprises a pedal which allows the user to trigger a predefined action, for example to start a polarimetric acquisition.

[0166] The colposcope head may also include a button to trigger another predefined action, for example starting and stopping video recording. Description of the figures

[0167] The invention may be better understood by reading the detailed description which follows, a non-limiting example of implementation of the different aspects of the invention, and by examining the attached drawing, in which:

[0168] [Fig-1] [Fig.l] is a block diagram of an exemplary multi-imaging system polarimetric pectral according to the invention,

[0169] [Fig.2] [Fig.2] is a view similar to [Fig.l] illustrating the possibility of delivering an image facilitating diagnosis,

[0170] [Fig.3] [Fig.3] illustrates the possibility of using artificial intelligence for generate diagnostic assistance,

[0171] [Fig.4] [Fig.4] illustrates different modes of displaying information in a colposcopy system according to the invention,

[0172] [Fig.5] [Fig.5] is a diagram illustrating the acquisition of successive series of measures,

[0173] [Fig.6] [Fig.6] illustrates the generation of artificial series from the series of the example of [Fig.5],

[0174] [Fig.7] [Fig.7] illustrates the reordering of measurements within the arti series functionals of [Fig.6],

[0175] [Fig.8] [Fig.8] illustrates the spatial registration of areas of the image corresponding to successive measurements,

[0176] [Fig.9] [Fig.9] illustrates the application of the method to measurements made in the framework of Mueller polarimetry,

[0177] [Fig. 10] [Fig. 10] illustrates the generation of artificial series of measurements in the case of the example of [Fig.9],

[0178] [Fig. 11] [Fig. 11] illustrates the assumptions used for the accelerated calculation of the parameters of interest,

[0179] [Fig. 12] [Fig. 12] represents comparative images obtained by implementing the Lu-Chipman decomposition on the one hand, and the additive decomposition used for the accelerated calculation of the polarimetric parameters on the other hand,

[0180] [Fig. 13] [Fig. 13] is a schematic and partial view of a colposcopy system according to the invention,

[0181] [Fig.14] [Fig.14] represents more particularly the head of the colposcope and the polarimetry and acquisition systems,

[0182] [Fig. 15] [Fig. 15] represents more particularly the lower part of the colposcopy system,

[0183] [Fig. 16] [Fig. 16] illustrates a detail of the system at the level of the head of the colposcope,

[0184] [Fig. 17] [Fig. 17] is a top view of the colposcopy system, arranged at proximity to an examination table,

[0185] [Fig. 18] [Fig. 18] is a partial elevational view of the colposcopy system,

[0186] [Fig. 19] [Fig. 19] represents the head of the colposcope with part of the system of polarimetry which is fixed on its front face,

[0187] [Fig.20] [Fig.20] is an exploded view showing different optical elements of the polarization state generator and the analyzer and their supporting parts,

[0188] [Fig.21] [Fig.21] illustrates the assembly of the parts supporting the optical elements of the polarization state generator,

[0189] [Fig.22] [Fig.22] illustrates the assembly of parts supporting optical elements of the polarization state analyzer,

[0190] [Fig.23] [Fig.23] represents in isolation one of the support parts, used for the fixing the filter holder used for calibration,

[0191] [Fig.24] [Fig.24] illustrates the assembly of two generator support parts of polarization states,

[0192] [Fig.25] [Fig.25] shows a front view of one of the support parts of the ge generator and the optical element received in the housing of this part,

[0193] [Fig.26] [Fig.26] is a frontal view of the front face of the colposcope head equipped with the polarimetric system,

[0194] [Fig.27] [Fig.27] is a partial and schematic sectional view of the system of illumination,

[0195] [Fig.28] [Fig.28] is a schematic and partial view, in perspective with section axial, of the illumination system of [Fig.27],

[0196] [Fig.29] [Fig.29] is a partial and schematic longitudinal section of the head of the colposcope at the level of the light guide connection,

[0197] [Fig.30] [Fig.30] is an exploded perspective view of elements used for assembly from the light guide to the head of the colposcope,

[0198] [Fig.31] [Fig.31] illustrates the spatial distribution of light intensity at the level of the observed area as a function of the diameter of the active part of the light guide used,

[0199] [Fig.32] [Fig.32] represents the calibration reflector and its cap,

[0200] [Fig.33] [Fig.33] represents the reflector in isolation, without its cap,

[0201] [Fig.34] [Fig.34] schematically and partially represents the filter holder calibration,

[0202] [Fig.35] [Fig.35] represents the part of the polarimetry system which is fixed on the front face of the head of the colposcope, equipped with its protective cover,

[0203] [Fig.36] [Fig.36] is a schematic and partial side view illustrating the mounting of the acquisition system on the head of the colposcope,

[0204] [Fig.37] [Fig.37] represents the filter wheel of the acquisition system and its drive motor,

[0205] [Fig.38] [Fig.38] represents in a partial and schematic manner the acquisition system from another angle of view, and

[0206] [Fig.39] [Fig.39] schematically and partially represents an example of a multi-sensor camera optical system. Detailed description Multispectral polarimetric imaging system

[0207] [Fig.l] shows a multispectral polarimetric imaging system 1 according to the invention, for example a Mueller polarimetric colposcope, it being understood that this aspect of the invention is not limited to a colposcope.

