Method and apparatus for the analysis of properties of a material and of their variation when the material is subjected to a conversion and / or processing procedure

Abstract

A method (1) for the analysis of physical and / or chemical and / or microstructural and / or structural properties of a material (90) and of their variation, in particular of an organic material or a tissue, even more particularly when the material is subjected to a transformation and / or processing treatment or procedure that varies the properties thereof, which comprises the steps of: a) striking the material (90) with ultrasonic sound waves; b) receiving echo sound waves (12), originating from the material (90) as a result of the preceding step, which are indicative at least of the acoustic impedance of the material (90); c) converting the received echo sound waves (12) to at least one electrical signal (13) which is indicative of the echo sound waves (12); d) converting the at least one electrical signal (13) to digital data (15) which are indicative of the at least one electrical signal (13); e) processing the digital data (15) in such a way as to obtain at least one piece of information that is indicative at least of the acoustic impedance of the material (90) and is adapted to be analyzed in order to detect the variation in one or more physical and / or chemical and / or microstructural and / or structural properties affecting the acoustic impedance of the material (90).

Description

[0001] METHOD AND APPARATUS FOR THE ANALYSIS OF PROPERTIES OF A MATERIAL AND OF THEIR VARIATION WHEN THE MATERIAL IS SUBJECTED TO A CONVERSION AND / OR PROCESSING PROCEDURE

[0002] The present invention relates to a method and an apparatus for the quantitative analysis of the physical and / or chemical and / or microstructural and / or structural properties of a material and of the variation of these properties, in particular an organic material or a tissue when it is subjected to a treatment or procedure that is capable of varying such properties, for example an industrial procedure of processing and / or transforming the material.

[0003] The present invention is therefore useful and practical, particularly but not exclusively for quality control within industrial processes, for example in the food and textile sector.

[0004] The present invention, in particular the apparatus that will be described below, can also be usefully applied in the medical sector for quantitative analysis in the context of medical treatments that act on tissues (for example chemotherapeutic treatments, radiotherapy, etc.).

[0005] The physical and / or chemical and / or microstructural and / or structural properties to which reference is made are in particular characteristics of the inner structure of the material and / or of its composition, such as for example softness, density, elasticity, hardness, concentration of water or of other substances, geometry and / or distribution and / or structure of the fibers or of other substructures that compose it, etc.

[0006] The procedures to which reference is made are in particular those the purpose of which is precisely to modify one or more of these physical and / or chemical and / or microstructural and / or structural properties of the material.

[0007] Non-limiting examples of this type of procedure are procedures in the food industry, such as cooking, desiccation, etc., and in the textile and tanning industry, such as chemical / physical functionalizing treatments, etc.

[0008] As is known, in this technical context, it is advisable, and often necessary, to perform checks on the material subjected to the industrial procedure, downstream of the method. These checks are aimed at evaluating the effect that the procedure had on the material, in order to check its efficacy, optionally correct some parameters of the process, and / or check whether the product falls within predetermined parameters, and / or in order to categorize and subdivide the products on the basis of the characteristics detected.

[0009] Nowadays, this type of check is usually done by way of invasive and destructive procedures, which involve taking samples and destroying part of the material, for example through core samples, and the subsequent analysis of the sample taken (generally by way of chemical or physical analyses performed in a laboratory externally to the processing line in which the industrial procedure is executed).

[0010] In fact, while for evaluating superficial characteristics of materials monitoring and control methods that are “inline” and non-destructive are provided, mainly based on optical readings, the same is not true when it comes to checking structural “internal” characteristics of the material (i.e. throughout its entire volume): for this type of check, in the prior art use is made of the above-mentioned destructive sampling controls which disadvantageous^ require considerable analysis time and costs, as well as holding up the entire batch being processed / produced until the results come back.

[0011] Among the checking methods available in the prior art are mass spectroscopy and chemical analyses using reagents. These methods, while giving good results, as previously mentioned, have the drawbacks of requiring long times and holding up the procedure, as well as being complex and costly to implement and, often, requiring equipment that is costly and complex. Furthermore, disadvantageously, the reliability of the conventional checking methods often depends on the skills and experience of the operator who implements them.

[0012] In general, the need is felt to have a method available for the quantitative analysis of the variation of the internal physical and / or chemical and / or microstructural and / or structural properties of a material subjected to a procedure, that can be executed in real time, inline, without requiring the procedure to be stopped, in a non-invasive manner, and which if possible is simple and economic to implement.

[0013] The use is also known, in ultrasound sensors, of signal processing methods and techniques, radiofrequency (RF) signal analysis, and timeseries analysis of radiofrequency ultrasound signals, feature extraction, feature selection and machine learning and / or techniques based on neural networks. However, in this type of conventional ultrasound sensors, variations in acquisition methods and techniques, in transmission frequency and in signal conversion can all significantly influence the results, resulting in processes and analyses that are unreliable and erroneous. In addition, artificial intelligence (Al) models, even if superior to conventional statistical methods, need to make a compromise between generalization, flexibility and performance.

