Method and computer system for predicting the stability of a sample.

Abstract

The present invention relates to a method (100) for predicting the stability of at least one sample, comprising the following steps: a. Selection (110) of at least one type of formula; b. Provision (120) of at least one list of ingredients, preferably via said input interface (210); c. Selection (130) of a first artificial intelligence model; d. Proposal (140) of at least one operating procedure; e. At each control point: i. Capture (150) of at least one image; ii. Analysis (160) of said captured image; iii. Classification (170) of said analyzed image, comprising at least one step of assigning (180) at least one first stability score; f. Prediction (190) of the stability of the sample. Fig. 1

Description

Title of the invention: Method and computer system for predicting the stability of a sample. technical field

[0001] The present invention relates to the field of artificial intelligence-assisted chemical formulation. Its application is particularly advantageous in the field of predicting the stability of a chemical sample. STATE OF THE ART

[0002] When a problem occurs during the formulation of a product, such as crystals, precipitates, solubility problems, finding the underlying cause can be a tedious task.

[0003] It is often necessary to call upon other collaborators from the R&D department, which can considerably prolong the development time.

[0004] Alternatively, it may be necessary to carry out many other tests, often to change only 1 to 3 ingredients in formulas that generally contain between ten and more than thirty ingredients. This is not cost-effective in terms of either time or raw materials.

[0005] One object of the present invention is therefore to overcome at least some of these various technical problems.

[0006] The other objects, features and advantages of the present invention will become apparent from an examination of the following description and accompanying drawings. It is understood that other advantages may be incorporated.

[0007] SUMMARY

[0008] The present invention relates to a method for predicting the stability of at least one sample, preferably comprising at least one chemical composition, preferably during the different stages of its development, said method being configured to be implemented by at least one computer prediction system, said method comprising at least the following steps: a. Selection, by at least one user, preferably via an input interface, of at least one type of formula, said type of formula being taken from at least one list of predetermined formula types, preferably said list of predetermined formula types may include at least the following formula types: i. creams, emulsions, lotions, gels and oils for the skin; ii. beauty masks; iii. foundations (liquids, pastes, powders); iv. powders for makeup, powders to be applied after bathing, powders for personal hygiene; v. solar products; vi. sunless tanning products; vii. products for whitening the skin; viii. Anti-wrinkle products. ix. shaving products (soaps, foams, lotions); x. makeup and makeup removal products; xi. products intended to be applied to the lips; xii. preparations for baths and showers (salts, foams, oils, gels); xiii. toilet soaps, deodorant soaps; xiv. dental and oral hygiene products; xv. external intimate hygiene products; xvi. Deodorants and antiperspirants: xvii. hair dyes; xviii. products for curling, straightening and fixing hair; xix. hair styling products; xx. hair cleaning products (lotions, powders, shampoos); 21. Hair care products (lotions, creams, oils); xxii. styling products (lotions, hairsprays, brilliantines); xxiii. perfumes, toilet waters and colognes; xxiv. depilatories; xxv. products for nail care and makeup; b. Provision, by said user, of at least one list of ingredients, preferably via said input interface; c. Automatic selection, preferably by at least one selection module, of a first artificial intelligence model AI1, preferably using a first convolutional neural network CNN1, from a first set of artificial intelligence models, said first artificial intelligence model AI1 being configured to match said chosen formula; d. Automatic proposal, preferably by at least one supervisory module, and advantageously using said first artificial intelligence model IAl, of at least one operating procedure adapted to said chosen formula and / or said ingredients supplied, said operating procedure comprising at least one series of control points, said control points being configured to trigger at least one image capture of at least part of said sample being prepared; e. At each checkpoint: i. Capture of at least one image, preferably by microscopy, of at least part of a sample, preferably during preparation, by at least one image capture module; ii. Analysis of said captured image, preferably in real time, by an analysis module, via a second artificial intelligence model AI2, preferably using a second convolutional neural network RNC2, so as to provide a first quality score; iii. Classification, by at least one classification module, of said image analyzed via a third artificial intelligence model AI3, preferably using a third convolutional neural network RNC3, said third artificial intelligence AI3 having been trained from at least one image database, said classification step comprising at least: • A step involving the assignment of at least one initial stability score to the sample such that: • If said first stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; • If the said first stability score is below a predetermined threshold, the said sample is classified as unstable; f. Prediction of sample stability, using at least one prediction module, by mathematically combining the initial stability scores of each control point with the initial quality scores of each control point to calculate a preliminary overall stability score.

[0009] The present invention thus enables monitoring and quality control of a sample during preparation. Indeed, by predicting the stability of a sample, the present invention makes it possible to anticipate the success or failure of the preparation. The present invention also makes it possible to improve stability during manufacturing in order to reduce the risk of failures.

[0010] According to another aspect, the present invention also relates to a computer program product, preferably stored on a non-transient memory medium, comprising instructions which, when executed by at least one of a processor, a computer, executes the method according to the present invention.