[0208] The system 1 comprises an optical illumination system 2 for illuminating a target T, for example the uterine cervix, and an optical system 3 for direct observation of the target and / or reconstruction of the image of the target T on an electronic medium.

[0209] The illumination and observation of the target T are carried out through a polarimetry system 4 which comprises a polarization state generator (PSG) crossed by the light coming from the illumination system and a polarization state analyzer (PSA) crossed by the light coming from the target T.

[0210] Preferably, the PSG and the PSA comprise liquid crystal polarization modulators, in particular chosen from those whose control frequency can be at least 60 Hz. Ferroelectric liquid crystal polarization modulators are preferred.

[0211] The system 1 also comprises an acquisition system comprising a multi-sensor camera 5, constituted for example by a tri-CCD or tri-CMOS camera, having for example sensors dedicated respectively to the red, green and blue domain of the visible spectrum.

[0212] The system 1 comprises computer means 6 for, among other things, managing the acquisition of intensity images for each sensor, controlling the operation of the liquid crystal polarization modulators, synchronizing the acquisitions of the camera with the modulation of the liquid crystals, processing the images obtained by each camera 5 sensors and calibrate the system for each of the chosen spectral bands, as will be detailed later.

[0213] These computing means 6 comprise for example one or more processors, one or more microcontrollers, specialized circuits such as FPGAs, or microcomputers, and the associated human-machine hardware interfaces, and can be programmed to carry out the acquisition of intensity images in the different spectral bands of interest, for example red, green and blue in the example considered, then the calculation, for all or part of the pixels of the images, of the Mueller matrices in these different spectral bands, and the calculation and display of the polarimetric parameters of interest.

[0214] The computer means 6 can thus comprise a controller making it possible to control the PSG and the PSA and to process the images coming from the camera, said controller comprising for example one or more FPGAs, and a workstation comprising a computer equipped where appropriate with at least one graphics card, and a human-machine interface comprising for example a screen, a keyboard, and one or more control buttons or pedals, as described below.

[0215] The illumination system is capable of emitting in each of the observation spectral bands.

[0216] In the case where the spectral bands are in the red, green and blue respectively, the illumination system may comprise a white light source such as a xenon source associated with spectral filters, or a set of LED diodes emitting respectively in the red, green and blue, or LED diodes emitting in the blue and yellow.

[0217] The use of a xenon lamp may be preferred for its ease of use and its power.

[0218] The camera of the acquisition system being multi-sensor, preferably a tri-CCD or tri-CMOS RGB camera, it can acquire several images at the same time without significant loss of light and without interference problems (“cross talking”) between the channels.

[0219] The image obtained on each of the sensors can be an intensity image in gray levels for the corresponding spectral band.

[0220] The grayscale intensity images of the different sensors are perfectly superimposable by construction, which simplifies the reconstruction of the final images after processing, in particular the addition of polarimetric information to the color images or the combination of polarimetric images at different wavelengths.

[0221] Images obtained by the multispectral polarimetric imaging system, applied to Mueller polarimetry in the different spectral bands, notably in red R, green G and blue B, can be generated by calculation in the different spectral bands, as shown in [Fig. 1]. These images can, if necessary, be combined to generate a Mueller polarimetry RGB color image.

[0222] In parallel, the intensity images acquired by the three sensors of the camera can be combined to form an RGB color image. This image is useful to allow the practitioner to select the area where the polarimetric analysis is to be carried out.

[0223] The determination of the Mueller matrices for each of the spectral bands R, G and B makes it possible to calculate in each of these bands intensity images translating the value of a polarimetry parameter (for example the retardance) for each of the pixels of the image, as illustrated in [Fig.2].

[0224] Thus, for example, for each spectral band, a depolarization image is generated, determined from knowledge of the Mueller matrix, as well as a retardance image, also determined from knowledge of the Mueller matrix.

[0225] This or these retardance and / or depolarization images can be merged with the color image to generate an image I in which the hue and / or contrast of certain areas is modified in order to provide additional information to the practitioner, helping with diagnosis.

[0226] The combination within the same imaging system of a polarimeter with liquid crystal polarization modulators, a multi-spectral illumination source, in particular a xenon lamp, and a tri-CCD or tri-CMOS camera makes it possible to obtain a particularly compact and high-performance imaging system, well suited to colposcopy in particular.

[0227] Being able to generate multi-mode images that are perfectly spatially aligned with each other facilitates deep learning, for example by means of a convolutional neural network 7, as illustrated in [Fig.3], this network receiving as input conventional color images, monochromatic intensity images and polarimetric images for the selected wavelength ranges and delivering as output one or more simplified images to aid diagnosis, for example in the form of increased contrast.

[0228] It is possible in particular during learning, which may be supervised learning, to provide the artificial intelligence system with non-polarimetric RGB color images as well as polarimetric images representative for example of images relating to depolarization, linear phase retardation and azimuth of linear phase retardation etc., the artificial intelligence system producing one or more images containing information helping with diagnosis.