[0014] In general, as is known, a model that is more generalized will be less flexible and will have lower levels of performance. It is difficult for conventional neural network models and machine learning models to maintain high levels of performance when the input is generalized, as described at length in the literature.

[0015] Furthermore, use of the above-mentioned conventional architectures, directly on the ultrasonic signal, which is subject to variations and noise, constitutes an obstacle to the training of the specific model, resulting in, among other things, an intrinsic risk of overfitting.

[0016] According to the known art, artificial intelligence acts directly on the input signal and therefore also on the noise. This causes noise and error to be propagated along the chain to the end result.

[0017] The aim of the present invention consists of providing a method for the analysis of the variation of physical and / or chemical and / or microstructural and / or structural properties of a material subjected to a procedure that varies these characteristics of the material, that solves the above technical problem, compensates for the above-mentioned drawbacks and overcomes the limitations of the prior art by making it possible to execute a check in real time and in a non-invasive manner.

[0018] Within this aim, an object of the present invention is to make available an analysis method of the mentioned type that does not require the procedure underway to be held up or the destruction of the material being analyzed.

[0019] Another object of the invention consists in providing an analysis method of the mentioned type that is simple and rapid to implement.

[0020] Another object of the invention consists in making available an analysis method of the mentioned type that can be executed in the absence of operators, therefore autonomously and without the presence of humans.

[0021] Another object of the invention consists in providing an analysis method of the mentioned type that offers reduced costs and if possible can be provided with an apparatus that is simple and economic to provide.

[0022] Another object of the invention consists in providing an analysis method of the mentioned type that is reliable.

[0023] Another object of the invention is to make available an apparatus that makes it possible to provide such a method.

[0024] Not least an object of the invention is to make available a valid alternative to the known art.

[0025] This aim and these and other objects which will become better apparent hereinafter are all achieved by a method according to claim 1.

[0026] This aim and these and other objects which will become better apparent hereinafter are also achieved by an apparatus according to claim 9.

[0027] Further characteristics and advantages of the invention will become more apparent from the description of some preferred, but not exclusive, embodiments of a method and of an apparatus according to the invention, which are illustrated by way of non-limiting example with the aid of the accompanying drawings wherein:

[0028] Figure 1 is a schematic block diagram of an embodiment of the analysis method provided by way of an apparatus according to the invention;

[0029] Figure 2 is a flowchart showing the steps of a preferred, but not exclusive, embodiment of the analysis method;

[0030] Figure 3 is a block diagram showing a quality control method of an industrial procedure in which the analysis method according to the invention is integrated;

[0031] Figure 4 is a block diagram showing the analysis method according to the invention.

[0032] With reference to the figures, the steps of a possible embodiment of the method, generally designated by the reference number 1, are shown in the flowchart of Figure 2.

[0033] The method 1 mainly serves to perform a (quantitative) analysis of the variation of one or more physical and / or chemical and / or microstructural and / or structural properties of a material 90, in particular an organic material or a tissue, when this is subjected to a processing and / or transformation treatment or procedure that varies such characteristics of the material. As will be explained more clearly below, in some embodiments, the method can optionally be used to analyze the properties of a sample of material 90, irrespective of the procedure, by performing a comparison with the predetermined standard reference values.

[0034] The analyzed material 90 can be a tissue, skin, organic tissues, or other material of animal or human origin, a plastic material, a food material, a material of plant origin, solids of various kinds in general.

[0035] As already clarified, the physical and / or chemical and / or microstructural and / or structural properties to which reference is made can be in particular characteristics of the inner structure of the material and / or of its composition, such as for example softness, density, elasticity, hardness, concentration of water or of other substances, geometry and / or distribution and / or structure of the fibers or of the layers or of other substructures that compose it, etc.

[0036] In the method 1 according to the invention, the operation is executed of striking the material 90, for example when the latter has already been subjected to the processing treatment / procedure with ultrasonic sound waves, i.e. ultrasound. Such ultrasonic waves are preferably emitted by a conventional probe 50, such as for example of the type used for digital medical ultrasound examinations, which comprises a series of transducer devices 2, 2’, 2” (such as piezoelectric crystals or other piezoelectric devices or other transducer devices, in particular MUT ultrasonic transducers, preferably pMUT or cMUT) which by vibrating, or in any case by being actuated, emit the sound waves.

[0037] The method 1 therefore entails receiving the echo sound waves 12 that are generated as a consequence of the interaction between the ultrasonic waves and the struck material 90; these echo waves 12 (which are formed by the at least partially reflected ultrasonic waves) therefore originate from the struck material 90 and are indicative at least of the acoustic impedance of such material 90.

