FIEED OF INVENTION

[0001] The invention concerns a computer-implemented method for estimating presence or absence of at least one biomarker in a urine sample, and/or estimating a value of at least one biomarker of a urine sample.

[0002] The invention applies to the field of urinalysis, and more precisely to the estimation of biomarkers concentration.

BACKGROUND OF INVENTION

[0003] Clinical urine tests, also known as urinalysis, are an examination of urine for certain biomarkers. Urinalysis is widely used for health check and/or diagnosis of diseases.

[0004] Examination of physical properties of urine is recommended as an initial evaluation of urine samples before more sophisticated tests (chemical, immunologic) are performed in laboratories. Currently, these urine physical properties examinations are mostly performed using different machines that measure different parameters or biomarker values, for instance based on an optical spectrum of the urine sample. More specifically, a high-resolution spectrum of a urine sample needs to be acquired using a dedicated laboratory equipment in order to derive the biomarker value from said high- resolution spectrum.

[0005] However, such method is not fully satisfactory.

[0006] Indeed, such method requires expensive pieces of equipment to measure different parameters, such as high-resolution optical spectra of the urine (with a resolution of typically 0.5 nm) and require human expertise for interpretation of said spectra. Such method is also time consuming, since the results must be aggregated with results of other chemical and/or physical analysis in a report and returned to the practitioner. Thus, the final results are generated long after the sampling, causing a delayed diagnosis and, consequently, unwanted and unnecessary stress to the patient.

[0007] Therefore, a purpose of the invention is to provide a method for estimating presence or absence of at least one biomarker in a urine sample, and/or estimating values of biomarkers of a urine sample that is more cost-effective and less time-consuming than the known methods.

SUMMARY

[0008] To this end, the present invention is a method of the aforementioned type, including, for each biomarker, the steps of:

- receiving a raw optical spectrum of the urine sample defined over a plurality of first wavelengths of a predetermined wavelength range,

- based on the raw optical spectrum of the urine sample, determining a corresponding high-resolution optical spectrum of the urine sample, the high-resolution optical spectrum being defined over a plurality of second wavelengths of the predetermined wavelength range, a wavelength interval between two successive second wavelengths being less than a wavelength interval between two successive first wavelengths; and

- implementing at least one corresponding biomarker model to estimate a value of the biomarker based at least on the determined high- resolution optical spectrum, each biomarker model being trained based on a training dataset including input data comprising a plurality of reference optical spectra defined over at least part of the predetermined wavelength interval, and output data comprising a plurality of reference values of the biomarker, each reference value of the biomarker being associated to a respective reference optical spectrum.

[0009] In other words, the method of the aforementioned type comprises the steps of: - receiving a raw optical spectrum of the urine sample defined over a plurality of first wavelengths of a predetermined wavelength range;

- associating to the raw optical spectrum of the urine sample a corresponding high-resolution optical spectrum of the urine sample, the high-resolution optical spectrum being defined over a plurality of second wavelengths of the predetermined wavelength range, a wavelength interval between two successive second wavelengths being less than a wavelength interval between two successive first wavelengths; and

- implementing at least one corresponding biomarker model to estimate a value of the biomarker based at least on the associated high- resolution optical spectrum, each biomarker model being trained based on a training dataset including input data comprising a plurality of reference optical spectra defined over at least part of the predetermined wavelength interval, and output data comprising a plurality of reference values of the biomarker, each reference value of the biomarker being associated to a respective reference optical spectrum.

[0010] Indeed, thanks to the invention, a precise determination of each biomarker value can be achieved without resorting to expensive equipment to measure multiple high- resolution optical spectra of the urine sample, then aggregating said multiple spectra in order to obtain a final optical spectrum on a broad wavelength range, which will serve as a basis for determining biomarker values. On the contrary, a relatively cheap device for point of care urinalysis, which has lower resolution (from 10 nm to 30 nm) but wider wavelength range (typically infrared, visible and ultraviolet), can be used: in this case, said device almost instantaneously provides a raw optical spectrum. The high-resolution optical spectrum associated to each biomarker is then quickly determined, based on the raw optical spectrum, over said broad wavelength range. Moreover, using biomarker models that are specific to each biomarker improves the relevancy of the determined high- resolution optical spectrum. As a result, quick and reliable measurements of biomarker values are achieved. [0011] According to other advantageous aspects of the invention, the method includes one or more of the following features, taken alone or in any possible combination:

[0012] - for each biomarker model, the input data of the corresponding training dataset also include a plurality of reference electroanalytical measurements, preferably a plurality of reference conductivity measurements, each reference value of the biomarker being further associated to a respective reference electroanalytical measurement, and for each biomarker, the method further comprising implementing each corresponding biomarker model to estimate a value of said biomarker based on the corresponding high-resolution optical spectrum and on a received electroanalytical measurement of the urine sample;

[0013] - the method further comprises correcting the received raw optical spectrum based on a temperature of the urine sample;

[0014] - the received electroanalytical measurement includes at least one conductivity value acquired between 0 Hz and 1 MHz;

[0015] - the method further comprises correcting the received conductivity measurement based on a temperature of the urine sample;