[0229] The computing means 6 can be configured to allow all or part

[0230]

[0231]

[0232]

[0233]

[0234]

[0235] of the following visualizations: - a parallel multimode visualization (figure 4A) of different polarimetric and non-polarimetric images, with for example display on the same screen of the RGB color image obtained by the camera (on the left), of an image of the linear phase retardance (in the middle), and of an image of the azimuth of the slow axis (on the right) - a multispectral parallel visualization (figure 4B), with for example on the left the reference RGB color image and on the right three series of three images in each of the spectral bands of interest, the first line representing the linear phase retardation, the second line representing the azimuth of the slow axis of the linear phase retardation and the third line representing the depolarization; - an interactive visualization (figure 4C) in which the overall RGB color image is displayed but a restricted display area, in this case circular in shape, for example centered on a pointer which can be moved on the image by the user, displays the information relating to a polarimetric parameter, in this case the linear phase delay, partially superimposed with the displayed RGB image, - a visualization (Figure 4D) with display in the left column of the RGB color image (top), the color polarimetric image (middle), and the image of the azimuth of the slow axis of the linear phase retardation at the bottom, as well as the enlarged image (right) of the three images of the left column merged together. We can still display: - color images of Mueller matrices, - color images of depolarization, - color images of linear phase retardation, - color images of azimuth of linear phase retardation, etc. Such an imaging system is advantageously applied to colposcopy, as will be detailed later, but can also be applied to other types of imaging systems, for example endoscopy, as well as microscopy and exoscopy for neurosurgery. Reminder on Mueller polarimetry Generally speaking, the polarized state of the incident light is related to the light exciting a sample S()Ut by the relation: ^out ~ M*Sjn with M the Mueller matrix. In order to obtain a Mueller matrix M of dimensions 4x4, the measure of coef- intensity coefficients is required. These intensity coefficients can be obtained through the use of the polarization state generator (PSG) which produces four independent polarization states characterized by four Stokes vectors (j — ] ,2, 3,4)-

[0236] Each represents a column of the modulation matrix W:

[0237] W = [Swr Swÿ Swÿ Sw4]

[0238] After interaction with the observed area, each polarization state produced by the polarization state generator is analyzed by four polarization configurations of the polarization state analyzer (PSA).

[0239] These four configurations are also described by four Stokes vectors (i = 1,2, 3,4)' 9ui represent the rows of the analysis matrix A: A=[SAt,SAÿSAs,SA4]T

[0240] After these steps we obtain the intensity matrix B:

[0241] B = AMW

[0242] from where:

[0243] M = AaBWa

[0244] B corresponds to a series of n measurements, in this case 16 measurements in the case of a 4x4 Mueller matrix.

[0245] The Mueller matrix can be processed according to various known methods in order to calculate polarimetric properties, namely depolarization, linear phase retardation, etc.

[0246] In the case of the polarimetric imaging system according to the invention, 16 successive measurements can be carried out for each pixel of the image, for each of the sensors of the multi-sensor camera, corresponding to the different coefficients of the intensity matrix B.

[0247] Preferably the Mueller matrix is ​​the classic 4x4 matrix, i.e. the complete Mueller matrix, but it is not outside the scope of the invention to use an incomplete 3x3 or 3x4 version. The invention can also be used for simplified polarimetric imaging techniques such as Stockes polarimetric imaging. Increased frame rate

[0248] According to one of its aspects, the invention makes it possible to increase a flow of results obtained for a physical quantity (or a set of physical quantities) X relating to a physical object O per unit of time from consecutive series p of n real measurements, called original series, succeeding one another in a predefined order at each given time interval, by generating additional series, called artificial series, of the original series.

[0249] These measurements can be the 16 intensity measurements of the matrix B in the case of the Mueller polarimetry, but this aspect of the invention is more general and can be applied to other types of imaging different from Mueller polarimetric imaging. Thus, X can be other than a Mueller matrix obtained by a Mueller polarimetry system, and in particular a non-square matrix of real numbers, a square matrix of real numbers, a vector or even a real number, or more generally any physical property or set of physical properties whose determination requires that several successive measurements be carried out. Otherwise, X can be obtained for an array of points (imaging). This array can be two-dimensional (2D) or three-dimensional (3D). This is the case for 2D or 3D imaging. For 2D imaging, each measurement point corresponds to a pixel. For 3D imaging, each measurement point corresponds to a voxel.For the result X=G(p) to be calculated correctly, the n measurements of the series p must respect a predefined order po, pB p2, ... . Pn-l*

[0250] Any other order, for example such that p' = (p2, p3, ..., p„ i, po), gives, in general, a value X'=G(p') where X' is not the correct result.

[0251] If the measuring device is capable of performing l times (IgN) all of the n measurements given by <. _ / H >> ] in 1 second, in all it provides a set p .... j of discrete measurements ni = (ni^ ..., n\_| ) in 1 second, where s=Zn.

[0252] These measured values ​​are for example the coefficients of the intensity matrix B in the case of Mueller polarimetry.

[0253] [Fig.5] represents as an example three real series p of measurements, each comprising four measured values ​​(n=4). We denote po the first value, pi the second, p2 the third and p3 the last, this order being predefined.

[0254] We first estimate that the result X (for example the Mueller matrix) does not vary or varies little over time. We can therefore assume that each series p has n measurement values ​​similar to those of the counterparts of the following series obtained in the same order.

[0255] New combinations of measurements can then be artificially formed to obtain new series of artificial measurements by associating measurements from different series, as illustrated in [Fig.6].