[0011] According to another aspect, the present invention also relates to a non-transient memory medium comprising at least a computer program produced according to the present invention.

[0012] According to yet another aspect, the present invention also relates to a computer prediction system configured to execute the method according to any one of the preceding claims, said computer prediction system comprising at least: a. An input interface configured to allow at least one user to: i. Choose at least one type of formula; ii. Provide at least a list of ingredients; b. A selection module configured to automatically select at least one first artificial intelligence model AI1 from a first set of artificial intelligence models, said first artificial intelligence model AI1 being configured to match said chosen formula; c. A monitoring module configured to automatically propose at least one operating procedure adapted to said chosen formula and / or supplied ingredients, preferably using said first AI1 artificial intelligence model, said operating procedure comprising at least one series of control points, said control points being configured to trigger at least one image capture; d. An image capture module configured to capture at least one image of at least part of a sample, preferably by microscopy; e. An analysis module configured to analyze at least one captured image, preferably in real time, via a second AI2 artificial intelligence model and to associate a first quality score with at least one captured image; f. A classification module configured for i. classify at least one image analyzed via a third AI3 artificial intelligence model, said third AI3 having been trained from at least one image database, said classification step comprising at least: ii. Assign at least one initial stability score to the sample such that: • If said first stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; • If the said first stability score is below a predetermined threshold, the said sample is classified as unstable; g. A prediction module configured to predict sample stability by mathematically combining the first stability scores of each control point with the first quality scores of each control point in order to calculate a preliminary overall stability score

[0013] The present invention thus enables the prediction of the stability of a sample during its manufacture, allowing intervention and therefore reducing the risk of failure. BRIEF DESCRIPTION OF FIGURES

[0014] The aims, objects, features and advantages of the invention will become clearer from the detailed description of an embodiment thereof, which is illustrated by the following accompanying drawings in which:

[0015] [Fig.1] Fig.1 represents certain steps of a prediction method according to an embodiment of the present invention.

[0016] [Fig.2] Fig.2 represents a prediction system according to an embodiment of the present invention.

[0017] [Fig.3] Fig.3 represents certain manufacturing steps of a sample according to an embodiment of the present invention.

[0018] [Fig.4] The [Fig.4] represents certain steps of a test phase according to an embodiment of the present invention.

[0019] [Fig.5] Fig.5 represents various stages of image capture and analysis of at least part of a sample according to an embodiment of the present invention.

[0020] The drawings are given by way of example and are not limiting of the invention. They constitute schematic representations of principle intended to facilitate understanding of the invention and are not necessarily to scale with practical applications. In particular, the dimensions are not representative of reality. DETAILED DESCRIPTION

[0021] Before proceeding with a detailed review of embodiments of the invention, optional features that may be used in combination or alternatively are listed below:

[0022] By way of example, the present invention comprises, after the step of predicting the stability of the sample, a stability testing phase comprising at least the following steps: a. Storage of said sample according to a first set of storage conditions, preferably using a storage device; b. Observation, by at least one observation module, preferably over a predefined period of time, of said sample so as to detect any modification of said sample, this observation step comprising at least: i. Capture, by at least said image capture module, of at least one image, preferably by microscopy, of at least part of a sample; ii. Analysis of said captured image, preferably in real time, via the second artificial intelligence model AI2, preferably using the second convolutional neural network RNC2, so as to provide a second quality score; iii. Classification, by at least said classification module, of said image analyzed via the third artificial intelligence model AI3, preferably using the third convolutional neural network RNC3, said third artificial intelligence AI3 having been trained from at least one image database, said classification step comprising at least: • A step involving the assignment of at least a second stability score to the sample such that: • If said second stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; • If said second stability score is below a predetermined threshold, said sample is classified as unstable; c. Prediction, by at least the prediction module, of the stability of the sample by mathematically combining the second stability scores of each observation with the second quality scores of each observation so as to calculate an overall stability score.

[0023] According to one example, the list of predetermined formula types includes at least: a. creams, emulsions, lotions, gels and oils for the skin; b. beauty masks; c. foundations (liquids, pastes, powders); d. powders for makeup, powders to apply after bathing, powders for personal hygiene; e. solar products; f. sunless tanning products; g. products for whitening the skin; h. anti-wrinkle products. i. shaving products (soaps, foams, lotions); j. makeup and makeup removal products; k. products intended to be applied to the lips; 1. preparations for baths and showers (salts, foams, oils, gels); m. toilet soaps, deodorant soaps; n. dental and oral hygiene products; o. external intimate hygiene products; p. deodorants and antiperspirants: q. hair dyes; r. products for curling, straightening and fixing hair; s. hair styling products; t. hair cleaning products (lotions, powders, shampoos); u. hair care products (lotions, creams, oils); v. styling products (lotions, hairsprays, brilliantines); w. perfumes, toilet waters and colognes; x. depilatory; y. products for nail care and makeup.