[0038] As is known, in fact, a beam of ultrasonic waves that encounters a material 90 is reflected, absorbed, deflected and / or diffracted, in accordance with the laws common to all wave propagation phenomena. Such phenomena are influenced by the characteristics of the material 90 and in particular by acoustic impedance, which is the force with which any medium resists the passage of ultrasound and is the product of the density of the medium multiplied by the speed of propagation of the ultrasound. As is known, characteristics like those listed above (softness, density, elasticity, hardness, etc.) are closely correlated with the acoustic impedance of the material.

[0039] In particular, at the separation surfaces between mediums with different acoustic impedance (acoustic interfaces), the phenomena of reflection, refraction and scattering arise which generate echoes that, in a known manner, can for example be used for the formation of ultrasound images.

[0040] The method entails converting the received echo sound waves 12 to at least one electrical signal 13, in particular a high-frequency RF (radio frequency) signal, which is indicative of the received echo sound waves 12; this conversion can be performed, in a known manner, by way of at least one and preferably a plurality of transducer devices 2, 2’, 2” and by an electronic signal control and conversion device 60, such as for example a conventional device commonly known as a “beamformer”, to which the transducer devices 2, 2’, 2” are connected. In more detail, purely for the purposes of non-limiting example, the transducer devices 2, 2', 2" can be: piezoelectric devices (in particular piezoelectric crystals or piezoelectric micromachined ultrasonic transducers, pMUT) or other ultrasonic transducers (such as capacitive micromachined ultrasonic transducers, cMUT) or any other type of transducer useful for the purpose of converting sound waves to electrical signals and vice versa.

[0041] In the preferred embodiments, including the embodiment illustrated, the same transducer devices 2, 2’, 2”, preferably comprised in a single probe 50, perform the functions both of emitting the ultrasonic waves and of receiving the echo waves 12, and are activated cyclically to emit the ultrasonic sound waves and to receive the echo sound waves 12 according to a suitable emission-reception pattern.

[0042] In more detail, in the preferred embodiments, the conversion of the echo sound waves 12 to one or more electrical signals 13 occurs in two steps: first a plurality of analog electrical signals 3, 3’, 3” are generated, each one by way of a respective device transducer 2, 2’, 2” (for example a respective piezoelectric crystal) which interacts with at least some of the echo sound waves 12; then, starting from this plurality of analog electrical signals 3, 3’, 3”, a single electrical RF signal 13 is generated, or a plurality of signals originating from each one of the channels of the transduction stage, by means of the electronic signal control and conversion device 60, which receives the analog electrical signals 3, 3’, 3” originating from the transducer devices 2, 2’, 2”.

[0043] In order to provide this conversion, the electronic signal control and conversion device 60 preferably used comprises, in a known manner, an analog interface (analog front-end) for receiving the analog signals 3, 3’, 3”, an analog-to-digital conversion module (ADC), a digital signal processor (DSP) and optionally any other necessary electronic module known in the prior art.

[0044] According to an optimal and advantageous solution, the electronic signal control and conversion device 60 is also configured to analyze and extract data from the analog signals 3, 3’, 3” before the latter are filtered.

[0045] The electronic signal control and conversion device 60 also controls the probe 50 or in any case the transducer devices 2, 2’, 2” in order to activate them alternately to emit and receive ultrasonic waves and to orchestrate the reception of the echo waves 12, according to a predetermined pattern.

[0046] In general, the emission of ultrasonic waves, the reception of the echo waves 12 and their conversion to an electrical RF signal can be executed similarly to what is done in conventional ultrasound imaging quantitative procedures (for creating ultrasound images) performed in the medical field for ultrasound-based diagnostics and using a conventional probe and beamformer which are adapted to this end. The electrical signal 13 thus generated is therefore converted to digital data 15 that is indicative of that electrical signal 13.

[0047] In the method 1 according to the invention, this digital data 15 is then processed so as to extract at least one piece of information that is indicative at least of the acoustic impedance of the material 90.

[0048] In the preferred embodiments, this information, which is provided by processing the digital data 15, comprises, or consists of, a series of coefficients a, 0, y, ... and / or one or more digital images which are indicative at least of the acoustic impedance of the material 90.

[0049] Such information (i.e. the coefficients a, 0, y, ... and / or the digital images) is adapted to be analyzed in order to obtain the variation in one or more physical and / or chemical and / or microstructural and / or structural properties affecting the acoustic impedance of the material 90.

[0050] In more detail, according to an optimal solution, the coefficients a, 0, y, ... (and / or the digital images) are compared with a set of predetermined coefficients a’, 0’, y’, ... (or digital images) which are indicative of the aforementioned one or more physical and / or chemical and / or microstructural and / or structural properties of the material 90 when it has not yet been subjected to the processing / transformation procedure, so as to evaluate the variation of one or more of the aforementioned properties as a consequence of the procedure.