[0256] In the example of figure 6, we associate for example for the series p the values ​​p^ p2> p3 of the first series with the value po of the second series.

[0257] We repeat these groupings on all n measurements.

[0258] The series g , p and p are identical to the series p of [Fig.5]. The order of the measurements is preserved. For these series it is not necessary to swap the measurements.

[0259] For the series p , p , p^, p , p^ and p the predefined order of the measurements is not respected; we then apply a permutation function P, this permutation function being illustrated in [Fig.7], in order to respect the predefined order of measurements within each series.

[0260] Following the permutation of the series p , .. .in order to preserve the predefined order, the function G can be applied to the set of new permuted series P(û ), P(û 4 5 ), ..., to calculate X.

[0261] In the general case of n measurements, if Tn is the time required to carry out the n measurements, Tt the time required to make a measurement, we can carry out 1 / Tn series of n measurements in one second; Tp is the time required to carry out a permutation, and Tc is the calculation time required to calculate X=G(p), then we preferably have Tc«Tt and Tp «Tt.

[0262] As indicated previously, the measurements can be successive intensity values ​​of each pixel of an image recorded by a given sensor of the camera of the acquisition system, and the method which has just been described can be applied to each pixel of this image.

[0263] However, this image can change over time, for example due to movements of the target.

[0264] In [Fig.8] (left) we have considered again the simplified case presented previously with n=4. Each small square represents a part of the image in the total field of vision represented by the large square. This part of the image can move in the field of vision over time.

[0265] It is then useful to apply a spatial registration function R to register the images acquired at different successive times, as illustrated (right), so that the successive intensity measurements used in the calculation of the Mueller matrix concern the same area.

[0266] The registration function R can be applied immediately after the acquisition of a new image, for example from a series of measurements of the image corresponding to m4 in [Fig.8].

[0267] The recalibration makes it possible to spatially coincide the pixels corresponding to the mB values ​​m2 and m3 with that corresponding to m4, those corresponding to the values ​​m2, m3 and m4 with that corresponding to m5jthose corresponding to the values ​​m3, m4 and m5 with that corresponding to m6jetc., as illustrated in [Fig.8].

[0268] The image registration function can involve any suitable image registration algorithm; an example of a registration function is described in the article M. Irani and S.Peleg “Improving Resolution by Image Registration”, 1991, CVGIP Graphical models and image processing, Elsevier.

[0269] The recalibration function may seek to determine the recalibration values ​​a in x and b between two images, so as to minimize a loss function between them.

[0270] An example of a loss function is given in the article MBA Haghighat, A Agha-golzadeth, and H. Seyedarabi, “A non-reference image fusion metric based on mutual information of image features,” Computers & Electrical Engineering, vol. 37, no. 5, pp. 744-756, Sep. 2011, doi:10.1016 / j.compeleceng.2011.07.012.

[0271] The permutation function P can be applied after the recalibration function R, to put the measurements in the correct order, then the function G can be applied to calculate the result X.

[0272] If we want to have more than eight times the result X calculated per second, in order to have a certain fluidity corresponding to imaging substantially in real time, we need 1 / Tn >=8.

[0273] The procedure described above can also be used in a more general case where the object O can be deformed. If Td is the characteristic time of the deformation, it is preferably Td»Tn.

[0274] If Tx is the characteristic duration of a change in the property materialized by the result X, it is also preferable to have Tx»Tn.

[0275] In the case of Mueller polarimetric imaging, the measured values ​​are the values ​​of the coefficients Bÿ (i,j = 1,...,4) of the intensity matrix B.

[0276] The sixteen coefficients of this matrix correspond to the measurements made for a given pixel, in order to then be able to calculate the Mueller matrix for this pixel, and form a series p of measurements as illustrated in [Fig.9].

[0277] Figure 10 illustrates the application of the permutation function P(p) to the series p.

[0278] The function G here corresponds to the calculation which makes it possible to obtain from the intensity matrix B the Mueller matrix M.

[0279] M = G ( p ) = X = A^BW*

[0280] The registration function R is applied before applying the permutation function P; the registration function R can be applied by taking as reference image, for example, that corresponding to the first image of a sequence of 16 consecutive images (in the case of a 4x4 Mueller matrix).

[0281] Preferably, TP + TR + Tc < Tt, in order to have real-time imaging, with

[0282] TP the time required for the permutation P,

[0283] Tr the time required to apply the recalibration function R,

[0284] Tc the time required to calculate M,

[0285] All the duration of a single measure. Post-processing of Mueller matrices

[0286] The conventional calculation of the Mueller Matrix can involve, for example, a so-called Lu-Chipman decomposition, as described in the article S.-Y. Lu and RA Chipman, “Interpretation of Mueller matrices based on polar decomposition”, Journal of the Optical Society of America A, vol 13, no.5, pll06, May 1996, doi: 10.1364 / JOSAA. 13.001106.

[0287] However, this decomposition is relatively expensive in terms of calculation time because it involves the calculation of the eigenvalues ​​of the matrix M.

[0288] Furthermore, when observing the uterine cervix, we mainly observe birefringence and depolarization effects.

[0289] It is assumed here that the birefringence effects are associated with the surface of the observed area while the depolarization effects are related to the volume of the observed area, as illustrated in [Fig.l 1].