[0024] According to one example, the first quality score is a function of the visible state of the sample.

[0025] According to one example, the second quality score is a function of the visible state of the sample.

[0026] According to one example, the first quality score is determined using at least a first database comprising a first plurality of sample images and a first plurality of quality scores, each image in said first plurality of images being associated with a quality score in said first plurality of quality scores.

[0027] According to one example, the second quality score is determined using at least a second database comprising a second plurality of sample images and a second plurality of quality scores, each image in said second plurality of images being associated with a quality score in said second plurality of quality scores.

[0028] According to one example, the determination of the first quality score includes the selection of a sample image from said first plurality of sample images and its associated quality score, then a modification of said associated quality score according to a plurality of parameters taken from at least: the presence of artifacts, the presence of droplets, the size of the droplets, the homogeneity of the droplets, etc.

[0029] According to one example, the determination of the second quality score includes the selection of a sample image from said second plurality of sample images and its associated quality score, and then a modification of said associated quality score according to a plurality of parameters taken from at least: the presence of artifacts, the presence of droplets, the size of the droplets, the homogeneity of the droplets, etc.

[0030] According to an example, the first quality score and the second quality score are a function of the visible state of the sample.

[0031] According to one example, the first quality score and the second quality score are determined using at least one database comprising a plurality of sample images and a plurality of quality scores, each image in said plurality of images being associated with a quality score in said plurality of quality scores.

[0032] According to one example, the determination of the first quality score and the second quality score includes the selection of a sample image from said plurality of sample images and its associated quality score, and then a modification of said associated quality score according to a plurality of parameters taken from at least: the presence of artifacts, the presence of droplets, the size of the droplets, the homogeneity of the droplets, etc.

[0033] According to one example, the step of providing at least one list of ingredients includes at least one step of comparing at least one ingredient from said list of ingredients with at least one database.

[0034] According to one example, the step of providing at least one list of ingredients includes at least one step of providing said user with at least one notification of incompatibility between at least one ingredient of said list of ingredients with at least one other ingredient of said list of ingredients.

[0035] According to one example, the step of providing at least one list of ingredients includes at least one step of automatic suggestion by at least one fourth artificial intelligence model, preferably using a fourth RNC4 convolutional neural network, of at least one ingredient absent from said list of ingredients and / or at least one substitute ingredient for at least one ingredient in the list of ingredients.

[0036] By way of example, the present invention comprises: a. after the step of proposing an operating procedure, a step of modifying said operating protocol by said user so as to generate a new operating protocol; b. after the step of modifying said operating protocol, a step of updating said control points according to said new operating protocol.

[0037] According to one example, the present invention comprises, if the sample obtains a stability score below said predetermined threshold, at least one step of proposing at least one measure for restoring the stability of said sample, said measure being taken from at least: a. Adding a new ingredient; b. Substitution of an ingredient; c. Change or adaptation of the manufacturing protocol; d. Change in temperature; e. Modification of the order in which the ingredients are introduced; f. ..etc...

[0038] According to one example, the present invention comprising at least one segmentation step prior to the classification step.

[0039] According to one example, the present invention comprising, at each control point and after the analysis step of said captured image, at least one step of identification and characterization of abnormal elements (crystallization, precipitation, aggregation, poor dispersion, etc...).

[0040] According to one example, the analysis step of said captured image includes a step of providing associated formulation advice relating to the identified abnormal elements.

[0041] According to one example, the present invention comprises at least one real-time analysis step, preferably in video mode (non-freezing image), comprising: a. a step of segmenting said image; b. a step of classifying said image; c. a step of highlighting a region to attract attention the user; d. a step of assigning a label to said region, such as: crystal, precipitate, object etc;

[0042] According to one example, the present invention includes, after the step of predicting the stability of said sample, at least one step of generating at least one report, preferably by a writing module, said report including at least said stability score.

[0043] According to one example, the present invention includes an observation module configured to detect any change in said sample, preferably over a predefined period of time.

[0044] According to one example, the present invention includes a writing module configured to generate at least one report, said report including at least said stability score.

[0045] Furthermore, the following description, listing the principles, aspects, and implementations of the present invention, along with their specific examples, is intended to encompass both their structural and functional equivalents, whether currently known or developed in the future. Thus, for example, it will be understood by those skilled in the art that all the functional diagrams herein represent conceptual views of illustrative circuits incorporating the principles of the present invention. Similarly, it will be understood that all the flowcharts, and the like, represent various processes that can be substantially represented on computer-readable media and thus executed by a computer or processor, whether or not that computer or processor is explicitly shown.

