JP4708839B2 - Biological information analyzer - Google Patents

Description

  The present invention relates to a biological information analysis apparatus, and more particularly to an apparatus that quantitatively analyzes a feature amount of biological information.

  The biological information includes a reception signal obtained by ultrasonic diagnosis, a measurement signal obtained by X-ray measurement, an electrocardiogram signal output from an electrocardiograph, and the like. Although it is possible to analyze biological information as it is and evaluate it, such analysis may make it difficult to sufficiently identify the difference in properties between tissues. Therefore, it is desirable to skillfully extract the feature values of the biological information so that the tissue properties can be appropriately evaluated.

  Patent Document 1 discloses an apparatus that divides time-series data into characteristic fluctuation units and categorizes them and accumulates them together with data generation time, thereby facilitating data retrieval. Patent Document 2 discloses an apparatus for analyzing biological information.

JP-A-9-34719 JP 2003-235814 A

  For example, there is a demand for more easily obtaining feature quantities for a large amount of data such as received signals obtained by ultrasonic diagnosis, but conventional correlation analysis or multivariate analysis in the time domain, analysis in the frequency domain, etc. In some cases, it is difficult to evaluate a data set composed of a large number of data itself or a difference between data sets.

  An object of the present invention is to provide a biological information analysis apparatus capable of remarkably expressing a difference in tissue properties with respect to biological information.

  The present invention provides an extraction means for extracting a plurality of time-series data in biological information obtained from a living body, and classifying individual feature amounts obtained from the respective time-series data, thereby obtaining a plurality of the plurality of time-series data. Symbol data conversion means for converting to the symbol data, symbol data string constituting means for constituting the symbol data string comprising the plurality of symbol data, and analyzing the symbol data string, thereby providing an overall feature amount for the biological information And an overall feature amount calculating means for obtaining

  In the above configuration, biological information is extracted from the living body. Particularly preferably, echo data as biological information is acquired by transmitting and receiving ultrasonic waves. Then, the following processing is performed on a plurality of time-series data constituting all or part of the biological information. First, each time series data is converted into symbol data based on the individual feature amount. In that case, it is determined which of the plurality of patterns the individual feature amount corresponds to. That is, classification or typification processing is performed on each time series data. Here, the symbol data corresponds to an identifier or a symbol representing each pattern. A symbol data string is composed of a plurality of symbol data (sequence) corresponding to the plurality of time-series data (sequence). The symbol data string is preferably a one-dimensional data string, but may be a two-dimensional or three-dimensional data string. By analyzing the symbol data string, the feature amount (overall feature amount) possessed by the symbol data string is obtained, and this is used as the evaluation value of the biological information. As the pattern classification method and the feature amount extraction method, various methods can be used based on the properties of the target biological information.

  According to the above configuration, the overall feature amount of the biological information can be obtained by comprehensively considering individual feature amounts for each time series data. In addition, it is possible to easily and appropriately obtain the entire feature amount included in the biological information by stepwise feature amount analysis. Therefore, for example, it is possible to obtain an evaluation value capable of remarkably identifying a difference in tissue properties.

  Preferably, the symbol data converting means is a return map creating means for creating a plurality of return maps based on the plurality of time series data, and a means for generating a plurality of symbol data from the plurality of return maps, Return map classification means for generating the symbol data by classifying the distribution state of the point set as the individual feature amount in each return map. Various types of return maps are known. For example, by mapping a point indicating a combination of two values to coordinates specified by fitting two adjacent values to two axes, respectively. A return map can be created. According to the return map, it is possible to recognize the feature amount of the signal as the basis as a spatial distribution. A process for typifying the spatial distribution is executed.

  Preferably, the symbol data converting means further includes a preprocessing means for analyzing the waveform of each time series data and converting the time series data into each discrete number sequence data, and the return map creating means includes each of the discrete number sequence data from the respective discrete number sequence data. Create a return map. For example, when the time series data is an RF reception signal, it is possible to create a return map from the signal as it is, but it is desirable to create the return map after extracting the waveform characteristics of the time series data. . Preferably, the preprocessing means obtains each discrete number sequence data by extreme value extraction processing or zero cross interval extraction processing for each time series data.