[0290] This figure schematically illustrates the propagation of light within the tissue. In this case, the Stokes vector Snd is associated with the non-depolarized component coming from the surface layer of the tissue, while the Stokes vector Sd is associated with the depolarized component of the light, coming from the volume of the tissue.

[0291] [Fig. 11] also represents the two Mueller matrices Mndet Md respectively related to the non-depolarized and depolarized components of light.

[0292] The factors q and p are intended to estimate the proportions of undepolarized and depolarized light, and must be determined.

[0293] The intensity matrix B can be formulated as the sum of the contributions from undepolarized and polarized light.

[0294] We then have:

[0295] B = B„„ + B„ = AÇqM^+pM^ W

[0296] and

[0297] M = qMnd + pMd = A, (Bml+Bd)W“l

[0298] The Mueller matrix of the undepolarized light Mnd, corresponds to a linear delay Mueller matrix, such as:

[0299] o ; Azimuthal orientation of the fast axis in degrees

[0300] 5: phase delay in degrees

[0301] 5e

[0180] eX.

[0302] 1 0 0 0 . . 0 cos22B + sit^Bcosd sin2Bcos2B{\-CQsd') -sin2Bsind Mret(e,ô) = 0 sin28cos2B(\-cos5) sùr26 + cos22(cos eos2Bsin5 .0 sinlBsind -cos2(kmd cosô

[0303]

[0304] The Mueller matrix of the depolarized light Md is equal to: 0 .0 0 0 0' 0 0 0 0 0 0 0 0 0.

[0305] On a

[0306] is ( m">A \ 1 / ^ / 115 \ 1 / sittW \ 0 = ^tan^ ) = ±tan\ \ l-^ci /

[0307] „ / \ / \ / ünô x J=atim^

[0308] 3+d'

[0309] The depolarization is given by:

[0310] Depolarization = 1 -

[0311] This decomposition method presents a significant time saving since it does not require the calculation of eigenvalues ​​and eigenvectors, which is time-consuming.

[0312] In addition, this method is matrix-based, which allows an algorithm to perform multiple calculations in parallel and not in series, which is a significant saving in time and memory.

[0313] The various polarimetric parameters can thus be obtained more quickly, which is very advantageous in the case of real-time polarimetric imaging.

[0314] In [Fig. 12] polarimetric images obtained on one side via a classical Lu-Chipman decomposition, and on the other via the additive decomposition described above, were compared.

[0315] The three top images are obtained by additive decomposition as described above and correspond respectively to the display of delta retardance (left image), theta retardance azimuth (middle image) and depolarization (right image), while the three bottom images represent the same parameters obtained via Lu-Chipman decomposition.

[0316] There is a great similarity between the images representing the same parameters, which reflects the performance of the process. Colposcopy system

[0317] The imaging system according to the invention can be arranged to carry out a colposcopy examination, for example using a colposcopy system 10 as illustrated in [Fig. 13],

[0318] This system 10 comprises a colposcope 20 having a head 21 carried by an articulated arm 22, itself carried by a rolling base 31.

[0319] A vertical mast 30 is also carried by the rolling base 31.

[0320] The colposcope 20 is connected to an illumination system comprising a light source 50 housed in a housing connected to the head 21 by a flexible optical guide 51, not visible in [Fig. 13] but apparent in [Fig. 14], the housing housing the source 50 resting for example on a workstation 60 comprising a computer.

[0321] The colposcopy system 10 comprises a screen 40 connected to the computer 60, which is advantageously a touch screen, preferably liquid-proof.

[0322] This screen is carried by the mast 30, as well as a keyboard 41, connected to the computer 60.

[0323] A controller 70 is connected to the computer 60 to control in particular the operation of a polarimetric system 90 carried by the head 21. This controller 70 can also be connected to an acquisition system 110, in particular to the camera 5.

[0324] Preferably, as seen in Figures 32 and 33 more particularly, the colposcopy system 10 comprises at least one reflector 81 housed in a support carried by the mast 30 and used for calibrating the polarimetry system. It is covered by a light shield 80, preferably carried by the mast 30, in particular at its top.

[0325] This reflector 81 is for example articulated around a vertical axis of rotation, and comprises for example on one face a frosted metal wall 82, for example made of aluminum, and on the opposite side a reference surface 83 having known spectral properties, for example a substrate with a neutral color surface with high Lambertian reflection called “spectralon”.

[0326] The rotary mounting of the reflector 81 makes it possible to easily replace one of the faces with the other without moving the head 21 of the colposcope, which facilitates calibration operations. During this, the head 21 is for example placed approximately 30 cm from the reflector 81, by maneuvering the articulated arm carrying the head.

[0327] The colposcopy system 10 advantageously comprises a pedal 100 visible in [Fig. 13] which allows the user to trigger a predefined action, for example to start a polarimetric acquisition.

[0328] The head 21 of the colposcope may also include a button 101, as illustrated in [Fig. 16], for triggering another predefined action, for example starting and stopping video recording.

[0329] The polarimetric system 90, illustrated in [Fig.19], comprises the polarization state generator PSG and the polarization state analyzer PSA, almost all of the optical elements of which are illustrated in [Fig.20].