[0046] The functions of the various elements shown in the figures, including any functional block referred to as a "processor" or "module," can be performed using dedicated hardware as well as hardware capable of executing software in conjunction with a computer program or appropriate instructions. When provided by a processor, the instructions can be provided by a single dedicated processor, a single shared processor, or by a plurality of individual processors, some of which may be shared. In certain embodiments of the present invention, the processor may be a general-purpose processor, such as a central processing unit (CPU), for example.Furthermore, the explicit use of the term "processor" should not be interpreted as referring exclusively to hardware capable of executing software and may implicitly include, but not be limited to, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), read-only memory (ROM) for storing software, random-access memory (RAM), and non-volatile storage. Other hardware, both conventional and / or custom, may also be included.

[0047] Software modules, or simply modules that are assumed to be software, can be represented herein as any combination of flowchart elements or other elements indicating the execution of process steps and / or a textual description. Such modules may be executed by hardware that is expressly or implicitly represented. Furthermore, it should be understood that the module may include, for example, but not limited to, computer program logic, computer program instructions, software, firmware, hardware circuits, or a combination thereof that provides the required capabilities.

[0048] According to one embodiment, the present invention relates to a method for predicting the stability of at least one sample. Preferably, this sample comprises at least one chemical composition. Advantageously, the present invention is configured to predict the stability of said sample during the various stages of its development.

[0049] Preferably, said method is configured to be implemented by at least one computer prediction system.

[0050] According to one embodiment, and as illustrated by Figures 1 and 2, said method 100 comprises at least the following steps: a. Selection 110, by at least one user, preferably via an input interface 210, of at least one type of formula; Preferably, said type of formula is taken from at least one list of predetermined formula types; Advantageously, said list of predetermined formula types may include at least the following formula types: i. creams, emulsions, lotions, gels and oils for the skin; ii. beauty masks; iii. foundations (liquids, pastes, powders); iv. powders for makeup, powders to be applied after bathing, powders for personal hygiene; v. solar products; vi. sunless tanning products; vii. products for whitening the skin; viii. Anti-wrinkle products. ix. shaving products (soaps, foams, lotions); x. makeup and makeup removal products; xi. products intended to be applied to the lips; xii. preparations for baths and showers (salts, foams, oils, gels); xiii. toilet soaps, deodorant soaps; xiv. dental and oral hygiene products; xv. external intimate hygiene products; xvi. Deodorants and antiperspirants: xvii. hair dyes; xviii. products for curling, straightening and fixing hair; xix. hair styling products; xx. hair cleaning products (lotions, powders, shampoos); xxi. hair care products (lotions, creams, oils); xxii. styling products (lotions, hairsprays, brilliantines); xxiii. perfumes, toilet waters and colognes; xxiv. depilatories; xxv. products for nail care and makeup. b. Provision 120, by said user, of at least one list of ingredients, preferably via said input interface 210; Preferably, the step of providing at least one list of ingredients may include at least one step of comparing at least one ingredient from said ingredient list with at least one database, preferably an ingredients database; c. Selection 130, preferably automatic, advantageously by at least one selection module 220, of a first artificial intelligence model AI1 from at least a first set of artificial intelligence models. Preferably, said first intelligence model is configured to use a first convolutional neural network CNN1. Advantageously, said first artificial intelligence model AI1 is configured to correspond to said chosen formula; Indeed, advantageously, the present invention comprises a plurality of artificial intelligence models AI, each having been trained to correspond to at least one type of formula; d. Proposal 140, preferably automatic, advantageously by at least one supervisory module 230, of at least one operating procedure, preferably using said first artificial intelligence model AI1; Preferably, said operating procedure is configured to be adapted to said chosen formula and / or said ingredients supplied; said operating procedure advantageously includes at least one series of control points; Preferably, said control points are configured to trigger at least one image capture of at least a part of said sample being prepared; e. At each control point, for example T0, Tl, ..., Tfin illustrated in [Fig.3]: i. Capture 150 of at least one image, preferably by microscopy, of at least part of a sample, preferably during preparation, advantageously by at least one image capture module 240; According to one embodiment, one or more image captures can be carried out automatically and / or manually by the user; ii. Analysis 160 of said captured image, preferably in real time, advantageously by at least one analysis module 250. This analysis is advantageously carried out via a second artificial intelligence model AI2; This second model AI2 artificial intelligence preferably uses a second convolutional neural network RNC2; This analysis is configured to provide a first quality score relative to said sample; Advantageously, this first quality score is a function of the visible state of said sample, preferably of the captured part of the sample; According to one embodiment, the first quality score can be determined using at least one database comprising a plurality of sample images and a plurality of quality scores; preferably, each image in said plurality of images is associated with a quality score from said plurality of quality scores;Preferably, the determination of the quality score includes the selection of a sample image from said plurality of sample images and its associated quality score, and then a modification of said associated quality score according to a plurality of parameters taken from at least: the presence of artifacts, the presence of droplets, the size of the droplets, the homogeneity of the droplets, etc...; iii. Classification 170 of said analyzed image, preferably in real time, advantageously by at least one classification module 260. This classification is advantageously carried out via a third artificial intelligence model AI3; this third artificial intelligence model AI3 preferably uses a third convolutional neural network RNC3. In one embodiment, this third artificial intelligence model AI3 has been trained from at least one image database. Preferably, said classification step comprises at least: • A step of assigning at least one initial stability score to the sample, preferably one currently being developed, such that: • If said first stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; • If the said first stability score is below a predetermined threshold, the said sample is classified as unstable; f. Prediction 190, preferably by at least one prediction module 270, of the stability of the sample, preferably during development and / or at the end of development, by mathematically combining the first scores of stability of each control point with the first quality scores of each control point in order to calculate a preliminary overall stability score.