  Preferably, the return map creating means creates each return map by mapping a plurality of points on a predetermined coordinate system based on a plurality of values constituting each discrete sequence data. Preferably, the return map classification means includes a distance calculation means for calculating a distance from a coordinate origin for each point constituting the point set, a histogram creation means for creating a histogram based on the distance for each point, Histogram analysis means for analyzing the histogram, and pattern determination means for determining a pattern corresponding to the analysis result of the histogram from a plurality of patterns, thereby obtaining the symbol data. Preferably, the histogram analysis unit obtains a plurality of analysis values regarding the form of the histogram, and the pattern determination unit is configured as a fuzzy inference unit having a plurality of membership functions corresponding to the plurality of analysis values.

  Preferably, the overall feature amount calculating means calculates the overall feature amount by analyzing a configuration of the symbol data string. Preferably, the biological information is information obtained from a one-dimensional region, a two-dimensional region, or a three-dimensional region in the living body by transmitting and receiving ultrasonic waves. Preferably, the site where the ultrasonic wave is transmitted and received is the liver, and the overall feature value correlates with the fibrosis index of the liver.

  The present invention also provides an extracting means for extracting a plurality of time-series data in biological information obtained from a living body, and classifying individual feature amounts obtained from the respective time-series data, thereby obtaining the plurality of time-series data. A symbol data conversion means for converting the symbol data into a plurality of symbol data, a symbol data string constituting means for constituting a symbol data string composed of the plurality of symbol data, a display process for each symbol data, and the symbol data string as a grayscale image Or display conversion means for converting into a color image. According to the above configuration, it is possible to recognize the entire feature amount of the biological information as an image that visually represents the symbol data string. When the image is a gray image, the overall feature quantity can be recognized as a gray distribution (light density pattern), and when the image is a color image, the overall feature quantity can be recognized as a hue distribution (hue pattern). However, when it is desired to perform quantitative evaluation, it is desirable to analyze the symbol data string and convert the overall feature value into some numerical value, that is, an evaluation value. When forming a grayscale image, each symbol data type is associated with a specific luminance (luminance mapping process), and when forming a color image, each symbol data type is associated with a specific hue. (Hue mapping process).

  As described above, according to the present invention, differences in tissue properties can be remarkably expressed in biological information.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described with reference to the drawings.

  First, the principle of this embodiment will be described with reference to FIG. Reference numeral 200 represents a scanning surface formed by electronic scanning of an ultrasonic beam. This scanning plane 200 corresponds to a two-dimensional tomographic image as a B-mode image. However, the present invention can also be applied to one-dimensional images or three-dimensional images. The present invention is particularly preferably applied to ultrasonic reception signals, but can also be applied to other biological information. On the scan plane 200, a region of interest (ROI) 202 is set by the user or automatically in the example shown in FIG. The region of interest 202 corresponds to an area for extracting biological information to be analyzed. In the example shown in FIG. 12, the region of interest 202 is set as a constant range in the r direction and a constant range in the θ direction, and the region of interest 202 is a fan-shaped region having a substantially trapezoidal shape as a whole. However, the shape of the region of interest 202 is not limited thereto, and may be a rectangle or a circle. A three-dimensional shape may also be used. If the region of interest 202 as shown in FIG. 12 is set, the region of interest 202 is configured as a set of a plurality of echo data 202A having the same length, and according to such conditions, the subsequent processing is simple. It becomes. Of course, the entire scanning plane 200 may be the analysis target.

  Each echo data 202A is configured as a time series of a plurality of data, that is, the echo data 202A is time series data. In an embodiment of the ultrasonic diagnostic apparatus according to the present invention described later, the time-series data 202A is a reception signal (RF reception signal) corresponding to a part of the ultrasonic beam. A symbolization process (symbolic process), which will be described in detail later, is performed for each time-series data. That is, the feature amount (individual feature amount) included in each time series data is analyzed, and a symbol corresponding to the classification to which the feature amount belongs is given. Thus, one symbol data is generated for each time series data.