[0330] The PSG polarization state generator comprises a succession of four optical elements placed on the path of the light exiting through a window 210 of the head 21 of the colposcope, namely in the direction going from the outside towards the exit window 210, a first 510nm half-wave liquid crystal polarization modulator HFLC 191, a 633nm quarter-wave plate QWP 192, a second 510nm quarter-wave liquid crystal polarization modulator QFLC 193, and a fixed linear polarizer 194.

[0331] The PSA polarization state analyzer comprises a succession of three optical elements placed on the path of the light reaching one of the input ports 211a and 211b associated respectively with the left and right eyepieces, in this case the input port 211b associated with the right eyepiece for the practitioner.

[0332] Starting from the outside and going towards the head of the colposcope, there are successively a first half-wave liquid crystal polarization modulator at 510 nm HFLC 221, a quarter-wave plate at 633 nm QWP 222 and a second quarter-wave liquid crystal polarization modulator at 510 nm QFLC 223.

[0333] The analyzer also includes a linear polarizer filter 224 similar to the polarizer 194 of the PSG, arranged downstream of the camera 5 within the acquisition system 110, as illustrated in [Fig.36].

[0334] Each optical component 191, 192 or 193 of the PSG has a circular shape in the illustrated example and is held between two support pieces attached to one another, one having a housing 233 for receiving the element and the other serving to hold the element in its housing.

[0335] All of these support pieces have a C-shaped upper part open upwards and leave the upper edge of each optical element clear.

[0336] The polarizer 194 has a square or rectangular outline, and is housed in a support 265 provided with an upwardly open slide.

[0337] The optical element 191 is placed between a first support piece 261 and a second support piece 262 which has a housing 233 receiving the element 191, the support piece 261 forming the holding cover.

[0338] The support piece 262 serves as a cover for a third support piece 263 having a housing 233 receiving the optical element 192, as visible in [Fig.21],

[0339] This third support part 263 serves as a cover for a fourth support part 264 having a housing 233 receiving the optical element 193.

[0340] The various support pieces can be superimposed and are fixed together with the support 265 through the base of the PSA on the front face of the head 21 of the colposcope by screws 242, as illustrated in [Fig. 19].

[0341] The optical elements 221, 222 and 223 of the PSA also have a circular shape and are held by support pieces 251, 252, 253 and 254 superimposed on each other.

[0342] Each support piece 251 to 254 has a base 255 provided with holes for the passage of screws 242 used for fixing to the front face of the head 21, and an upper part 256 in the general shape of a C oriented substantially at 45° downwards, the upper part 256 being connected by an upright 257 to the base 255, as illustrated in [Fig.22],

[0343] The support pieces 252 to 254 each have a housing 258 accommodating a respective optical element.

[0344] It can be seen in [Fig.22] that the optical element 221 is held between the parts of support 251 and 252, the support piece 251 serving as a cover, that the optical element 222 is held between the support pieces 252 serving as a cover and 253, and that the element 223 is held between the support pieces 253 serving as a cover and 254.

[0345] The support part 252 carries in the upper part two branches 270 for fixing a filter holder 310 shown in [Fig.34]. This filter holder 310 carries in the example considered three filters 311, 312 and 313 which are respectively a phase retarder L30 whose fast axis is oriented at 30° relative to the linear polarizer PO, a linear polarizer P90 whose transmission axis is oriented at 90° relative to the polarizer PO, and a polarizer PO whose transmission axis is oriented at 0° relative to a reference.

[0346] The filter holder 310 also includes a cleared area placed in front of the PSA once the calibration is complete in order to allow the acquisition of images.

[0347] The filter holder 310 is movable in translation in the example considered, in a generally horizontal direction.

[0348] An electronic circuit 320 allows the computer 60 to know the position of the filter holder, and therefore the filter which is active if applicable.

[0349] The filter holder 310 can be motorized or moved manually, during calibration, so as to successively place the different filters 311 to 313 in front of the PSA.

[0350] Calibration can be carried out using the so-called eigenvalue calibration method (ECM).

[0351] It can be seen in [Fig.25] in particular that the housing 233 is open laterally so as to allow the control wires of the optical element received therein to exit through a passage 233a.

[0352] The housing 233 also has, opposite the passage 233a, a clearance 233b for receiving the control wire at its connection to the optical element. The wire follows the contour of the optical element in a semi-circular channel 233c which matches the profile of the optical element.

[0353] It can be seen in [Fig.23] that the optical elements 221 and 223 of the PSA, of larger diameter, are almost entirely superimposed on the two input ports 211a and 211b, while the optical element 222, of smaller diameter than the elements 221 and 223, is entirely superimposed on the port 221b but partially on the port 211a.

[0354] However, this does not unduly affect the quality of the observed image, taking into account the focusing distance, because the focusing is carried out at a distance much greater than the distance of the optical element from the port 221a.

[0355] [Fig.27] is a schematic and partial section of the light source 50.

[0356] This comprises a xenon lamp 52 and a set of filters 53 and 54, i.e. a first bandpass filter (Edmund Optics #84-728) which allows only light between 400 and 750 nm to be transmitted with a transmittance of approximately 95% and a Second long-pass filter (Edmund Optics #84-754) which allows only light at a wavelength greater than 400 nm to be transmitted with a transmittance of approximately 93%. The choice of these filters allows for a relatively high transmittance, around 90% in the 400-750 nm range.