[0051] The present invention thus allows for real-time analysis of the stability of a sample during and at the end of processing. This results in considerable time savings.

[0052] According to a preferred embodiment, the present invention may include, after the step of predicting the stability of the sample, a stability testing phase.

[0053] Advantageously, this phase of testing the stability of the sample may include at least the following steps, as illustrated in [Fig.4] for example: a. Storage of said sample under a first set of storage conditions, preferably using a storage device; These storage conditions may further include: i. Storage temperature; ii. Storage atmosphere; iii. Humidity level; iv. ...etc... b. Observation, by at least one observation module 280, preferably over a predefined period of time, of said sample so as to detect any modification of said sample, this observation step comprising at least: i. Capture, by at least said image capture module, of at least one image, preferably by microscopy, of at least part of a sample; ii. Analysis of said captured image, preferably in real time, via the second artificial intelligence model AI2, preferably using the second convolutional neural network RNC2, so as to provide a second quality score; Advantageously, this second quality score is a function of the visible state of said sample, preferably of the captured, i.e., observed, part of the sample; In one embodiment, the second quality score can be determined using at least one database comprising a plurality of sample images and a plurality of quality scores; preferably, each image in said plurality of images is associated with a quality score from said plurality of quality scores; preferably, the determination of the quality score includes selecting a sample image from said plurality of sample images and its associated quality score, then a modification of said quality score associated with a plurality of parameters taken from at least: the presence of artifacts, the presence of droplets, the size of the droplets, the homogeneity of the droplets, etc... iii. Classification, by at least said classification module, of said image analyzed via the third artificial intelligence model AI3, preferably using the third convolutional neural network RNC3, said third artificial intelligence AI3 having been trained from at least one image database, said classification step comprising at least: • A step involving the assignment of at least a second stability score to the sample such that: • If said second stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; • If said second stability score is below a predetermined threshold, said sample is classified as unstable; c. Prediction, by at least said prediction module, of the stability of the sample by mathematically combining the second stability scores of each observation with the second quality scores of each observation so as to calculate an overall stability score.

[0054] The present invention thus makes it possible, once the sample has been made, to predict its stability over time.

[0055] According to one embodiment, said method may include, after the step of predicting the stability of said sample, at least one step of generating, preferably by at least one writing module 290, at least one report; said report may include information concerning the preparation of said sample such as the ingredients, or the protocol followed; this report may also include at least one image taken during the preparation of said sample; said report advantageously includes the stability score.

[0056] According to one embodiment, the step of providing the list of ingredients, preferably by the user, may include at least one step of comparing at least one ingredient from said ingredient list with at least one database. This comparison, or cross-referencing, thus makes it possible to check for a potential incompatibility, for example, or to suggest a replacement ingredient, for example.

[0057] Preferably, the step of providing the list of ingredients may include at least one step of providing, preferably to said user, at least one notification of incompatibility between at least one ingredient of said list of ingredients and at least one other ingredient of said list of ingredients. Advantageously, the present invention makes it possible to anticipate potential incompatibilities between the various ingredients of the list of ingredients.

[0058] Advantageously, the step of providing the list of ingredients may include at least one step of automatically suggesting at least one ingredient absent from said list of ingredients and / or at least one substitute ingredient for at least one ingredient in the list of ingredients. This suggestion step may be carried out, for example, via a fourth artificial intelligence model AI4, preferably using a fourth convolutional neural network RNC4; this automatic suggestion step is advantageously carried out by at least one suggestion module.

[0059] Thus the present invention provides, even before the start of the preparation of a sample, valuable information to the user, such as a proposal to replace an ingredient with an equivalent or better quality ingredient or one having a lower price.

[0060] Advantageously, the fourth AI4 artificial intelligence model was trained via a database comprising lists of ingredients and their characteristics as well as their uses in relation to various formulas.

[0061] According to one embodiment, the present invention also allows the user to update the operating procedure. Preferably, the method may include, after the step of proposing an operating procedure, a step in which the user modifies said operating procedure so as to generate a new operating procedure. This allows the user to make adjustments to an automatically proposed operating procedure. Advantageously, this type of adjustment is followed by a step of updating the control points. Thus, advantageously, the present invention may include, after the step of modifying said operating procedure, a step of updating said control points according to said new operating procedure.