  A plurality of symbol data 204A corresponding to the plurality of echo data 202A constitutes a symbol data string 204 as an entire arrangement thereof. Here, the arrangement of the plurality of symbol data 204A corresponds to the arrangement of the plurality of echo data 202A in the θ direction in the example shown in FIG.

  A feature amount (overall feature amount) included in the symbol data string 204 is extracted, and the feature amount is used as an evaluation value 206. Such an evaluation value 206 can be used, for example, as an index for evaluating the properties of a living tissue such as the liver. In an example described later, the degree of fibrosis of the liver is objectively determined depending on the size of the evaluation value 206. It becomes possible to judge. In other words, in the present embodiment, unique information regarding tissue properties that is not necessarily clarified when the biological information included in the region of interest 202 is directly analyzed is extracted as individual features and overall features. This is obtained by the stepwise process. In that case, based on a known symbolic technique, it is further developed. As described above, it is particularly desirable to analyze a two-dimensional image or information corresponding thereto, but it is of course possible to process three-dimensional information or the like. In the example illustrated in FIG. 12, for example, a two-dimensional tomographic image of the liver is acquired, the region of interest 202 is set as a portion of interest in the liver, and information within a range defined by the region of interest 202 is processed. Become. When extracting a plurality of time-series data from the region of interest 202, data extraction processing can be performed along an arbitrary direction. However, if data extraction processing or data extraction processing is performed along an ultrasonic beam, the data extraction processing can be easily performed. There is an advantage that an operation can be executed.

  Next, the configuration of the ultrasonic diagnostic apparatus according to the present embodiment will be described based on FIG. This ultrasonic diagnostic apparatus is used in the medical field, for example, to image a living tissue such as the liver and objectively evaluate the properties of the living tissue as a numerical value.

  The probe 10 is an ultrasonic probe, and ultrasonic waves are transmitted and received by the probe 10. Specifically, the probe 10 is provided with an array transducer including a plurality of transducer elements. An ultrasonic beam is formed by the array transducer, and the ultrasonic beam is scanned electronically. As the electronic scanning method, electronic sector scanning, electronic linear scanning, and the like are well known. The array transducer may be a 1D array transducer or a 2D array transducer. FIG. 1 shows a state in which ultrasonic waves are being transmitted and received by the probe 10 to the living tissue 11. The probe 10 is used in contact with the surface of a living body or inserted into a body cavity of a living body. The living tissue 11 is, for example, a liver.

  The transmission / reception unit 12 functions as a transmission beam former and a reception beam former. That is, the transmission / reception unit 12 supplies a plurality of transmission signals to the array transducer. As a result, a transmission beam is formed in the array transducer. On the other hand, a plurality of reception signals output from the array transducer are input to the transmission / reception unit 12, and a phasing addition process is performed on the plurality of reception signals. As a result, a reception beam is electronically formed. Incidentally, an ultrasonic beam is generally recognized as a combination of a transmission beam and a reception beam. It is also possible to apply a technique for simultaneously forming a plurality of reception beams per transmission beam.

  In addition to being output to the tomographic image processing unit 14, the received signal (RF received signal) after the phasing addition output from the transmission / reception unit 12 is input to the symbolic dynamics processing unit 22 via the switch 18 in the present embodiment. The First, the formation of tomographic images will be described. The tomographic image processing unit 14 includes modules such as a detector, a logarithmic compressor, a noise processing circuit, a coordinate conversion circuit, and an interpolation circuit. A mode tomographic image is formed. In the present embodiment, the tomographic image processing unit 14 includes a so-called DSC (digital scan converter). Of course, you may make it provide the circuit which forms a two-dimensional blood-flow image and another image based on a received signal. Image data output from the tomographic image processing unit 14 is output to the display processing unit 16. The display processing unit 16 will be described in detail later. The display processing unit 16 has a function of displaying an evaluation value as an overall feature amount calculated by a method described below together with a tomographic image. . A display unit 26 is connected to the display processing unit 16.