[0016] - associating to each raw optical spectrum a high-resolution optical spectrum comprises performing a spectral reconstruction algorithm on the corresponding raw optical spectrum based on a transmission curve of each input optics of an optical spectrometer used to acquire said raw optical spectrum;

[0017] - each biomarker is associated to a corresponding relevant wavelength range, and for each raw optical spectrum, associating to the raw optical spectrum the corresponding high-resolution optical spectrum includes identifying a base high-resolution optical spectrum, among a plurality of stored base high-resolution optical spectra, that is the closest to said raw optical spectrum over the relevant wavelength range with regard to a predetermined metric, the identified base high-resolution optical spectrum forming the high-resolution optical spectrum;

[0018] - for each raw optical spectrum associating to the raw optical spectrum the corresponding high-resolution optical spectrum includes identifying a base high- resolution optical spectrum, among a plurality of stored base high-resolution optical spectra, that is the closest to said raw optical spectrum over the predetermined wavelength range with regard to a predetermined metric, the identified base high-resolution optical spectrum forming the high-resolution optical spectrum;

[0019] - the predetermined wavelength range overlaps with at least one of the following wavelength sub-ranges: [270 nm; 400 nm], [390 nm; l lOO nm] and/or [800 nm; 2600 nm];

[0020] - the method further comprises:

- step of determining whether the estimated values of at least two biomarkers satisfy a predetermined physiological relationship; and

- generating a warning signal if not;

[0021] - the method further comprises a step of data clustering, based on at least one raw optical spectrum, to assign the urine sample to one of a plurality of predetermined patient classes, at least one predetermined patient class being associated to a corresponding pathology or health status, the biomarker model depending on the assigned patient class;

[0022] - each biomarker is associated to at least two distinct biomarker models, and, for each biomarker, the corresponding estimated value is a function of a result provided by each biomarker model;

[0023] - the method further comprises a step of prescreening including:

- performing data analysis on each raw optical spectrum to determine, based on features of the raw optical spectrum, whether the corresponding urine sample is associated to a patient having a predetermined atypical status;

- generating an alert signal if the urine sample is found to be associated to a patient having a predetermined atypical status.

[0024] The invention also relates to a computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method as defined previously. [0025] The invention also relates to a system for estimating presence or absence of at least one biomarker in a urine sample, and/or estimating a value of at least one biomarker of a urine sample, the system comprising a processing unit configured to, for each biomarker:

- receive a raw optical spectrum of the urine sample acquired over a plurality of first wavelengths of a predetermined wavelength range,

- associate to the raw optical spectrum of the urine sample, a corresponding high-resolution optical spectrum of the urine sample, the high-resolution optical spectrum being defined over a plurality of second wavelengths of the predetermined wavelength range, a wavelength interval between two successive second wavelengths being less than a wavelength interval between two successive first wavelengths; and

- implement at least one corresponding biomarker model to estimate a value of the biomarker based at least on the determined high- resolution optical spectrum, each biomarker model being trained based on a training dataset including input data comprising a plurality of reference optical spectra defined over at least part of the predetermined wavelength interval, and output data comprising a plurality of reference values of the biomarker, each reference value of the biomarker being associated to a respective reference optical spectrum.

BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematical representation of a urinalysis system according to the invention;

[0027] Figure 2 is a graph showing a raw optical spectrum and a corresponding high- resolution optical spectrum determined by the urinalysis system of figure 1; and

[0028] Figure 3 is a chart showing a urinalysis method according to the invention. DETAILED DESCRIPTION

[0029] A urinalysis system 2 according to the invention is shown on figure 1.

[0030] The urinalysis system 2 is configured to acquire data relating to a urine sample of a patient, and to measure presence or absence of at least one biomarker in a urine sample, and/or measure a value of at least one biomarker of said urine sample. The urinalysis system 2 is able to perform a qualitative measurement (determination of presence/absence of a biomarker) and/or quantitative measurement (determination of concentration of a biomarker). In particular, when the concentration of a biomarker in a urine sample is below a predetermined value, it indicates that said biomarker is not detectable from the urine sample. Inversely, when the concentration of a biomarker in a urine sample is above the limit of detection, it indicates that said biomarker is present in the urine sample, detectable and quantitatively measurable.

[0031] The value of a biomarker refers herein to the concentration or concentration range of said biomarker in the urine sample.

[0032] More precisely, the urinalysis system 2 comprises at least one probe 4 and a processing unit 6. Optionally, the urinalysis system 2 includes a docking station 8 for allowing communication between the probe 4 and the processing unit 6. Alternatively, the processing unit 6 and each probe 4 are configured to directly communicate with each other.

[0033] Probe 4

[0034] The urinalysis system 2 comprises a probe 4 configured to acquire at least one raw optical spectrum of the urine sample, over a plurality of first wavelengths of a predetermined wavelength range. In this case, the probe 4 may be named optical probe 4. In other words, the raw optical spectrum comprises a plurality of couples, each couple comprising a first wavelength and a corresponding measured light intensity. Said intensity may be representative of a transmission of the urine sample at the corresponding first wavelength, or a fluorescence of the urine sample at the corresponding first wavelength. An example of such raw optical spectrum is shown on figure 2, and corresponds to the circles on the graph. In this example, a first wavelength interval between two successive first wavelengths is inferior or equal to 30 nm.