[0357] The optical guide 51 is held at one end 51a in the axis of the lamp 52 by means of an end piece 56.

[0358] As illustrated in [Fig.29], the optical guide 51 is received at its other end 51b in the axis of an aspherical lens 180 of the colposcope, so as to give an image of the liquid guide approximately 30 cm from the head of the colposcope.

[0359] The lens 180 is placed in front of a prism 181 which reflects the light towards the exit window of the colposcope.

[0360] The optical guide 51 can be held in place if necessary using a set of supports 191 and 192 whose position relative to the body of the head of the colposcope is adjustable in X and Y respectively, so as to allow the position of the end 51b to be precisely adjusted in the axis of the lens 180.

[0361] It is advantageous to use a liquid guide 51 whose core has a diameter of 3 mm, because this allows, as illustrated in [Fig. 31], to benefit from more intense illumination in the center of the area illuminated by the colposcope, compared to a conventional guide whose core is 5 mm in diameter. In particular, the liquid guide with a diameter of 3 mm makes it possible to obtain higher illumination, compared to that obtained with a liquid guide of 5 mm in diameter, over a surface of 3 cm in diameter corresponding approximately to the diameter of the uterine cervix.

[0362] The acquisition system 110 is fixed, as can be seen in particular in figures 36 to 38, on the head 21 of the colposcope and receives the light reaching one of the eyepieces (the same as the one in front of which the PSA is placed) thanks to one or more reflecting prisms integrated into the head 21.

[0363] The camera 5 is equipped with a lens 415, and a filter wheel 411 carrying in the example considered three filters 421, 422 and 423.

[0364] This is for example a tri-band filter adapted to the camera 5, a band-pass filter centered on 650 nm and for example 40 nm of spectral width, and a band-pass filter centered on 700 nm and 50 nm of spectral width.

[0365] The filter wheel 411 is driven in rotation by a motor 410. An angular positioning system by contactor can allow the system to know the angular position of the wheel, and therefore of the filter which must be positioned exactly in the path of the light reaching the camera 5.

[0366] The camera 5 may comprise, as illustrated in [Fig.39], three dichroic prisms, to separate the wavelengths towards the different sensors. The camera comprises for example dichroic prisms 605, 604 and 606, the sensor 603 for the detection blue being fixed on prism 605, sensor 601 for detecting red being fixed on prism 604, itself fixed on prism 605, and sensor 602 for detecting green being fixed to prism 606, itself fixed to prism 604.

[0367] To use the system 10, the user can carry out the polarimetric calibration in a manner known per se using the two faces of the reflector 81 and the different filters of the filter holder 310.

[0368] Then, the head 21 can be positioned so as to image the cervix after dilation of the vagina with the use of a speculum.

[0369] At any time, the practitioner can start recording a video by pressing the button 300 and / or trigger the polarimetric acquisition by pressing the pedal 100. Pressing the pedal triggers the polarimetric acquisition by stopping the video recording, which automatically resumes after the end of the polarimetric acquisition. Pressing the button again stops the video recording.

[0370] Of course, the invention is not limited to the example which has just been described.

[0371] In particular, the optical system of the colposcope can be modified, for example by removing the eyepieces, with observation being carried out solely on screen.

Claims

Claims

1. Multispectral polarimetric imaging system (10), in particular for colposcopy, comprising: - an illumination system comprising at least one light source (50), this illumination system emitting in at least two, in particular at least three, spectral bands, - a polarization state generator (PSG) arranged downstream of the light source and upstream of a target (T) to be imaged, - a polarization state analyzer (PSA) arranged downstream of the target to be imaged, - a multi-sensor camera (5) comprising at least two, better still at least three, sensors (601, 602, 603) for recording respectively at least two, better still at least three, images in said spectral bands.

2. System according to claim 1, the spectral bands being three in number and preferably ranging from 445 nm to 475 nm for the first, from 510 nm to 550 nm for the second, and from 600 nm to 660 nm for the third, these spectral bands being preferably centered on 460 nm, 530 nm and 630 nm respectively.

3. System according to one of claims 1 and 2, the camera (5) being a multi-CCD or multi-CMOS camera, in particular bi-CCD, bi-CMOS, tri-CCD, tri-CMOS, 4-CMOS, 4-CCD, better tri-CCD or tri-CMOS, preferably tri-CMOS.

4. System according to claim 2, the spectral bands being three in number and respectively in the red / near infrared, green and blue.

5. System according to any one of the preceding claims, the camera comprising at least two, better at least three, dichroic prisms, to separate the wavelengths towards the different sensors.

6. A system according to any preceding claim, the polarization state generator (PSG) comprising electrically controllable liquid crystal polarization modulators.

7. A system according to any preceding claim, the polarization state analyzer (PSA) comprising modulators of electrically controllable liquid crystal polarization.

8. System according to one of claims 6 and 7, the liquid crystal polarization modulators being ferroelectric or nematic liquid crystal polarization modulators, preferably ferroelectric.

9. System according to any one of the preceding claims, the light source being a white light source, better a xenon lamp.