  On the other hand, the reception signal output from the transmission / reception unit 12 is input to the symbolic dynamics processing unit 22 via the switch 18. As conceptually shown in FIG. 1, the switch 18 outputs an input received signal to the a terminal or the b terminal. Here, when the entire scanning plane is the processing target, the switch 18 outputs the received signal to the symbolic dynamics processing unit 22 via the a terminal. In this case, the received signal obtained for each received beam is input to the symbolic dynamics processing unit 22 as time series data. On the other hand, when a region of interest is set on the scanning plane, the switch 18 outputs an input received signal to the division processing unit 20 via the b terminal in order to extract data in the region of interest. The division processing unit 20 cuts out a signal corresponding to the region of interest in one received signal in this embodiment, and outputs the partial signal to the symbolic dynamics processing unit 22. In this case, each segmented portion corresponds to time-series data.

  The symbolic dynamics processing unit 22 performs conversion into symbol data for each time-series data that is sequentially input, as described with reference to FIG. 12, and further, the symbol data as the set or the entire sequence thereof An evaluation value is calculated by obtaining a feature value after forming a column. This will be described in detail later with reference to FIG. The calculated evaluation value, that is, the overall feature amount is sent to the display processing unit 16. When the above-mentioned region of interest is set when displaying the tomographic image, the display processing unit 16 synthesizes a frame indicating the region of interest as a graphic image on the ultrasound image and evaluates it adjacent to the ultrasound image. Combining values as numerical data. In this case, if a plurality of regions of interest are set on the scanning plane and an evaluation value is obtained for each region of interest, it is also possible to display each evaluation value overlaid on the tomographic image in association with the hue, for example. It is. The display unit 26 displays an image generated by the display processing unit 16 or synthesized.

  The operation panel 24 includes a keyboard, a trackball, and the like, and a user can set processing conditions and a region of interest using the operation panel 24. The symbolic dynamics processing unit 22 shown in FIG. 1 is preferably configured as a program that runs on the CPU. For example, an RF reception signal can be taken out from an existing ultrasonic diagnostic apparatus and taken into a personal computer, symbolic dynamics processing can be executed on the personal computer, and the execution result can be displayed on a display on the personal computer. The processing result can be returned to the ultrasonic diagnostic apparatus and displayed together with the ultrasonic image.

  FIG. 2 shows a specific configuration example of the symbolic dynamics processing unit 22 shown in FIG. Each time series data described above is sequentially input to the time series data input unit 30. Each time-series data corresponds to a reception signal for one ultrasonic beam or a signal portion partially extracted from one ultrasonic beam. Each time series data is input to the pattern classification processing unit 32.

  A specific configuration of the pattern classification processing unit 32 will be described later with reference to FIG. 3. In the present embodiment, the pattern classification processing unit 32 converts the time series data into discrete sequence data for each time series data. By constructing a return map and evaluating the distribution of the point set in the return map using fuzzy reasoning, symbol data corresponding to each discrete sequence data is finally obtained.

  The symbol string construction unit 34 constructs a symbol data string by aligning symbol data generated for each time series data. The processing contents will be described later with reference to FIGS. The feature amount extraction unit 36 analyzes the feature amount (overall feature amount) included in the symbol data string as will be described later with reference to FIG. The overall feature amount is output as an evaluation value to the display processing unit 16 via the analysis result output unit 38. It is also possible to display the black and white grayscale image or color image on the screen by graphically expressing the symbol data string by giving brightness or hue to each symbol data constituting the symbol data string. This will be described later with reference to FIGS.