[0035] To acquire said raw optical spectrum, the probe 4 includes an optical spectrometer (not shown), such as a miniaturized optical spectrometer. Preferably, for each first wavelength, the optical spectrometer includes a corresponding light detection member, which preferably includes a photodetector. Preferably, each photodetector is provided, at an optical input thereof, with input optics. For instance, said input optics comprise an optical filter, said optical filter having a known transmission curve. Alternatively, or in addition, the input optics comprise an interferometer, such as a Fabry-Perot interferometer, preferably having an adjustable transmission curve. In other words, the optical probe 4 includes a lighting module configured to emit light in the urine sample, and an optical sensor configured to receive light emitted by said urine sample, light transmitted through said urine sample and/or light scattered in said urine sample.

[0036] Preferably, the predetermined wavelength range overlaps with at least one of the following wavelength sub-ranges: [270 nm; 400 nm] (corresponding to UltraViolet range); [390 nm; l lOO nm] (corresponding to near infrared-visible range); and/or [800 nm; 2600 nm] (corresponding to infrared range). In other words, the predetermined wavelength range is at least partially included in the [270 nm; 2600 nm] interval. This is advantageous, because the main urine components can be identified in these wavelength ranges. Moreover, is such wavelength range, the cost of the optical and electronic components is generally lower compared to other wavelength ranges: the cost of the probe 4.

[0037] The optical probe 4 is configured to receive light transmitted through (and/or reflected by) the urine sample such as visible and near infrared light, thus allowing the urinalysis system 2 to perform visible spectrometry, near infrared spectrometry on a urine sample. Concerning visible spectrometry, it is possible to detect biomarkers such as, for example, minerals (e.g. Na, K, Ca, Mg, Cl, P), creatinine, urea, urine osmolality, urine specific gravity, uric acid, urine pH, ammonium, citrate, oxalate, albumin and microalbuminuria, total proteins, bilirubin, urobilinogen, uroporphyrin, coproporphyrin, porphobilin, red blood cells, hemoglobin, white blood cells, ketones, glucose or presence of bacteria or crystals. Concerning near infrared spectrometry, it is possible to detect biomarkers such as, for example, minerals (e.g. Na, K, Ca, Mg, Cl, P), creatinine, urea, urine osmolality, urine specific gravity, uric acid, urine pH, ammonium, citrate, oxalate, albumin, total proteins, bilirubin, urobilinogen, uroporphyrin, coproporphyrin, porphobilin, red blood cells, hemoglobin, white blood cells, ketones, glucose or presence of bacteria or crystals. Urine osmolality and urine specific gravity are biomarkers of hydration, very useful to determine how well the kidneys are working. Creatinine is also a biomarker that indicates the proper functioning of kidneys.

[0038] The optical probe 4 is also configured to receive light emitted by the urine sample to anticipate autofluorescence of urine, thus allowing the urinalysis system 2 to perform autofluorescence spectrometry on a urine sample. Fluorescence spectrometry allows the detection of biomarkers such as, for example, red blood cells, bacteria, heavy metals, NADH (hydrogenated nicotinamide adenine dinucleotide), NADPH (Nicotinamide adenine dinucleotide phosphate), FAD (flavin adenine dinucleotide), elastin, collagens, tryptophan, porphyrins, riboflavin, bilirubin, flavoproteins, pteridines compounds (neopterin, pterine, xanthopterin, isoxanthopterin) or other endogenous fluorophores.

[0039] Preferably, the urinalysis system 2 further comprises a probe 4 configured to acquire an electroanalytical measurement of the urine sample. In this case, the probe 4 may be named electroanalytical probe 4. The electroanalytical probe 4 allows the urinalysis system 2 to perform electrochemical measurement, such as conductimetry, coulometry, potentiometry, amperometry, or voltammetry, on a urine sample. As an example, the conductivity of urine arises mainly from the mobility of the constituents (hydrated ions) present in the sample and therefore gives a measure of the ability of the sample to conduct a charge applied to it. Thus, by measuring the conductivity of a urine sample, it is possible to determine the concentration of ions (for example Na ⁺ , K ⁺ , Ca ²⁺ , Mg ²⁺ , H ⁺ /CO3‘ or CT). Other electroanalytical methods allow the urinalysis system 2 to perform analysis of uric acid, oxalate, citrate, creatinine, bilirubin, nitrites, ascorbic acid, amino acids such as tryptophane or tyrosine, BNP (brain natriuretic peptide), hormones such as Beta-HCG, cortisol steroid hormones, anti-mullerian hormone (AMH) or thyroid hormones, and marker of collagen degradation, in said sample. [0040] Preferably, the electroanalytical probe 4 is a conductivity probe, configured to perform conductivity measurement on the urine sample. Said conductivity probe comprises two or more electrodes.