10. A system according to any preceding claim, comprising a tri-band dichroic filter downstream of the light source.

11. A system according to any preceding claim, comprising a filter wheel containing three filters (421, 422, 423), arranged in front of the camera.

12. System according to the preceding claim, the filter wheel carrying a multi-band filter, in particular tri-band, preferably to allow acquisition at wavelengths in spectral bands centered around 460nm, 530nm and 630nm by the camera (5), as well as one or more monochromatic dichroic filters, preferably to allow acquisition of images at 650nm and 700nm respectively.

13. A system according to any preceding claim, comprising a processor for generating at least one Mueller polarimetric image of the target in each of the spectral bands.

14. System according to the preceding claim, the processor being arranged to superimpose at least partially, on an image corresponding to the non-polarimetric observation by the camera, an image containing at least one piece of polarimetric information.

15. System according to one of claims 13 and 14, being arranged to make real series of n measurements corresponding to the acquisition of at least 9, in particular 12, better 16, coefficients of an intensity matrix B, the processor being arranged to - generate series of artificial measurements from a grouping of measurements coming from a real series of measurements of a given rank, and from at least one real series of a different rank, in particular following, and from a permutation P of the measurements within each artificial series so as to respect the predefined order of the n measurements within each series, generating a stream of polarimetric images at a frequency higher than that which would be possible with said real series of measurements without the artificial series of measurements, from the polarimetric images produced from the real series of measurements and from the polarimetric images generated from the artificial series interposed between the real series.

16. System according to the preceding claim, the processor being arranged to, before carrying out the permutation P, carry out a spatial resetting of the images corresponding to the measurements in order to take into account possible movements of the areas observed in the field of vision of the camera.

17. A system according to any one of claims 13 to 16, the processor being arranged to carry out an additive decomposition of the Mueller matrix M in the form M = qMnd+pMd where Mnd is the non-depolarizing component and Md is the depolarizing component, the parameters q and p being the weights of the two components of the Mueller matrix, Mnd being the Mueller matrix of a linear phase retarder given by: '10 0 0 . . 0 cos22ü + sitr20cos5 sin26cas20(I - cosô) -sin23sin5 M„d = M(e,5) = rc 0 sin20cosHd(\-cosS) sin"20 + cos220cos co&OsinS .0 sin2üsin8 ~cos20sin5 cosô and Md is the Mueller matrix of a pure depolarizer given by: '1 0 0 0' 0 0 0 0 0 0 0 0 .0 0 0 0. 9 is the azimuthal orientation of the fast (or slow) axis in degrees and delta the phase delay in degrees (ranging from 0 to 180°) with 6* = ^atan(^ ) (-ùtâilsfnô •W ô = atan / «42 sin(20)w44 = atan = 4atan(^î) 2 \ cos 2.0 / M J _ çosô _ 1 _ œw) 1 _ 1,4.1 U “ ^44 1 “ 1 “ 1 “ u J- the coefficients mÿ (i,j=l,2,3,4) being those of the Mueller matrix M, mu being the unnormalized and unpolarized intensity coefficient of M, the other coefficients mÿ (i,j = 1,2,3,4) being normalized with respect to m

18. h- System according to claim 17, the processor being arranged to calculate the depolarization by performing the following operation: Depolarization = 1--^

19. System according to any one of the preceding claims, being arranged to generate the parallel display of at least one non-polarimetric image in at least one of the spectral bands, and of at least one polarimetric image.

20. System according to any one of the preceding claims, being a colposcopy system, this system comprising: - a rolling base (31), - a mast (30) carried by the rolling base, - a colposcope head (21) carried by an articulated arm connected to the base, - at least one reference reflector (81) carried by the mast (30), used for calibrating the polarimetry system, and preferably also: - a workstation comprising a computer (60), carried by the rolling base, - a keyboard (41) carried by an articulated arm connected to the mast, - a screen (40) carried by an articulated arm connected to the mast at a height greater than that of the arm supporting the keyboard.

21. System according to claim 20, the reference reflector (81) being arranged in a light shield (80), at the top of the mast, being articulated around a vertical axis, and comprising on one face a frosted metal wall, preferably made of aluminum, and on the opposite side a reference surface having known spectral properties.

22. System according to one of claims 20 and 21, comprising a pedal (100) which allows the user to trigger a predefined action, in particular to start a polarimetric acquisition, the head (21) of the colposcope comprising a button (101) making it possible to trigger another predefined action, in particular the start and stop of the video recording.

23. A system according to any preceding claim, being

24.

25.

26. a colposcopy system, this system comprising an illumination system comprising • At least one light source (50), • a liquid light guide (51), to be connected at one end to the light source, this light guide having a core diameter less than or equal to 5 mm, the system comprising a colposcope head (21), comprising an optical system having a light input port connected to the other end of the liquid light guide, and a light output port for illuminating an area to be observed. The system of claim 23, the optical system comprising between the light input port and the light output port an aspherical lens (180) and a reflecting prism (181). System according to claim 23 or 24, the internal diameter of the liquid optical guide (51) being between 2.5 and 3.5 mm. A method of training an artificial intelligence system (7), preferably comprising at least one convolutional neural network, in which the artificial intelligence system receives as input non-polarimetric images and polarimetric images generated by the polarimetric imaging system according to any one of the preceding claims.

Images