Data operation system for inspection, and method for adjusting, storing and analyzing data

The computer-based data management system efficiently processes and analyzes biomolecule test data using HDF5 format and statistical analysis, addressing the challenge of handling large datasets for personalized medical care and disease-specific treatment, enhancing medical care efficiency and reducing costs.

WO2024247883A1PCT designated stage expired Publication Date: 2024-12-05NAT INST FOR MATERIALS SCI
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2024/019029
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-30
Filing Date
2024-05-23
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Current systems lack efficient methods for organizing, formatting, storing, and analyzing large amounts of data from biomolecule measuring devices and in vitro diagnostic tests, hindering advanced data utilization and personalized medical care.

Method used

A computer-based data management system that includes data input, filtering, analysis, metadata generation, metadata analysis, and visualization, utilizing HDF5 format for high-speed data processing and storage, with features for abnormal data handling and statistical analysis methods like time series and correlation analysis.

Benefits of technology

Enables high-speed data management and analysis, improving medical care efficiency by providing organized and visualized data for personalized patient care and disease-specific treatment, while reducing manual processing time and personnel costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2024019029_05122024_PF_FP_ABST
    Figure JP2024019029_05122024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention provides a data operation system for inspection comprising a data input unit (11), a filtering unit (12), a data analysis / metadata generation unit (13), a metadata analysis unit (14), a metadata storage unit (15), and a display unit (16), wherein: the data input unit (11) captures inspection data; the filtering unit (12) selects and captures data determined to be valid and identifies abnormal data; the data analysis / metadata generation unit (13) analyzes the data and then generates metadata; the metadata analysis unit (14) analyzes the metadata to obtain the analyzed metadata; the metadata storage unit (15) stores the metadata and the analyzed metadata; and the display unit (16) organizes, visualizes, and displays the stored data in the metadata storage unit (15), wherein the filtering unit (12), the data analysis / metadata generation unit (13), the metadata analysis unit (14), and the metadata storage unit (15) store a file in an HDF5 format.
Need to check novelty before this filing date? Find Prior Art

Description

Inspection data management system and method for formatting, storing and analyzing data

[0001] The present invention relates to a testing data management system and a method for formatting, storing, and analyzing data. This application claims priority to Japanese Patent Application No. 2023-088702, filed on May 30, 2023, the contents of which are incorporated herein by reference.

[0002] With the demand for more advanced and efficient medical care, there is a demand for a testing data management system that organizes, formats, stores, and analyzes large amounts of data obtained from in vitro diagnostic tests and other methods, and visualizes and provides the data organized at the individual, population, case, and typological levels, such as those with the same partial DNA sequence. With this testing data management system, for example, at the individual level, it would be possible to provide medical care that is more suited to a patient's constitution and condition based on their past medical history and test data. At the population level, it would become possible to provide more capital-efficient medical care and insurance, and at the case and typological levels, it would become possible to provide medical care that is more suited to the patient's condition, such as selecting medications.

[0003] Biomolecular testing equipment is being developed for a variety of target biomolecules, but the market-standard ELISA and PCR methods are currently only used for sequential testing. Progress in making advanced use of the data obtained is slow, and it is difficult to say that data management technology is equipped to automatically collect data and analyze it from multiple perspectives. Systems that package and provide images and medical information using biological samples are disclosed in Patent Documents 1 to 3, and a test chip that detects biological samples with high sensitivity is disclosed in Patent Document 4.

[0004] Japanese Patent Publication No. 2016-184438 (A) Japanese Patent Publication No. 2018-206444 (A) Japanese Patent Publication No. 2022-105730 (A) International Publication No. 2021 / 131331 (A)

[0005] In view of the above background, the present invention aims to provide a test data management system that organizes, formats, stores, analyzes, and visualizes large amounts of data obtained through biomolecular measurement devices and in vitro diagnostic tests, and that has high-speed data read and write operations, as well as a method for organizing, storing, and analyzing data. Here, data organizing means organizing and integrating original data of different formats in accordance with standards for quantifying or symbolizing them.