[0041] Said electroanalytical measurement, preferably a conductivity measurement, includes at least one electroanalytical value (conductivity value), preferably acquired at 0 Hz (z.e., a conductivity value associated to a DC current), and/or at frequencies higher than 0 Hz (z.e., a conductivity value associated to an AC current), for instance up to 1 MHz. For instance, the electroanalytical measurement, preferably a conductivity measurement, is an electroanalytical spectrum (a conductivity spectrum) acquired at one or multiple frequencies in a frequency range included in the [0 Hz; 1 MHz] range, preferably in the [1 Hz; 1 MHz] range, more preferably in the [1 Hz; 100 kHz] range. In the case the electroanalytical measurement is a conductivity measurement, the measured conductivity is the urine conductivity, due to the presence of ions in the urine.

[0042] The urinalysis system 2 thus provides a non-invasive scan of urine samples based on four technologies: visible spectrometry, near infrared spectrometry (or infrared spectrometry), autofluorescence spectrometry and electroanalytical methods (preferably conductimetry). It allows a thorough physico-chemical characterization of a urine sample.

[0043] Combining biomarkers detection by electroanalytical method, fluorimetry and optical spectrometry is particularly advantageous as it allows for a thorough, and yet fast, scan of the urine sample, determining simultaneously the presence/absence and/or concentration of several biomarkers in a single urine sample. It also provides a better result, i.e. more precise, less false negative, than processing separately optical and electric measures. Finally, concomitant measurement of sample temperature allows to improve precision of optical and electroanalytical measurement.

[0044] For example, near-infrared spectrometry and electroanalytical measurement are complementary to determine precisely mineral concentration (e.g. Na, K, Ca, Mg, Cl, P). Indeed, inorganic ions in aqueous solutions do not directly absorb NIR light but influence spectral patterns at specific wavelength through ion- water interactions. Similarly, urine saturation and crystallization (e.g. presence of Calcium Oxalate crystals) can be detected optically. Thus, the optical spectrum brings both qualitative and quantitative information. This first estimation of each mineral concentration is completed by electroanalytical measurement, such as, for example, conductimetry measurement, that reflects total concentration of cations and anions in solution, each ion having a specific molar conductivity. The electroanalytical measures at different frequencies allowing the precise determination of ion concentration. Moreover, as the mobility of ions increase with temperature, a simultaneous measurement of temperature on top of optical spectrum and electroanalytical measurement allows an even more precise measure of mineral concentration.

[0045] Moreover, visible, NIR and IR spectra contain specific wavelengths that are heavily associated with similar urine biomarkers. Thus, combination of these spectral information significantly improves biomarker concentration prediction. For example, information can be found on osmolality below 700 nm, between 800 nm and 850 nm, around 1000 nm, around 1150 nm and above 1200 nm, meaning all ranges contain information, sometimes redundant and sometimes not, allowing to improve osmolality measurement. Finally, fluorescence spectroscopy is used to identify specific biomarkers in combination with visible spectra.

[0046] The urinalysis system 2 may further comprise a probe 4 configured to acquire a temperature value of the urine sample. In this case, the probe 4 may be named temperature probe 4. Said temperature probe 4 is preferably an infrared temperature sensor. The temperature probe 4allows the measurement of the temperature of the urine sample which will then be converted into an electrical signal associated with said temperature.

[0047] The urinalysis system 2 may further comprise a probe 4 configured to emit and/or detect ultrasound. In this case, the probe 4 may be named ultrasound probe 4. Said ultrasound probe 4 provides information on physical, mechanical and chemical properties of the urinesample such as density, presence of cells and/or crystals.

[0048] The urinalysis system 2 may further comprise a probe 4 configured to perform Raman spectroscopy measurement, Electrochemical Impedance Spectroscopy, or Fourier transform infrared spectroscopy measurement (FTIR) on the urine sample. [0049] The probe 4 is, for instance, similar to the probe disclosed in European patent application EP 20 306 576.

[0050] The urinalysis system 2 may be used to measure presence or absence of at least one biomarker in a biological sample, and/or measure a value of at least one biomarker of a biological sample, said biological sample being selected in the group comprising blood, urine, lymph, fluidified feces, adipose tissue, bone marrow, cerebrospinal fluid, sperm, cord blood, breast milk, tears, water, or saliva. The biological sample can be provided by a human or an animal, e.g. cattle, sheep, pigs, horses or any other animal.

[0051] Processing unit 6

[0052] The processing unit 6 will now be disclosed.

[0053] According to the invention, the expression “processing unit” should not be construed to be restricted to hardware capable of executing software, and refers in a general way to a processing device, which can, for example, include a microprocessor, an integrated circuit, or a programmable logic device (PLD). The processing unit 6 may also encompass one or more Graphics Processing Units (GPU) or Tensor Processing Units (TSU), whether exploited for computer graphics and image processing or other functions. Additionally, the instructions and/or data enabling to perform associated and/or resulting functionalities may be stored on any processor-readable medium such as, e.g., an integrated circuit, a hard disk, a CD (Compact Disc), an optical disc such as a DVD (Digital Versatile Disc), a RAM (Random- Access Memory) or a ROM (Read-Only Memory). Instructions may be notably stored in hardware, software, firmware or in any combination thereof.