[0051] Preferably, the predetermined coefficients a’, 0’, y’, ... and / or digital images are obtained in advance in a preliminary step of calibration in which the same detection operations previously described (emission of ultrasonic waves which strike the material 90, reception of echo waves, conversion of the latter to at least one RF signal 13, extraction of data items and processing thereof) are executed on a sample of the material 90, before the latter has been subjected to the processing procedure.

[0052] Alternatively, the predetermined coefficients a’, 0’, y’ can be reference coefficients already mapped in advance on the basis of the known physical and / or chemical and / or microstructural and / or structural properties of the material (for example known from studies or previous analyses, from the literature or from suitable databases).

[0053] The extraction of the digital data from the RF signal 13 and the processing of these data (to obtain and interpret the coefficients and / or the digital images) occurs by way of a computer-based data processing apparatus 70 (such as for example a computer or an electronic controller) connected to the electronic signal control and conversion device 60 in order to receive the digital signal 13 generated by it. Conveniently, such computer-based data processing apparatus 70 is configured to execute one or more algorithms 77 which preferably perform a statistical analysis of the coefficients obtained.

[0054] In more detail, in the preferred embodiments of the method, the digital RF signals 13 supplied by the electronic signal control and conversion device 60 are processed so as to obtain a normal distribution of coefficients a, 0, y and even more preferably a statistical analysis is then executed of these coefficients a, 0, y which produces the statistical data in output, such as the average of the values of the coefficients and the variance. In the same way, the same statistical analysis can be executed in the preliminary calibration step, using the signals 13 obtained from the echo waves 12 originating from the material 90 not yet processed, and therefore the statistical data obtained before and after the procedure can be compared, as for example is shown in Figure 1.

[0055] Figure 1 shows an example of output in the form of a statistical distribution of the coefficients in a graph wherein the values of the coefficients are on the ordinate. In the graph, shown downstream of the computer-based data processing apparatus 70, two statistical distributions of coefficients can be seen: the first (indicated with the number 15’) relates to the coefficients a’, 0’, y’, ... obtained from the material 90 before it is subjected to the processing / transformation procedure and the second (indicated with the number 15) relates to the coefficients a, 0, y, ... obtained from the material 90 after it has been subjected to the procedure. Note that there are evident significant statistical differences: in the second distribution the coefficients have a higher average value and a wider distribution.

[0056] These differences are indicative of the variation of one or more properties of the material and can be used to quantitatively evaluate the variation of one or more properties of the material 90.

[0057] The data processing just described is preferably executed in a process based on artificial intelligence 77, in particular on machine learning and on neural networks. In other words, the algorithm 77 is implemented in the software program of the computer-based data processing apparatus 70 and is based on artificial intelligence algorithms, in particular machine learning and the use of neural networks.

[0058] A particularly advantageous aspect of the invention consists in that the coefficients a, 0, y, ... (or, more precisely, the statistical data obtained from them) are correlated with one or more of the physical and / or chemical and / or microstructural and / or structural properties of the material 90; this correlation can be established empirically, for example by way of a machine learning and deep learning procedure.

[0059] According to an optimal solution, the correlation between the coefficients (or, more precisely, the statistical data obtained from them) and one or more of the said above properties of the material is established, on a once-off basis, in a preliminary step of mapping, wherein the coefficients are compared with measurements of such properties performed with conventional experimental methods (for example with laboratory analysis).

[0060] It is therefore important to emphasize that it is possible to perform the comparison between the coefficients a, 0, y, ... and a’, 0’, y’, ... (or, more precisely, the statistical data obtained from them) obtained using specific algorithms, in at least two ways:

[0061] I) between before and after the processing / transformation procedure for the same sample of material 90;

[0062] II) comparison of the coefficients only before or only after the processing / transformation procedure, respectively with some reference coefficients, already mapped on reference chemical / physical and structural properties.

[0063] In the above second way, it is therefore possible to compare the target sample of material 90 with one or more standard reference values in order to analyze the properties of the sample of material 90 irrespective of the procedure (for example only before or only after the procedure).

[0064] Therefore, the method 1 and the analysis apparatus 10, according to the invention, can be used not only to analyze the variation of the properties of the material 90 after a procedure, but also to analyze the properties of a sample of material 90 by comparing it with predetermined standard reference values.

[0065] The possibility is not ruled out of combining the two ways for a more structured analysis.

[0066] Turning now in more detail to the signal processing performed in the method, this entails a kind of processing which is illustrated in detail in Figure 4 which shows the functional blocks that represent the operating steps of the process.

[0067] The signal 3 originating from the transducers is constituted by components of the ultrasonic signal, in phase I and in quadrature Q, which are stabilized in an innovative manner through direct correlation with the two-dimensional image. This is advantageously used as additional reference input and as an instrument for orienting the probe, with the function of improving the reproducibility and reliability of acquisition on the same area. Preferably a multitrack ultrasonic signal is used.