  FIG. 3 shows a specific configuration example of the pattern classification processing unit 32 shown in FIG. As will be described later with reference to FIG. 4, the discrete number sequence processing unit 40 executes preprocessing for converting input time-series data into discrete number sequence data so that the feature amount is reflected. The return map construction unit 42 constructs a return map by performing a mapping process on a predetermined coordinate system based on the discrete sequence data. The fuzzy inference unit 44 classifies the return map into one of the patterns based on a plurality of analysis values or feature amounts obtained by analyzing the return map, as will be described later with reference to FIGS. . When a pattern is determined by such classification processing, symbol data corresponding to the pattern is determined. The processing is performed in the symbolization processing unit 46.

  Next, the operation of each component shown in FIGS. 2 to 3 will be described in more detail.

  In FIG. 4, (A) shows time series data 100. As described above, the time series data 100 is an RF reception signal in the present embodiment. As described above, discrete number sequence data is generated as preprocessing for each time series data 100. In that case, feature values such as each extreme value or each zero cross point interval in the time series data 100 are referred to. In (A), an array of extreme values is shown as reference numeral 102. If a return map, which will be described later, is generated from the RF reception signal itself, the feature amount included in the RF reception signal cannot be remarkably expressed. However, according to such a technique, the characteristics of the RF reception signal can be expressed very clearly. It can be reconfigured as a dimensional distribution display.

Here, the normalization process is executed so that the minimum value X _min becomes 0 with respect to the discrete number sequence data {X ₁ , X ₂ , X ₃ ,... X _i , X _{i + 1 ,.} The That is, prior to mapping each discrete number sequence data, a process of matching the coordinate system is performed in order to compare each feature quantity with each other.

FIG. 4B shows an example of the return map 104. In the present embodiment, the return map 104 is a map in which points are plotted on coordinates corresponding to a set of two values X _i and X _{i + 1} that are adjacent in time in discrete sequence data. That is, point mapping is performed for every two adjacent values (pairs) in the discrete sequence data. When such mapping is performed, it has been confirmed that the point set spreads with a specific distribution on the return map 104 according to the properties of the living tissue. However, although it is possible to visually recognize or identify such a tendency of the point set spread, quantification cannot be performed as it is. Therefore, as will be described below, the content of the return map is analyzed in the fuzzy inference unit 44 shown in FIG.

  The fuzzy inference unit 44 in FIG. 3 has a function of analyzing a return map to create a histogram, a function of analyzing the histogram to obtain variances and peak values, etc. And a function of generating symbol data corresponding to the determined classification. The specific contents will be described in detail below.

  In order to classify the distribution degree of the point set on the return map, first, the distance from the coordinate origin is obtained for each point on the return map, and a histogram representing the distance for each point is created. A specific example of this will be described later with reference to FIG.

For the histogram, the variance is first calculated. Here, when the distance obtained for each point on the return map is r _i (i = 1 to n: n is the number of data) and the average of the distance is r with a bar, the variance σ is defined as follows: Is done.

  As for the histogram, it is detected how many constant peaks there are when viewed as a whole. And a peak value is calculated | required as needed. Here, in the present embodiment, when the values of the histogram are sequentially referred from the origin side along the horizontal axis in the histogram, the peak value of the peak currently obtained is higher than the peak value of the peak obtained in the past. When the peak value is small, the peak value of the peak obtained in the past is invalidated, and the peak value of the peak currently obtained is validated. That is, attention is paid to a peak at a distance far from the origin, and the peak value is used in fuzzy inference.

As described above, when the variance and the number of peaks are specified for each return map, that is, each histogram, each membership value μ _S , μ _L , μ _T is used by using the following three membership functions. Is required.

When the number of peaks in the histogram is one and the variance is small, the following first membership function is applied.

  In the above, α is a coefficient that determines the inclination of the membership function, A indicates the inflection point of the inclination, and Z indicates the variance of each data. For the RF reception signal, for example, α = 0.15 and A = 50.

Further, when the number of peaks in the histogram is one and the variance is large, the following second membership function is used.

  In the above, β is a coefficient that determines the inclination of the membership function, B indicates the inflection point of the inclination, and Z indicates the variance of each data. For example, β = 0.15 and B = 50 for the RF reception signal.