[0054] Biomarkers

[0055] The processing unit 6 is configured to estimate presence or absence of at least one biomarker in a urine sample, and/or estimate a value of at least one biomarker of the urine sample.

[0056] The aforementioned value may be a concentration or a concentration range of a predetermined chemical compound, i.e. biomarker, in the urine sample. [0057] The biomarker may be selected among minerals (e.g. sodium, potassium, calcium, magnesium, chloride, phosphorus), creatinine, urea, urine osmolality, urine specific gravity, uric acid, urine pH, ammonium, citrate, oxalate, albumin and micro-albuminuria, total proteins, bilirubin, urobilinogen, urobilin, porphobilin, porphobilinogen, uroporphyrin, coproporphyrin, red blood cells, hemoglobin, white blood cells, ketones, glucose, bacteria, heavy metals, NADH (hydrogenated nicotinamide adenine dinucleotide), NADPH (Nicotinamide adenine dinucleotide phosphate), FAD (flavin adenine dinucleotide), elastin, collagens, tyrosine, tryptophan, porphyrins, riboflavin, bilirubin, flavoproteins, pteridines compounds (neopterin, pterine, xanthopterin, isoxanthopterin) or other endogenous fluorophores, BNP (brain natriuretic peptide), hormones such as Beta-HCG, cortisol steroid hormones, anti-mullerian hormone (AMH) or thyroid hormones, or crystals.

[0058] In particular, the biomarker may be a biomarker representative of a fluid balance of the patient, for instance at least one of: urine specific gravity and osmolality. Urine osmolality and urine specific gravity are biomarkers of hydration, very useful to determine how well the kidneys are working.

[0059] Alternatively, or in addition, the biomarker may be a biomarker representative of patient kidney function. For instance, such biomarker is at least one of: creatinine (estimation of concentration or concentration range), albumin (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range), ammonia, proteins (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range).

[0060] Alternatively, or in addition, the biomarker may be a biomarker indicative of the presence of kidney stone(s) in one or both kidney(s) of the patient. For instance, such biomarker is at least one of citrate, oxalate (estimation of concentration or concentration range) and crystals (indication of presence or absence in the urine sample).

[0061] Alternatively, or in addition, the biomarker may be a biomarker representative of metabolism of the patient. For instance, such biomarker may be at least one of: a pH or a pH range of the urine sample, uric acid, ketones (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range), bilirubin (estimation of concentration or concentration range), urobilinogen (estimation of concentration or concentration range) and urea (estimation of concentration or concentration range).

[0062] Alternatively, or in addition, the biomarker may be a biomarker indicative of hyperglycemia in the patient. For instance, such biomarker may be glucose (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range.

[0063] Alternatively, or in addition, the biomarker may be a biomarker indicative of an infection of the patient. For instance, such biomarker may relate to at least one of: hematuria (z.e., an indication relating to the presence or the absence thereof, or a concentration range or a concentration thereof), leukocytes (z.e., an indication relating to the presence or absence, or a concentration range or concentration thereof), and bacteria (z.e., an indication relating to the presence or the absence, range of concentration or concentration thereof).

[0064] Alternatively, or in addition, the biomarker may be a biomarker indicative of a specific pathology such as, for example, rare disease. For instance, such biomarker may be at least one of: porphyrin (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range), porphobilinogen (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range), porphobilin (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range), hemoglobin (indication of presence or absence in the urine sample, and/or estimation of concentration or concentration range), oxalate (indicating hyperoxaluria, estimation of concentration or concentration range), and endogenous fluorophores(indication of presence or absence in the urine sample).

[0065] Alternatively, or in addition, the biomarker may be a biomarker indicative of patient oxidative stress. [0066] Alternatively, or in addition, the biomarker may be a biomarker indicative of patient immunity status.

[0067] Alternatively, or in addition, the biomarker may be a biomarker indicative of patient diet quality.

[0068] Configuration of the processing unit 6

[0069] More specifically, the processing unit 6 is configured to estimate (determine or indicate) presence or absence of at least one biomarker in a urine sample, and/or estimate (or determine) the value of each biomarker based at least on a raw optical spectrum of the urine sample. Said raw optical spectrum is, for instance, received from the probe 4, in particular the optical probe 4.

[0070] The processing unit 6 is configured to perform an estimation (or determination) method, a workflow of which is schematically shown on figure 2. Said estimation method comprises a high-resolution optical spectrum determination step 20 (hereinafter “determination step 20”) and a biomarker value estimation step 30 (hereinafter “estimation step 30”). More precisely, the processing unit 6 is configured to associate, during the determination step 20, and for each biomarker, to the raw optical spectrum a corresponding high-resolution optical spectrum. Moreover, for each biomarker, the processing unit 6 is configured to estimate presence or absence in the urine sample and/or a respective value, being the concentration or concentration in the urine sample, based at least on the high-resolution optical spectrum associated to the raw optical spectrum.

[0071] A detailed description of steps 20 and 30 will now be provided.