[0062] The present invention, by providing a preliminary stability score during the preparation of a sample, allows the user to improve his sample and therefore his formulation by seeking to improve said score, preferably during the preparation of said sample.

[0063] Thus, according to one embodiment, the present invention may include, if the sample obtains a preliminary stability score below said predetermined threshold, at least one step of proposing at least one measure of restoration of the stability of said sample. Advantageously, said measure may include at least: a. Adding a new ingredient; b. Substitution of an ingredient; c. Change or adaptation of the manufacturing protocol; d. Change in temperature; e. Modification of the order in which the ingredients are introduced; f. ...etc...

[0064] In addition to assisting the user in developing a stable sample, the present invention can also provide decision-making support during sample development.

[0065] According to one embodiment, and as illustrated in [Fig.5], the present invention provides support for the development of a sample via the use of image analysis techniques, preferably in real time, and advantageously based on at least one artificial intelligence model.

[0066] Thus, for example, the present invention may include at least one segmentation step before and / or after the classification step. Figure 5 illustrates, for example, various forms of a sample, such as a microscopic image of an emulsion, an emulsion containing objects, a transparent sample, and finally, a transparent sample containing objects, by way of example. It should be noted that various types of lighting can be used (normal or polarized light). This segmentation step can graphically isolate one or more areas of the captured image corresponding to possible structural anomalies of the sample. Figure 5 illustrates, for example, the segmentation and / or classification of captured images. Subsequently, an analysis can be performed to classify the identified objects, for example.Advantageously, at each control point and after the analysis step of said captured image, the present invention may include at least one step of identifying and characterizing abnormal elements, also called the object classification step. These abnormal elements may include, for example: crystallization, precipitation, aggregation, poor dispersion, etc.

[0067] These various image analyses allow the assignment of a quality score on a sample at various times during its manufacturing phase.

[0068] In the event that one or more abnormal elements are identified by the present invention, it may include a step of providing associated formulation advice relating to the identified abnormal elements, preferably in order to eliminate these possible abnormal elements and / or to correct the formulation of said sample.

[0069] Advantageously, the present invention can also enable the user to be informed, preferably in real time, with information concerning the sample being prepared.

[0070] Thus, according to one embodiment, and as illustrated in [Fig. 5], said method may comprise at least one analysis step, preferably in real time, advantageously by the analysis module, and preferably using said second artificial intelligence model AI2. Advantageously, this analysis step may preferably be performed in video mode, that is to say, from a video, i.e., from a plurality of images, i.e., from a non-static image. This analysis step may comprise, for example: a. a segmentation step of said image, preferably by the analysis module; This segmentation step is configured to identify one or more areas of one or more images in order to identify optical differences; Preferably, this segmentation step can be carried out by conventional image analysis techniques; this segmentation step allows the identification of anomalies from an image of at least a part of a sample being prepared; b. a classification step of said image, preferably by the analysis module; This classification step is configured to classify one or more elements identified during the segmentation step; this classification can be done on the basis of a list of anomaly types, for example; This classification step is advantageously executed using said at least second artificial intelligence model AI2; c. Optionally, a step to highlight a region to attract the user's attention; This step to highlight a region can be carried out by visually delimiting a portion of said image, a portion including at least one anomaly; d. Optionally, a step of assigning a label to said region, such as: crystal, precipitate, object, etc...; This step of assigning a label thus allows the user to visually identify and spot an anomaly, for example, in the sample being prepared.

[0071] The present invention thus makes it possible, in addition to indicating to the user the stability score of the sample, to provide him with information on possible anomalies and their nature.

[0072] The present invention thus provides valuable assistance during the manufacture of a sample. Preferably, the present invention also provides valuable assistance during the stability testing phase, after the manufacturing steps.

[0073] According to one embodiment, the present invention also relates to a computer prediction system configured to execute the method according to the present invention. Said computer prediction system advantageously comprises at least one processor; said processor is preferably configured to execute a series of instructions.

[0074] According to one embodiment, said computer prediction system comprises at least: a. An input interface configured to allow at least one user to: i. Choose at least one type of formula; ii. Provide at least a list of ingredients; b. A selection module configured to automatically select at least said first AI1 artificial intelligence model from a first set of artificial intelligence models; c. A monitoring module configured to automatically propose at least one operating mode adapted to said chosen formula and / or supplied ingredients, preferably using said first artificial intelligence model AI1; d. An image capture module configured to capture at least one image of at least part of a sample, preferably by microscopy; e. An analysis module configured to analyze at least one captured image, preferably in real time, via said second AI2 artificial intelligence model and to associate said first quality score with at least one captured image; f. A classification module configured for i. Classify at least one image analyzed via said third artificial intelligence model AI3; g. A prediction module configured to predict the preliminary stability of the sample by mathematically combining the first stability scores of each control point with the first quality scores of each control point so as to calculate an overall preliminary stability score; Preferably, said prediction module is also configured to predict the stability of the sample by mathematically combining the second stability scores of each observation with the second quality scores of each observation so as to calculate an overall stability score; Advantageously, said prediction module is configured to predict the stability of the sample by mathematically combining the stability scores of each control point with the quality scores of each control point in order to calculate a stability score; h. Preferably, an observation module configured to detect any change in said sample, preferably over a predefined period of time, advantageously during a stability testing phase; i. Preferably, a writing module configured to generate at least said report including at least said stability score;