[0068] The signal 3 (thus stabilized and representable by a two-dimensional image plus a matrix of raw data) enters the first block 62 in which it is broken down and cleaned of noise. In this block 62, a first step of initializing is executed, which consists in extracting the binary data and information like the number of frames, the frame size and the sampling frequency. Subsequently, the cycle of reading and processing begins for each binary frame acquired. The frame data are processed, by applying algorithms, such as for example demodulation algorithms which are per se known. The processed data are stored in a buffer. The data are then subjected to reshaping according to specific dimensions and the sampling period is calculated. Using some parameters, such as, for example, the speed of sound and the sampling period, some time and space parameters are calculated.

[0069] Subsequently, in output from block 62 a two-dimensional synthetic image 63 is generated, deriving from the creation of two two-dimensional data matrices and a logarithmic conversion, using conventional methods.

[0070] Furthermore, in parallel, again starting from the data obtained in block 62, an additional histogram data structure 64 is generated, which is organized with intensity values along the axis of the ordinate and with signal delay values along the axis of the abscissa. Through the above data structure 64, it is possible to extract statistical parameters which are used to further reduce noise and out-of-scale data, during the subsequent processing steps. This data structure 64 is different from the two-dimensional image.

[0071] This additional data structure 64 is fed into an additional processing block 65. In this processing block 65, the additional data structure 64 is manipulated again and a matrix is obtained by applying (in a known manner) the Hilbert transform on each row of the additional data structure 64, so as to extract the overall envelope of the signal 66.

[0072] The envelope of the signal 66 is subsequently stabilized in amplitude, using conventional methods.

[0073] The two-dimensional synthetic image 63 and the overall envelope of the (stabilized) signal 66 are then fed into the processing and orchestration block 67 where they are conveniently manipulated again in order to obtain a graph composed of multiple mutually coherent signals 68 in output.

[0074] The use of multiple correlated and coherent signals makes it possible to filter the noise components, eliminating them, and to amplify the significant components of interest. This makes it possible to improve accuracy and performance over known signal processing techniques.

[0075] The multiple coherent signals 68 obtained in output are subsequently fed in input to the block 69, where they are further filtered and subjected to an operation of composition into a single signal, in accordance with conventional methods.

[0076] The signal 13 that is obtained (which corresponds to the one indicated with the number 13 in Figure 1) is a complex signal and brings useful information about the sample of material 90 analyzed, and the noise component is almost completely canceled out.

[0077] The blocks 62-69 described so far correspond to steps that can be implemented in the electronic signal control and conversion device 60. Preferably, the device 60 is connected to a dedicated computer, which executes the steps described for blocks 62 to 69.

[0078] An additional subsequent block 80 receives the complex signal 13 in input, in output from block 69, and processes it with operations described below.

[0079] As a first operation the block 80 executes a Fourier transform on the complex signal 13, in order to obtain the frequency domain spectrum.

[0080] As a second operation, the amplitude of the complex signal 13 is calculated, as the square root of the sum of the squares of the real and imaginary components of that complex signal 13.

[0081] As a third operation, the vector parameter a is calculated as the arc tangent of the ratio between the imaginary vector component and the real vector component of the complex signal 13.

[0082] As a fourth operation, the vector parameter 0 is calculated as the arc sine of the ratio between the imaginary vector component and the real vector component of the complex signal 13.

[0083] As a fifth operation, the Fourier transform is executed on vector parameters a and 0 in order to obtain the frequency spectra as a function of the associated frequencies.

[0084] Subsequently, a three-dimensional matrix is created in which each element represents the amplitude calculated as the square root of the sum of the squares of each pair of values corresponding to each of the real and imaginary components of the complex signal 13. By way of non-limiting example, the above operations make it possible to perform a spatial clustering that can, by way of non-limiting example, take advantage of a known spatial clustering algorithm (known as K-means) on the 3D matrix in order to identify groups of points with similar amplitudes, supplying information on the spatial distribution of the structures in the signal. Finally, an optically readable / analyzable chart 82 is generated in output from block 80, and more precisely a two-dimensional bar chart 82. The two- dimensional bar chart 82 contains key information on the underlying structure of the complex signal 13 in input to the block 80.

[0085] In brief, this method advantageously combines a Fourier analysis in the frequency domain with spatial clustering in order to obtain a fuller understanding of the structure of the signal than conventional methodologies and techniques.

[0086] Advantageously, the bar chart 82 is optically readable, using known optical and computer vision techniques.

[0087] Advantageously, the application of neural networks and techniques of machine learning and deep learning, i.e. the artificial intelligence process block 77, occurs exclusively starting from this step. The advantage deriving from this approach consists of a richer information content and significantly reduced error and noise components. This makes it possible to overcome the compromise, known in the state of the art, between performance and flexibility. The use of multiple input signals 68, as described, makes it possible to filter out the noise components and amplify the useful components of interest.