[0072] Determination step 20

[0073] As mentioned previously, the processing unit 6 is configured to associate to the raw optical spectrum, during the determination step 20, and for each biomarker, a corresponding high-resolution optical spectrum.

[0074] Such high-resolution optical spectrum is defined over a plurality of second wavelengths of the predetermined wavelength range. Moreover, a second wavelength interval between two successive second wavelengths is less than a first wavelength interval between two successive first wavelengths. In other words, the high-resolution optical spectrum has a spectral resolution that is greater than a spectral resolution of the raw optical spectrum, while being similar in shape to the raw optical spectrum. An example of such high-resolution optical spectrum is shown on figure 2, and corresponds to the dashed line on the graph. In this example, a second wavelength interval between two successive first wavelengths is 0.5 nm.

[0075] In other words, the high -resolution optical spectrum is the spectrum having a higher resolution than the raw optical and that most closely matches the data couples of the raw optical spectrum.

[0076] The high-resolution optical spectrum was previously acquired on model samples in medical laboratories using different machines that measure different parameters or biomarkers concentration, then it was stored in a data library comprising a plurality of high-resolution optical spectra.

[0077] In one configuration, in order to determine the high-resolution optical spectrum associated to a given biomarker, the processing unit 6 is configured to perform spectral reconstruction on the corresponding raw optical spectrum based on a predetermined reconstruction algorithm. Said spectral reconstruction algorithm is based on a transmission curve of the input optics of each optical spectrometer used to acquire said raw optical spectrum (generally the optical spectrometer included in the probe 4).

[0078] For instance, the processing unit is configured to solve M equations of the following form: where bi is an amplitude of a detection signal output by light detection member i, “i” being an integer comprised between 1 and M, M being an integer lower than or equal to N, N being the total number of points of the raw optical spectrum, that is the total number of light detection members; f( ) is the light spectrum reaching each light detection member (z.e., the high-resolution optical spectrum); (|)i( ) is the transmission curve of the input optics of light detection member i; h(X) is a response function of the photodetector of light detection member i; and ei is a measurement error.

[0079] Advantageously, spectral reconstruction allows to train very quickly models with incomplete spectra, in particular in the case of a large database of complete spectra. This also allows to use standard algorithm models developed on reference spectra, it facilitates the visualization of the results, and it allows to have a single reference library and several measurement tools with different sensors and different discrete point measurements depending on the biomarkers they measure.

[0080] In an alternative and preferred configuration, at least one biomarker is associated to a corresponding relevant wavelength range. In this case, starting from the raw optical spectrum, and in order to associate the high-resolution optical spectrum associated to a given biomarker to the raw optical spectrum, the processing unit 6 is configured to identify, among a plurality of stored base high-resolution optical spectra, the base high- resolution optical spectrum that is the closest to said raw optical spectrum over the relevant wavelength range with regard to a predetermined metric. In this case, the identified base high-resolution optical spectrum forms the high-resolution optical spectrum. This direct identification between raw spectrum and high resolution spectrum is advantageously faster, more scalable as it allows the reference spectra database to evolve/expand without changing the reconstruction algorithm and more agile as it allows the use of devices producing different degraded data (generating degraded data on different locations of the spectra depending on the biomarker(s) they measure).

[0081] As an example, the processing unit 6 is configured to implement a closest distance matching algorithm (such as a ^-nearest neighbors algorithm) in order to determine the high-resolution optical spectrum, based on the raw optical spectrum and each base high- resolution optical spectrum.

[0082] In this preferred configuration, no interpolation models or reconstructions of spectrum (such as the application of a super-resolution algorithm) are needed as a direct correspondence is performed between the raw optical spectrum and a high-resolution optical spectrum from the data library. In this preferred configuration, the raw optical spectrum is directly compared to high resolution optical spectra from the data library. The raw optical spectrum presents values acquired over a plurality of first wavelengths of a predetermined wavelength range. On the contrary, the high-resolution optical spectra from the data library are defined over a plurality of second wavelengths, the interval between two successive second wavelengths being smaller than the interval between two successive first wavelengths.

[0083] Advantageously, the processing unit 6 is configured to correct, prior to the association of a high-resolution optical spectrum to the received raw optical spectrum, the received raw optical spectrum based on a temperature of the urine sample. This allows to compensate for any potential temperature-related bias of the optical spectrometer of the probe 4, notably temperature-induced spectral variations such as the position and intensity of spectral absorption bands of the chemical compounds in the urine sample.

[0084] Estimation step 30

[0085] Moreover, for each biomarker, the processing unit 6 is configured to estimate, during the estimation step 30, presence or absence in the urine sample and/or estimate a respective value (i.e. concentration or concentration range) based at least one the determined high-resolution optical spectrum.

[0086] More precisely, in order to estimate presence or absence in the urine sample and/or estimate a respective value (i.e. concentration or concentration range) of a given biomarker, the processing unit 6 is configured to implement at least one corresponding biomarker model to estimate said presence/absence or value based at least on the high- resolution optical spectrum.