[0075] Thus, the present invention makes it possible to predict the stability of a sample at various times during its manufacture and storage.

[0076] The invention is not limited to the embodiments previously described and extends to all embodiments covered by the claims.

[0077] Numerical references

[0078] 100 Prediction Method

[0079] 110 Choice of a type of formulas

[0080] 120 Provision of a list of ingredients

[0081] 130 Selection of a first artificial intelligence model AI1

[0082] 140 Proposal for a procedure

[0083] 150 Image capture

[0084] 160 Analysis of a captured image

[0085] 170 Classification of an analyzed image

[0086] 180 Assignment of a first stability score

[0087] 190 Prediction of a preliminary stability score

[0088] 200 Prediction System

[0089] 210 Input interface

[0090] 220 Selection Module

[0091] 230 Supervisory Module

[0092] 240 Capture Module

[0093] 250 Analysis Module

[0094] 260 Classification Module

[0095] 270 Prediction Module

[0096] 280 Observation module

[0097] 290 Writing Module

Claims

1. Demands Method (100) for predicting the stability of at least one sample, preferably comprising at least one chemical composition, preferably during the different stages of its development, said method (100) being configured to be implemented by at least one computer prediction system (200), said method comprising at least the following steps: a. Choice (110), by at least one user, preferably via an input interface (210), of at least one type of formula, said type of formula being taken from at least one list of predetermined formula types; b. Provision (120), by said user, of at least one list of ingredients, preferably via said input interface (210); c. Automatic selection (130), preferably by at least one selection module (220), of a first artificial intelligence model AI1, preferably using a first convolutional neural network CNN1, from a first set of artificial intelligence models, said first artificial intelligence model AI1 being configured to match said chosen formula; d. Automatic proposal (140), preferably by at least one supervisory module (230), using said first artificial intelligence model AI1, of at least one operating procedure adapted to said chosen formula and / or said ingredients supplied, said operating procedure comprising at least one series of control points, said control points being configured to trigger at least one image capture of at least a part of said sample being prepared; e. At each checkpoint: i. Capture (150) of at least one image, preferably by microscopy, of at least part of a sample, preferably in preparation, by at least one image capture module (240); ii. Analysis (160) of said captured image, preferably in real time, by an analysis module (250), via a second artificial intelligence model AI2, preferably using a second convolutional neural network RNC2, so as to provide a first quality score; iii. Classification (170), by at least one classification module (260), of said analyzed image via a third artificial intelligence model AI3, preferably using a third convolutional neural network RNC3, said third artificial intelligence AI3 having been trained from at least one image database, said classification step comprising at least: • A step of assigning (180) at least one first stability score to said sample such that: f.If said first stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; g- If said first stability score is less than a predetermined threshold, said sample is classified as unstable; h. Prediction (190) of the stability of the sample, by at least one prediction module (270), by mathematically combining the first stability scores of each control point with the first quality scores of each control point so as to calculate a preliminary overall stability score.

2. Method (100) according to the preceding claim comprising, after the sample stability prediction step (190), a stability testing phase comprising at least the following steps: a. Storage of said sample according to a first set of storage conditions, preferably using a storage device; b. Observation, by at least one observation module (280), preferably over a predefined period of time, of said sample so as to detect any modification of said sample, this observation step comprising at least: i. Capture, by at least said image capture module, of at least one image, preferably by microscopy, of at least part of a sample; ii. Analysis of said captured image, preferably in real time, via the second artificial intelligence model AI2, preferably using the second convolutional neural network RNC2, so as to provide a second quality score; iii. Classification, by at least said classification module, of said image analyzed via the third artificial intelligence model AI3, preferably using the third convolutional neural network RNC3, said third artificial intelligence AI3 having been trained from at least one image database, said classification step comprising at least: • A step involving the assignment of at least a second stability score to the sample such that: c. If said second stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; d. If said second stability score is below a predetermined threshold, said sample is classified as unstable; e. Prediction, by at least the prediction module, of the sample stability by mathematically combining the second stability scores of each observation with the second quality scores of each observation in order to calculate an overall stability score.

3. Method (100) according to any one of the preceding claims wherein the first quality score and the second quality score are a function of the visible state of the sample.