[0088] The resulting two-dimensional bar chart 82 is less influenced by error and noise components. This two-dimensional bar chart 82 is then used as input to the artificial intelligence algorithm 77, improving its precision and performance, with superior outcomes to the known art in which algorithms are applied on the signal 3 containing noise.

[0089] The method also has a differential reading mode which consists of comparing the two-dimensional bar chart 82 with one or more other two- dimensional bar charts 82' preferably obtained from the same material 90, in steps prior to the processing procedure; this enables the differential analysis of the properties of the materials before and after treatments or transformations of the physical and / or chemical and / or microstructural and / or structural properties of a material 90.

[0090] Turning now to describe the artificial intelligence algorithm 77 in greater detail, this is configured to be a “past-informed deep neural network” especially trained to read and interpret two-dimensional bar charts 82.

[0091] This algorithm 77 (neural network) processes and interprets the two- dimensional bar chart 82 in input, also drawing on information originating from a reference two-dimensional bar chart 82'. The advantages consist of a further improvement in terms of accuracy and resolution of analysis. The neural network 77 receives the two-dimensional bar chart 82 in input, together with a reference two-dimensional bar chart 82', and generates a differential prediction through two stages 71, 72.

[0092] In the first stage 71, the neural network 77 interprets the input (the two-dimensional bar chart 82) and generates a prediction based thereon. This prediction can be represented by a function f(x) that takes as input the data item x (obtained from the two-dimensional bar chart 82) and generates a prediction y.

[0093] In the second stage 72, the neural network 77 uses the reference two- dimensional bar chart 82', supplied by the data orchestration and database block 83, to generate a new prediction y' that respects the constraints deriving from the differential reading and interpretation between the two- dimensional bar chart 82 and the reference two-dimensional bar chart 82'. This prediction can be represented by a function g(y, r) that takes as input the prediction y and the reference r (from the reference two-dimensional bar chart 82 supplied by the block 83) and generates a new prediction y'.

[0094] The function g(y, r) can be implemented in different ways and preferably by way of a convolutional neural network (CNN) architecture in order to extract the characteristics of the input (the chart 82) and of the reference chart 82' and use them to modify the prediction y.

[0095] In the event of a non-differential reading, in the absence of a reference two-dimensional bar chart 82', the orchestration block 83 deactivates block 72, which is simply skipped.

[0096] In brief, the Al algorithm 77 executes:

[0097] - a first operating stage 71 in which predictive data Y are generated as a function of the data obtained by analyzing the two-dimensional bar chart 82, which is optically readable using computer vision Al techniques;

[0098] - if in differential mode, the orchestration block 83 activates a second operating stage 72, in which the predictive data Y are modified as a function of reference data r obtained from the analysis of the reference two- dimensional bar chart 82' for use by the past-informed neural network.

[0099] Then, in a third operating stage 73, the predictive data, modified Y' in the second operating stage 72, are converted to semantic representations, in the end obtaining a classification result that is understandable and readable by an operator. In the operating stage 73, by way of non-limiting example, known methods can be used such as Softmax functions, sigmoid functions, margin maximization functions, Bayesian networks and the like. Basically, the coefficients a, 0, y to which reference is made above are the fruit of analysis of the bar chart 82, and the coefficients a’, 0’, y’ used as a reference for comparison are the fruit of analysis of the reference bar chart 82'.

[0100] This architecture described above offers several advantages. The "past-informed" approach makes it possible to separate the process of interpretation of the input from the process of generation of the differential prediction, to reduce common intrinsic effects in the original input complex signal 13, and to reduce noise. Furthermore, this architecture facilitates the design and training of neural networks, as well as the reliability and accuracy of those neural networks.

[0101] It is therefore possible to overcome the limitations of the known art of methods that use raw ultrasonic signals, radiofrequency ultrasonic signals and / or two-dimensional ultrasound images, which do not allow repeatability and are not sufficient to obtain quantitative and reliable information on the structure of the material analyzed.

[0102] The method according to the invention makes it possible to implement a quality control that is “closed-loop”, non-invasive (therefore without extraction of samples to be analyzed in a laboratory), in real time and continuous (not sample-based) on the entire batch.

[0103] In particular, in an application of undoubted industrial utility, the analysis method 1 can be used to provide a quality control and management method 100, in real time, for a processing procedure, as shown in Figure 3.

[0104] In this quality control and management method 100 for an industrial processing procedure, the material 90 is analyzed with an analysis method 1 according to the invention (for example following the steps a-e shown in Figure 2) both before and after the processing procedure. The coefficients a’, P’, y’, ... , a, 0, y, ... obtained are analyzed, subjected to statistical analysis, and compared by a data processing and control system (for example by a computer-based data processing apparatus, conveniently configured) with the methods described above.