[0087] Each biomarker model has been trained prior to the implementation of the estimation step 30. More precisely, for each biomarker, each corresponding biomarker model has been trained based on a training dataset including input data comprising a plurality of reference optical spectra defined over at least part of the predetermined wavelength interval, and output data comprising a plurality of reference values of the biomarker, each reference value of the biomarker being associated to a respective reference optical spectrum. Preferably, the reference optical spectra have the same resolution as the high-resolution optical spectra.

[0088] Advantageously, for at least one biomarker, the processing unit 6 is further configured to determine the corresponding value based on the electroanalytical measurement of the urine sample, such as for example conductimetry, coulometry, potentiometry, amperometry, or voltammetry, preferably conductivity measurement (also called conductimetry). In this case, for at least one biomarker model corresponding to a given biomarker, the input data of the corresponding training dataset also include a plurality of reference electroanalytical measurements, preferably conductivity measurements. As a result, each reference value of the biomarker is further associated to a respective reference electroanalytical measurement, preferably conductivity measurement. In this case, for each concerned biomarker, the processing unit 6 is further configured to implement each corresponding biomarker model to estimate a value of said biomarker based on the corresponding high-resolution optical spectrum and on the received electroanalytical measurement of the urine sample. This is advantageous, as the inventors have realized that an electroanalytical measurement leads to a more precise estimation of the value of almost each of the aforementioned biomarker. It allows a thorough physico-chemical characterization of a urine sample. The combination of optical estimations and electroanalytical estimations advantageously allows to estimate (or measure) different characteristics of each molecule resulting in a better characterization of the sample, such as for example, its fluorescence, its absorbance, its electrochemical oxidation, etc. These responses constitute a unique fingerprint for each molecule. These results, in association with modeling, allow to differentiate distinct markers and to quantify them more precisely.

[0089] In this case, the processing unit 6 is advantageously also configured to correct the received electroanalytical measurement based on a temperature of the urine sample prior to the estimation of each biomarker. This is advantageous, as such step allows to compensate for the temperature-dependency of the electroanalytical measurement of a solution. [0090] Advantageously, each biomarker is associated to at least two distinct biomarker models. In order to be distinct from each other, said biomarker models have been trained based on different training datasets and/or are of different types (e.g., an artificial neural network, a decision tree, a support-vector machine, a Bayesian network and so on). In this case, for each biomarker, the corresponding estimated value is a function of a result provided by each of the associated biomarker models. For instance, said estimated value is equal to the mean value of the results provided by all the biomarker models. Computing the mean value of the results allows reducing some uncertainties hold by each biomarker model, and thus renders the determination of the biomarker concentration more accurate. According to another example, the number of biomarker models is greater than or equal to 3, and said estimated value is equal to the value which has the highest number of occurrences. Preferably, said estimated value is equal to the mean value of the results provided by all the biomarker models if a relative difference between the results provided by the biomarker models is within a predetermined tolerance range. In this case, if said mean value is outside the tolerance range, the processing unit 6 may be configured to output an alert signal.

[0091] Advantageously, for at least two biomarkers, the processing unit 6 is configured to determine, after the values of said biomarkers have been estimated, whether the corresponding estimated values satisfy a predetermined physiological relationship. In this case, the processing unit 6 is further configured to generate a warning signal if the predetermined physiological relationship is not satisfied. This is advantageous, as such step allows to check for consistency between the estimated values of the biomarkers, and/or allow detection of abnormalities (such as health issues) on the patient side.

[0092] For example, the urine osmolality value can be estimated using the concentration of the major osmotically active solutes in solution, i.e sodium, potassium and urea: estimated Uosm = 2x(UNa + UK) + Uurea where UNa, UK and Uurea are the urinary concentrations of sodium, potassium and urea, respectively, in mmol/L.

[0093] Advantageously, the processing unit 6 is configured to perform, prior to the implementation of each biomarker models, a step of data clustering, based on at least one raw optical spectrum, to assign the urine sample to one of a plurality of predetermined patient classes, at least one predetermined patient class being associated to a corresponding pathology or health status. In this case, at least one biomarker model depends on the assigned patient class. In other words, the raw optical spectrum is assigned to a predetermined patient class, and as a consequence, a specific set of biomarker models is selected to determine the corresponding biomarker values or concentrations. This step of data clustering corresponds to a step of profiling the raw optical spectrum, in order to make the determination of the biomarker concentrations more relevant. For instance, for a given biomarker, the processing unit 6 is configured to select a biomarker model for estimating the value of said biomarker depending on the patient class to which the patient has been assigned. As a result, estimation of each biomarker value is more reliable. As an example, glycosuria can be encountered in patient suffering diabetes, requiring the use of specific algorithms to estimate glucose concentration.

[0094] For instance, such data clustering includes projecting the raw optical spectra in a predetermined clustering latent space. Such projection is, for example, based on the implementation of a trained artificial intelligence model. Alternatively, such projection includes performing a predetermined data transformation algorithm, such as principal component analysis. By comparing the result of such projection to predetermined patient classes, a patient class is assigned to the current patient. The assigned patient class is, for instance, the patient class that is the closest to the projected raw optical spectrum, with respect to a predetermined metric.