Further, when the number of peaks in the histogram is two (or more), a membership function according to Table 1 is defined as follows.

  However, the peak value is determined as described above. In applying the three membership functions, in the present embodiment, the first membership function is applied first, and the other two membership functions are applied as necessary.

  FIG. 5 shows the three membership functions described above in (A) to (C). The first membership function (A) and the second membership function (B) have an inverted relationship with each other, and one can be substituted for the other.

When each membership value is obtained as described above, the histogram is classified into one of types A to D according to the following conditions.

  FIG. 6 shows the above conditions. Here, Condition 1 is shown in (A), Condition 2 is shown in (B), and Condition 1 is determined first as described above. Will be.

  As described above, it is possible to classify each histogram into any pattern according to its characteristics using fuzzy inference. Symbol A is given for the target time series data. Similarly, when type B is determined, symbol B is given. When type C is determined, symbol C is given, and type D Is determined, the symbol D is given. The symbol assignment is performed by the symbolization processing unit 46 shown in FIG. 3, and symbol data representing the given symbol is output from the symbolization processing unit 46.

  In this embodiment, the distribution state of the point set in the return map is classified into a plurality of patterns according to the above conditions, but the classification method is not limited to the above, and is reflected on the return map. It is desirable to appropriately determine the classification conditions so that appropriate classification, that is, symbolicization is performed according to the distribution state corresponding to the properties of the tissue. For example, the above conditions may be corrected according to the type of biological tissue to be evaluated or according to transmission / reception conditions.

  When the flow up to this point is organized using FIG. 7, a plurality of return maps are generated as shown in (B) from a plurality of time-series data as shown in (A), and each return map is further generated. Corresponding to the above, a plurality of histograms are created as shown in FIG. Then, a plurality of feature amounts representing the features of each histogram are obtained, and a symbol (that is, a pattern) representing the feature amount (individual feature amount) for each time series data as shown in (D) by fuzzy inference with respect to them. ) Will be determined. The content shown in FIG. 7 is an example for explaining the invention.

  In FIG. 2, the symbol string construction unit 34 constructs a symbol data string by aligning the symbol data sequentially output from the pattern separation processing unit 32 in the previous stage. Here, a color pattern or a gray scale pattern representing a symbol data string can be generated by giving a predetermined hue or luminance to each symbol.

  (A) in FIG. 8 shows a symbol string. Here, a color pattern in which red is assigned to the symbol A, green is assigned to the symbol B, blue is assigned to the symbol C, and white is assigned to the symbol D ( B). Further, a gray scale pattern in which dark gray is assigned to the symbol A, light gray is assigned to the symbol B, black is assigned to the symbol C, and white is assigned to the symbol D is represented by (C). In addition, if D does not generally occur as a symbol, the symbol D may be excluded from the classification target.

  FIG. 9 shows color patterns corresponding to various symbol strings. Here, (A) and (B) show a color pattern of a healthy example of the liver. (C) and (D) show color patterns in the case where chronic hepatitis has developed in the liver. As shown in the figure, the color patterns clearly differ between healthy cases and disease cases. For example, it is possible to visually diagnose a disease by displaying such color patterns or displaying them in contrast. . However, in the case of visual judgment, since quantification is difficult, in this embodiment, a symbol string or the above-described color pattern is quantitatively analyzed. The feature amount extraction unit shown in FIG. 2 performs this processing. This will be described with reference to FIG.

  In FIG. 10, (a) shows a color pattern, and various feature quantity extraction methods can be applied to such a color pattern or symbol string. Examples thereof are shown in (b) to (h).

  (B) shows the frequency distribution of each color in the color pattern shown in (a). Each frequency value can also be expressed numerically. In (c), the longest row portion for each color in the color pattern shown in (a) is extracted. For each color, the length of the column can be calculated numerically. (D) shows the average length of each color band in the color pattern shown in (a). That is, it focuses on the continuity of colors, and each length can be expressed numerically.