[0105] Following this analysis and computation, in particular based on the comparison between the coefficients obtained before and after the processing procedure a’, 0’, y’, ... , a, 0, y, the data processing and control system 70 (or an operator) can subsequently automatically execute corrective actions on the processing procedure (such as correcting process parameters or stopping the procedure in order to allow corrections).

[0106] In general, therefore, information indicative at least of the acoustic impedance of the material 90 is obtained before and after the processing procedure, and this information is analyzed and compared and, on the basis of this analysis and comparison, corrective actions are optionally and subsequently executed on the processing procedure. In particular, the corrective actions are executed if a predetermined condition is highlighted by the comparison / analysis (for example if a statistical data item, such as average or variance, of the coefficients a, 0, y after the procedure shifts by a predetermined value from the same statistical data item of the coefficients a’, P’, y’ obtained before the procedure).

[0107] The present invention also relates to an analysis apparatus 10 which is configured to execute the analysis method 1 described, and this apparatus 10, as shown in Figure 1, comprises the following main components, which are mutually functionally connected:

[0108] - a probe 50 which is configured to emit ultrasonic sound waves, to receive echo sound waves 12 and to convert the received echo sound waves 12 to analog electrical signals 3, 3', 3”;

[0109] - an electronic signal control and conversion device 60 which is configured to control the probe 50 so as to induce the emission of ultrasonic sound waves and to convert the analog electrical signals 3, 3', 3" to an electrical RF signal 13 that is indicative of the echo sound waves 12;

[0110] - a computer-based data processing apparatus 70 which is configured to convert the electrical signal 13 to digital data which are indicative of the at least one electrical signal 13 and to process the digital data so as to produce a series of coefficients and / or one or more digital images that are indicative at least of the acoustic impedance of the material 90 and are adapted to be analyzed in order to obtain the variation of one or more physical and / or chemical and / or microstructural and / or structural properties that influence the acoustic impedance of the material 90 being analyzed.

[0111] The details of implementation and the operation of the analysis apparatus 10 have already been described with reference to the method.

[0112] It should be noted finally that the method and the apparatus that are the subject matter of the present invention can optionally advantageously also be used in the medical sector for quantitative analysis in the area of medical treatments that act on tissues (chemotherapy treatments, radiotherapy etc.), in particular to analyze variations in tissue properties before, during and after the treatment, by way of non-limiting example, for the characterization of tumors and formations deriving from carcinogenesis.

[0113] In practice it has been found that the method and the apparatus, according to the present invention, achieve the intended aim and objects in that they make it possible to execute checks in real time and in a non- invasive manner of variations in the structural characteristics of a material subjected to a processing procedure.

[0114] Another advantage of the method and apparatus, according to the invention, consists in that they do not require the processing procedure in progress to be stopped.

[0115] Another advantage of the method and apparatus, according to the invention, consists in that the analysis can be executed in the absence of staff, therefore autonomously and without the presence of humans, in that the analysis is based on a comparison between digital data, which can be executed autonomously by the apparatus.

[0116] Another advantage of the method, according to the invention, consists in that it is simple, rapid and economic to provide. Furthermore, the method and apparatus, according to the invention, make available a valid alternative to the known art.

[0117] The analysis apparatus and the method thus conceived are susceptible of numerous modifications and variations, all of which are within the scope of the appended claims.

[0118] Moreover, all the details may be substituted by other, technically equivalent elements.

[0119] The disclosures in Italian Patent Application No. 102023000009798 from which this application claims priority are incorporated herein by reference.

[0120] Where technical features mentioned in any claim are followed by reference signs, those reference signs have been included for the sole purpose of increasing the intelligibility of the claims and accordingly, such reference signs do not have any limiting effect on the interpretation of each element identified by way of example by such reference signs.

Claims

CLAIMS1. A method (1) for the analysis of physical and / or chemical and / or microstructural and / or structural properties of a material (90) and / or of their variation, in particular of an organic material or a tissue, even more particularly when subjected to a processing and / or transformation treatment or procedure that varies said properties, which comprises the steps of: a) striking the material (90) with ultrasonic sound waves; b) receiving echo sound waves (12), originating from said material (90) as a result of the preceding step, which are indicative at least of the acoustic impedance of said material (90); c) converting the received echo sound waves (12) to at least one electrical signal (13) which is indicative of said echo sound waves (12); d) converting said at least one electrical signal (13) to digital data (15) which are indicative of said at least one electrical signal (13); e) processing said digital data (15) in such a way as to obtain at least one piece of information that is indicative at least of the acoustic impedance of the material (90) and is adapted to be analyzed in order to detect the variation in one or more physical and / or chemical and / or microstructural and / or structural properties affecting the acoustic impedance of said material (90).

2. The method according to claim 1, wherein said at least one piece of information comprises, or consists of, a series of coefficients (a, 0, y) and / or one or more digital images that are indicative at least of the acoustic impedance of the material (90) and which are obtained from the analysis of an optically-readable graphic (82) which is generated by processing said electrical signal (13).