[0095] According to an advantageous embodiment, the processing unit 6 is also configured to perform a step of prescreening, before the determination step 20, that is to say based on each raw optical spectrum. Such step of prescreening is aimed at detecting a patient health status prior to, or without, computing each biomarker value. Such prescreening is based on the fact that, for a predetermined set (hereinafter referred to as “atypical status” or “atypical health status”) of health statuses, the raw optical spectrum exhibits specific features. In other words, during prescreening, the processing unit 6 performs data analysis on each raw optical spectrum to determine, based on the features of the raw optical spectrum, whether the urine sample is associated to a patient having a predetermined atypical status. [0096] For instance, the atypical status may be one of: metabolic malfunction, drug use, doping (z.e., use of performance enhancing drugs), supplement intake, absorption of one or more heavy metals, absorption of one or more drugs belonging to predetermined classes (such as: analgesics, antibiotics, anticoagulants, antidepressants, anticarcinogens, anticonvulsants, neuroleptics, antiviral drugs, sedatives, stimulants, diabetes medications, or cardiovascular drugs).

[0097] Preferably, the step of prescreening includes projecting each raw optical spectrum on a prescreening latent space (e.g., by performing dimensionality reduction techniques, like principal component analysis or embedding methods). In this case, the processing unit 6 is configured to determine that the raw optical spectrum is associated to a patient having an atypical status if the projected raw optical spectrum belongs to a predetermined region of the prescreening latent space, associated to said atypical status.

[0098] Alternatively, or in addition, the step of prescreening comprises determining whether, for a predetermined set of wavelengths, the amplitude of the raw optical spectrum is higher than a respective threshold. In this case, the processing unit 6 is configured to determine that the raw optical spectrum is associated to a patient having an atypical status if, for at least part of the predetermined set of wavelengths, the amplitude of the raw optical spectrum is higher that the respective threshold.

[0099] This latter aspect of the invention results from the fact that, due to the presence of exogenous chemical components in the urine sample, a plurality of fluorescence peaks is observed in the raw optical spectrum, for instance, in the [500 nm; 1100 nm] wavelength range. This is, for instance, the case if the patient suffers from porphyria.

[0100] Preferably, the processing unit 6 is also configured to so that it does not perform steps 20, 30 if the patient associated to the current urine sample is found to have an atypical status.

[0101] The processing unit 6 is, preferably, further configured to generate an alert signal if the patient associated to the current urine sample is found to have an atypical status.

[0102] Operation of the urinalysis system 2 will now be described. [0103] The processing unit 6 receives a raw optical spectrum of the urine sample of a patient. As mentioned previously, the raw optical spectrum includes a plurality of couples, each couple comprising a first wavelength and a corresponding measured light intensity.

[0104] Advantageously, the processing unit 6 corrects the received raw optical spectrum based on a temperature of the urine sample.

[0105] Then, during an optional step of prescreening, the processing unit 6 performs data analysis on each raw optical spectrum to determine, based on the features of the raw optical spectrum, whether the urine sample is associated to a patient having a predetermined atypical status. If the patient has an atypical status, the processing unit 6 preferably generates an alert signal and does not perform further processing of the raw optical spectrum.

[0106] Then, during the determination step 20, and for each biomarker, the processing unit 6 determines a corresponding high-resolution optical spectrum based on the raw optical spectrum. Each high-resolution optical spectrum includes a plurality of couples, each couple comprising a second wavelength and a corresponding computed light intensity. A second wavelength interval between two successive second wavelengths is smaller, for instance at least ten times smaller, than a first wavelength interval between two successive first wavelengths of the raw optical spectrum.

[0107] In a specific configuration, in order to determine the high-resolution optical spectrum associated to a given biomarker, the processing unit 6 performs a spectral reconstruction algorithm on the corresponding raw optical spectrum.

[0108] In an alternative and preferred configuration, in order to determine the high- resolution optical spectrum associated to a given biomarker, the processing unit 6 identifies, among a plurality of stored base high-resolution optical spectra, the base high- resolution optical spectrum that is the closest to said raw optical spectrum over the relevant wavelength range with regard to a predetermined metric. [0109] Advantageously, the processing unit 6 also performs a step of data clustering, based on at least one raw optical spectrum, to assign the urine sample to one of a plurality of predetermined patient classes.

[0110] Then, during the estimation step 30, for each biomarker, the processing unit 6 estimates a respective value based at least on the determined high-resolution optical spectrum. More precisely, to estimate a value of a given biomarker, the processing unit 6 implements at least one corresponding biomarker model to estimate said value. Preferably, each biomarker model is selected based on the assigned patient class.

[0111] Advantageously, for at least one biomarker, the processing unit 6 is further configured to determine the corresponding value based on an electroanalytical measurement, preferably a conductivity measurement of the urine sample (which has advantageously also been corrected based on a temperature of the urine sample). In this case, the electroanalytical measurement was performed by the electroanalytical probe 4 and the resulting signal was received by the processing unit 6. [0112] As a result, the presence or absence of each biomarker in the urine sample, and/or the value, i.e. concentration or concentration range of each biomarker is estimated.