  In (e) to (g), a frequency distribution, that is, a histogram for each partial tone for each color in the color pattern shown in (a) is shown. For example, in the example shown in the figure, as shown in FIG. (H) shows the frequency distribution of color change between adjacent bands in the color pattern shown in (a). That is, the frequency is counted for each color change such as the transition from the symbol A to the symbol B, the transition from the symbol D to the symbol C, and the transition from the symbol C to the symbol A.

  The feature amount extraction method described above is, of course, an example, and various methods can be applied. For example, the ratio of green, that is, the amount of B on the symbol string can be quantified and used as an overall feature amount.

  The horizontal axis in FIG. 11 represents the fibrosis index. The vertical axis in FIG. 11 indicates the amount of green, that is, the ratio of the symbol B in the entire symbol string. However, the numerical values shown on the vertical axis are only relative values. As can be understood from the graph shown in FIG. 11, there is a certain correlation between the percentage of green and the fibrosis index. By the way, each point in the graph in FIG. 11 represents actual patient data, where the fibrosis index was actually identified by performing a liver biopsy, ie, tissue sampling, and the percentage of green is It was obtained by ultrasonic diagnosis using the method as described above. Therefore, if the correlation as shown in FIG. 11 is used, when the green ratio is obtained as an evaluation value for a certain patient, the fibrosis index for the patient's liver can be specified as a numerical value from the evaluation value. It is. That is, by displaying the percentage of green or other information as an overall feature quantity of biological information, or by displaying a fibrosis index as an evaluation value as shown in FIG. This has the advantage of providing useful information for diagnosis.

  According to the present embodiment, for example, the characteristics of a tissue that is difficult to recognize on a B-mode tomographic image or the characteristics of a tissue that is difficult to quantitatively observe are finally determined using a stepwise feature extraction method. It is possible to obtain an evaluation value to be expressed, and there is an advantage that information for quantitatively analyzing the properties of a living tissue can be obtained. Therefore, if such information is used, it is possible to more appropriately determine a tissue disease in a living body, and a beneficial effect in medicine can be expected.

  In the above embodiment, an example of a liver in a living body has been taken up, but it is of course possible to process other tissues. For example, an evaluation value for determining benign or malignant about a tumor can be obtained, and blood flow or the like can be targeted.

  In the above-described embodiment, numerical values or information corresponding thereto are finally displayed as images as overall feature amounts. However, as described above, the color patterns and grayscale patterns as shown in FIGS. It is also possible to display on the screen, and it is possible to recognize the tissue properties even in such display. In that case, in the above embodiment, a one-dimensional graphic pattern corresponding to a one-dimensional symbol string is displayed. However, coloring or luminance expression corresponding to each part in a two-dimensional tomographic image may be adopted. It is possible, and various other expression methods can be adopted.

1 is a block diagram showing a preferred embodiment of an ultrasonic diagnostic apparatus as a biological information analyzing apparatus according to the present invention. It is a block diagram which shows the specific structural example of the symbolic dynamics process part shown in FIG. FIG. 3 is a block diagram illustrating a specific configuration example of a pattern classification processing unit illustrated in FIG. 2. It is explanatory drawing which shows the relationship between time series data and a return map. It is a figure which shows three membership functions used by fuzzy reasoning. It is a figure which shows the relationship between two conditions and patterns used by fuzzy reasoning. It is explanatory drawing for demonstrating the flow of the pattern classification | category process which concerns on this embodiment. It is explanatory drawing which shows a symbol string and the color pattern and gray scale pattern corresponding to it. It is a figure which shows the color pattern of a healthy example and a disease example. It is a figure for demonstrating the evaluation method for extracting the feature-value from a color pattern. It is a graph which shows the correlation with the green ratio in a color pattern, and a fibrosis index. It is explanatory drawing for demonstrating the principle which concerns on embodiment.