3. The method according to claim 2, wherein in step e) at least said coefficients (a, 0, y) are obtained, and comprising the further step of: f) comparing said coefficients (a, 0, y) with a set of predetermined coefficients (a', 0', y') which are indicative of said one or more physicaland / or chemical and / or microstructural and / or structural properties of the material (90) before it is subjected to said processing procedure, and which are obtained from a reference optically-readable chart (82'), preferably a two-dimensional bar chart.

4. The method according to claim 3, wherein said optically-readable chart is a two-dimensional bar chart.

5. The method according to claim 3, wherein said reference optically- readable chart (82'), from which said predetermined coefficients (a’, 0’, y’) are obtained, is obtained in a preliminary calibration step in which the steps a)-e) are performed on a sample of the material (20) before it is subjected to the processing procedure.

6. The method according to one or more of the preceding claims, wherein said digital data are processed by means of an artificial intelligence algorithm (77), in particular an algorithm that uses machine learning, deep learning and neural networks, which compares the obtained optically- readable chart (82) with the reference chart (82').

7. The method according to claim 6, wherein the artificial intelligence algorithm (77) comprises executing:- a first operating stage (71) in which predictive data (Y) are generated as a function of the data obtained by analyzing the optically- readable chart (82);- a second operating stage (72) in which said predictive data (Y) are modified as a function of reference data (r) obtained from the analysis of said reference chart (82');- a third operating stage (73) in which the predictive data (Y') modified in the second operating stage are converted to output information readable by an operator.

8. The method according to one or more of the preceding claims, wherein said step c) of converting the received echo sound waves (12) to one or more electrical signals comprises the steps of:cl) generating a plurality of analog electrical signals (3, 3', 3"), each one by means of a respective piezoelectric device (2, 2', 2") which interacts with at least part of said echo sound waves (12); c2) generating a single electrical signal (13) on the basis of said plurality of analog electrical signals (3, 3', 3") by means of an electronic signal control and conversion device (60); wherein said step c2) of generating a single electrical signal (13) comprises breaking down analog electrical signals, filtering out noise components and amplifying significant components of interest.

9. The method according to the preceding claim, wherein said step c2) of generating a single electrical signal (13) also comprises the steps of:- extracting binary data, processing said binary data and saving said binary data (62);- reconstructing a binary image (63) starting from said saved binary data;- generating an additional data structure (64) starting from said saved binary data;- extracting a matrix by applying the Hilbert transform on said additional data structure (54), so as to extract an overall envelope of the signal (66);- processing said two-dimensional image (63) and said overall envelope of the signal (66) so as to obtain in output a chart composed of multiple mutually coherent signals (68);- processing said multiple mutually coherent signals (68), filtering said multiple mutually coherent signals (68) and forming them into a single signal so as to obtain a single multidimensional vector representing the signal (13) that indicates said echo sound waves (12).

10. The method according to one or more of the preceding claims, wherein a same plurality of transducer devices (2, 2', 2") is activated alternately to emit said ultrasonic sound waves and to receive said echosound waves (12).

11. A quality control and management method (100) for an industrial processing procedure, which comprises the steps of:- performing an analysis method (1) according to one or more of the preceding claims on a material (90), before it is subjected to the processing procedure, in order to obtain information (a', P', y') that is indicative at least of the acoustic impedance of the material (90);- performing an analysis method (1) according to one or more of the preceding claims on the material (90), after it has been subjected to the processing procedure, in order to obtain information (a', P', y') that is indicative at least of the acoustic impedance of the material (90);- performing an analysis of said information (a', P', y', a, P, y) in order to verify at least one predetermined condition;- on the basis of said analysis, executing, in feedback, a corrective action on the processing procedure when said condition arises.

12. An analysis apparatus (10) for the execution of a method (1) according to one or more of claims 1 to 7, comprising a sound wave emission and reception probe (50) which is functionally connected to an electronic signal control and conversion device (60) which is functionally connected to a computer-based data processing apparatus (70); wherein:- said probe (50) is configured to emit ultrasonic sound waves, to receive echo sound waves (12) and to convert said received echo sound waves (12) to analog electrical signals (3, 3', 3");- said electronic signal control and conversion device (60) is configured to control said probe (50) so as to induce the emission of ultrasonic sound waves and to convert said analog electrical signals (3, 3', 3") to at least one electrical signal (13) that is indicative of said echo sound waves (12);- said computer-based data processing apparatus (70) is configured to convert said electrical signal (13) to digital data which are indicative of saidat least one electrical signal (13) and to process said digital data so as to produce a series of coefficients (a, 0, y) and / or one or more digital images that are indicative at least of the acoustic impedance of the material (90) and are adapted to be analyzed in order to obtain the variation of one or more physical and / or chemical and / or microstructural and / or structural properties that influence the acoustic impedance of said material (90).

Images