4. Method (100) according to any one of the preceding claims wherein the first quality score and the second quality score are determined using at least one database comprising a plurality of sample images and a plurality of quality scores, each image in said plurality of images being associated with a quality score in said plurality of quality scores.

5. Method (100) according to the preceding claim wherein the determination of the first quality score and the second quality score comprises the selection of a sample image from said plurality of sample images and its associated quality score, and then a modification of said associated quality score as a function of a plurality of parameters taken from at least: the presence of artifacts, the presence of droplets, the size of the droplets, the homogeneity of the droplets, etc.

6. Method (100) according to any one of the preceding claims wherein the step of supplying (120) at least one list of ingredients includes at least one step of comparing at least one ingredient of said ingredient list with at least one database.

7. Method (100) according to any one of the preceding claims wherein the step of supplying (120) at least one list of ingredients includes at least one step of supplying said user with at least one notification of incompatibility between at least one ingredient of said list of ingredients with at least one other ingredient of said list of ingredients.

8. A method (100) according to any one of the preceding claims, wherein the step of supplying (120) at least one list of ingredients comprises at least one step of automatic suggestion by at least one fourth artificial intelligence model, preferably using a fourth RNC4 convolutional neural network, of at least one ingredient not in said list of ingredient and / or at least one substitute ingredient for at least one ingredient in the ingredient list.

9. Method (100) according to any one of the preceding claims comprising: a. after the step of proposing (140) a procedure, a step of modifying said procedure by said user so as to generate a new procedure; b. after the step of modifying said procedure, a step of updating said control points according to said new procedure.

10. Method (100) according to any one of the preceding claims comprising, if the sample obtains a stability score below said predetermined threshold, at least one step of proposing at least one measure for restoring the stability of said sample, said measure being taken from at least: a. Addition of a new ingredient; b. Substitution of an ingredient; c. Change or adaptation of the manufacturing protocol; d. Modification of the temperature; e. Modification of the order of introduction of the ingredients.

11. Method (100) according to any one of the preceding claims comprising at least one segmentation step before the classification step.

12. Method (100) according to any one of the preceding claims comprising, at each control point and after the analysis step (160) of said captured image, at least one step of identification and characterization of abnormal elements (crystallization, precipitation, aggregation, poor dispersion, etc.).

13. Method (100) according to the preceding claim wherein the analysis step (160) of said captured image includes a step of providing associated formulation advice with respect to the identified anomalous elements.

14. Method (100) according to any one of the preceding claims comprising at least one real-time analysis step, preferably in video mode (non-freezing image), comprising:

15.

16.

17.

18. a. a step of segmenting said image; b. a step of classifying said image; c. a step of highlighting a region to attract the user's attention; d. a step of assigning a label to said region, such as: crystal, precipitate, object etc; Method (100) according to any one of the preceding claims comprising, after the step of predicting the stability of said sample, at least one step of generating at least one report, by a writing module (290), said report comprising at least said stability score. Product computer program, preferably stored on a non-transient memory medium, comprising instructions which, when executed by at least one of a processor, computer, executes method (100) according to any one of the preceding claims. Non-transient memory support comprising at least one computer program product according to the preceding claim. A computer prediction system (200) configured to execute the method (100) according to any one of claims 1 to 15, said computer prediction system (200) comprising at least: a. an input interface (210) configured to allow at least one user to: i. Choose at least one type of formula; ii. Provide at least a list of ingredients; b. A selection module (220) configured to automatically select at least one first artificial intelligence model AI1 from a first set of artificial intelligence models, said first artificial intelligence model AI1 being configured to match said chosen formula; c. A supervisory module (230) configured to automatically propose at least one operating procedure adapted to said chosen formula and / or said ingredients supplied, said operating procedure comprising at least a series of points control points, said control points being configured to trigger at least one image capture; d. An image capture module (240) configured to capture at least one image of at least part of a sample, preferably by microscopy; e. An analysis module (250) configured to analyze at least one captured image, preferably in real time, via a second AI2 artificial intelligence model and to associate a first quality score with at least one captured image; f. A classification module (260) configured for i. classify at least one image analyzed via a third AI3 artificial intelligence model, said third AI3 having been trained from at least one image database, said classification step comprising at least: ii. Assign at least one initial stability score to the sample such that: g. If said first stability score is greater than or equal to a predetermined threshold, said sample is classified as stable; h. If said first stability score is less than a predetermined threshold, said sample is classified as unstable; i. A prediction module (270) configured to predict the stability of the sample by mathematically combining the first stability scores of each control point with the first quality scores of each control point so as to calculate a preliminary overall stability score.

19. A computer prediction system (200) according to the preceding claim comprising an observation module (280) configured to detect any change in said sample, preferably over a predefined period of time.

20. A computer prediction system (200) according to any one of the two preceding claims comprising a writing module (290) configured to generate at least one report, said report including at least said stability score.

Images