Explanation of symbols

  10 probe, 12 transmission / reception unit, 16 display processing unit, 18 switch, 20 division processing unit, 22 symbolic dynamics processing unit, 30 time-series data input unit, 32 pattern classification processing unit, 34 symbol string construction unit, 36 feature extraction 38, analysis result output unit, 40 discrete sequence processing unit, 42 return map construction unit, 44 fuzzy inference unit, 46 symbolization processing unit, 100 time series data, 102 array of extreme values, 104 return map.

Claims (11)

An extraction means for extracting a plurality of time-series data as a plurality of RF reception signals arranged in the ultrasonic beam scanning direction by scanning an ultrasonic beam on a living body;
Symbol data conversion for converting the plurality of time-series data into a plurality of symbol data by classifying individual feature amounts obtained from a plurality of values arranged in time-series order in the depth direction constituting each time-series data Means,
A symbol data string constituting unit that arranges the plurality of symbol data in an ultrasonic beam scanning direction and constitutes a symbol data string representing the whole of the plurality of time series data ;
By analyzing the symbol data string, an overall feature amount computing means for obtaining an overall feature amount for the biological information;
A biological information analyzing apparatus comprising: The apparatus of claim 1.
The symbol data conversion means includes:
Return map creating means for creating a plurality of return maps based on the plurality of time series data;
Return map classification means for generating a plurality of symbol data from the plurality of return maps, wherein the return map classification means generates the symbol data by classifying a distribution state of the point set as the individual feature amount in each return map; ,
A biological information analyzing apparatus comprising: The apparatus of claim 2.
The symbol data conversion means further includes pre-processing means for performing waveform analysis on each of the time series data and converting it into discrete number sequence data,
The biometric information analysis device, wherein the return map creating means creates each return map from each discrete number sequence data. The apparatus of claim 3.
The biological information analyzer according to claim 1, wherein the preprocessing means obtains each discrete number sequence data by an extreme value extraction process or a zero cross interval extraction process for each time series data. The apparatus of claim 3.
The biometric information analysis apparatus, wherein the return map creating means creates each return map by mapping a plurality of points on a predetermined coordinate system based on a plurality of values constituting each discrete sequence data . The apparatus of claim 2.
The return map classification means includes
Distance calculating means for calculating the distance from the coordinate origin for each point constituting the point set;
A histogram creating means for creating a histogram based on the distance for each point;
A histogram analysis means for analyzing the histogram;
Pattern determination means for determining a pattern corresponding to the analysis result of the histogram from a plurality of patterns, thereby obtaining the symbol data;
A biological information analyzing apparatus comprising: The apparatus of claim 6.
The histogram analysis means obtains a plurality of analysis values for the form of the histogram,
The biological information analysis apparatus, wherein the pattern determination means is configured as a fuzzy inference unit having a plurality of membership functions corresponding to the plurality of analysis values. The apparatus of claim 1.
The biometric information analyzing apparatus characterized in that the global feature quantity computing means computes the global feature quantity by analyzing a configuration of the symbol data string. The apparatus of claim 1.
The biological information analysis apparatus, wherein the biological information is information obtained from a one-dimensional region, a two-dimensional region, or a three-dimensional region in a living body by transmitting and receiving ultrasonic waves. The apparatus of claim 1.
The site where the ultrasonic wave is transmitted and received is the liver,
The biological information analysis apparatus characterized in that the overall characteristic amount correlates with the fibrosis index of the liver. An extraction means for extracting a plurality of time-series data as a plurality of RF reception signals arranged in the ultrasonic beam scanning direction by scanning an ultrasonic beam on a living body;
Symbol data conversion for converting the plurality of time-series data into a plurality of symbol data by classifying individual feature amounts obtained from a plurality of values arranged in time-series order in the depth direction constituting each time-series data Means,
A symbol data string constituting unit that arranges the plurality of symbol data in an ultrasonic beam scanning direction and constitutes a symbol data string representing the whole of the plurality of time series data ;
Display conversion means for displaying each symbol data and converting the symbol data string into a grayscale image or a color image;
A biological information analyzing apparatus comprising:

Images