[0006] The configuration of the present invention to solve the problems is shown below. (Configuration 1) A computer-based data formatting, storage, and analysis system for managing test data, the system comprising: a data input unit, a filtering unit, a data analysis and metadata generation unit, a metadata analysis unit, a metadata storage unit, and a display unit; the data input unit is configured to import test data; the filtering unit is configured to select and import data determined to be valid according to predetermined criteria from the data imported by the data input unit, and identify abnormal data; the data analysis and metadata generation unit is configured to analyze the imported data and generate metadata; the metadata analysis unit is configured to analyze the generated metadata to obtain analysis metadata; the metadata storage unit is configured to store the metadata generated by the data analysis and metadata generation unit and the analysis metadata obtained by analysis by the metadata analysis unit; the display unit is configured to have the function of extracting, organizing, visualizing, and displaying data stored in the metadata storage unit for an item selected by a user; and the filtering unit, the data analysis and metadata generation unit, the metadata analysis unit, and the metadata storage unit store files in HDF5 format. (Configuration 2) The inspection data management system of Configuration 1, further comprising an abnormal data storage unit that stores the abnormal data, and configured to have a function of displaying the abnormal data stored in the abnormal data unit on the display unit. (Configuration 3) The inspection data management system of Configuration 2, further comprising an abnormal data analysis unit that analyzes whether the abnormal data is abnormal data to be eliminated or data that is to be treated as data representing a new event, and configured to have a function of displaying details of the new event on the display unit if the abnormal data is a new event. (Configuration 4) The inspection data management system of any one of Configurations 1 to 3, wherein the metadata analysis means is statistical analysis.(Configuration 5) The test data management system according to any one of configurations 1 to 4, wherein the metadata analysis is one or more selected from the group consisting of time series analysis, regression analysis, principal component analysis, correlation analysis, factor analysis, cluster analysis, canonical analysis, t-analysis, analysis of variance, and random forest. (Configuration 6) The test data management system according to any one of configurations 1 to 5, wherein the file format of the display unit is a CSV file. (Configuration 7) The test data management system according to any one of configurations 1 to 6, wherein the test data includes personal attribute data. (Configuration 8) The test data management system according to any one of configurations 1 to 7, wherein the test data includes image data. (Configuration 9) The test data management system according to configuration 8, wherein the image data includes biomolecule fluorescence image data. (Configuration 10) The inspection data management system of Configuration 9, wherein the inspection data is biomolecule fluorescence image data, and at least one of the filtering unit and the data analysis / metadata generation unit has the following functions: automatically select an analysis region according to predetermined criteria based on the maximum value of the fluorescence intensity of the biomolecule fluorescence image data, perform smoothing processing on the fluorescence image data within the analysis region, determine a background intensity distribution within the analysis region, calculate the difference between the fluorescence intensity spatial distribution of the fluorescence image data after the smoothing processing and the intensity distribution of the background to determine the fluorescence intensity spatial distribution of the net fluorescence image data, and calculate the average intensity and full width at half maximum of the net fluorescence intensity spatial distribution. (Configuration 11) The inspection data management system of Configuration 10, wherein the metadata includes one or more selected from the group consisting of the average intensity, full width at half maximum, and difference amount of the fluorescence intensity spatial distribution, the biomolecule being the measurement target, and attribution information of the biomolecule.(Configuration 12) A method for fully automatically formatting, saving, and analyzing test data using a computer, comprising: performing data input, filtering, data analysis / metadata generation, metadata analysis, metadata storage, and display; the data input step inputs test data into the computer; the filtering step selects data determined to be valid according to predetermined criteria from the data obtained in the data input step and inputs the data into the computer, and identifies abnormal data; the data analysis / metadata generation step analyzes the input data and generates metadata; the metadata analysis step analyzes the generated metadata and generates analysis metadata; the metadata storage step stores the metadata generated in the data analysis / metadata generation step and the analysis metadata obtained by analysis by the metadata analysis means in the computer; and the display step extracts, organizes, visualizes, and displays the metadata-stored data for an item selected by a user; and the method for formatting, saving, and analyzing test data uses HDF5 format as a file format for the filtering, data analysis / metadata generation, metadata analysis, and metadata storage steps.

[0007] According to the present invention, a test data management system is provided that is a high-speed system that organizes, formats, stores, analyzes, and visualizes large amounts of data obtained through biomolecular measuring devices and in vitro diagnostic tests, as well as a method for formatting, storing, and analyzing data.

[0008] 1 is an explanatory diagram showing the overall configuration of a first inspection data management system of the present invention. FIG. 2 is an explanatory diagram showing the overall configuration of a second inspection data management system of the present invention. FIG. 3 is an explanatory diagram showing the overall configuration of a third inspection data management system of the present invention. FIG. 4 is an explanatory diagram showing a file configuration of the present invention. FIG. 5 is a flowchart showing the data creation step of the present invention. FIG. 6 is a flowchart showing the filtering image analysis step of the present invention. FIG. 7 is a flowchart showing the normal image analysis step of the present invention. FIG. 8 is a flowchart showing the visualization step of the present invention. FIG. 9 is an example of metadata creation. FIG. 10 is a bird's-eye view showing an overview of the inspection device used in the examples. FIG. 11 is image data showing fluorescent inspection acquisition example 1. FIG. 12 is image data showing fluorescent inspection acquisition example 2. FIG. 13 is a characteristic distribution map created based on image data, showing signal intensity distribution. FIG. 14 is a characteristic distribution map created based on image data, showing a distribution map of pixels whose signal intensity exceeds a reference value. FIG. 15 is a characteristic distribution map created based on image data, showing a distribution map of pixels whose signal deviation exceeds a reference value. FIG. 16 is a characteristic map showing a histogram of fluorescence intensity measured after being determined to be normal. FIG. 17 is a characteristic map showing a QQ plot (Quantile-Quantile Plot) of fluorescence intensity measured after being determined to be normal. 14A and 14B are image data showing an example of acquisition of an abnormal fluorescent inspection. 15A and 15B are characteristic distribution maps created based on the image data shown in FIG. 14A, showing signal intensity distribution. 15B and 15C are characteristic distribution maps created based on the image data shown in FIG. 14A, showing a distribution map of pixels whose signal intensity exceeds a reference value. 15B and 15C are characteristic distribution maps created based on the image data shown in FIG. 14A, showing a distribution map of pixels whose signal deviation exceeds a reference value. 15C and 15D are characteristic maps showing an example of a distribution map of pixels whose signal intensity measured after being determined to be abnormal exceeds a reference value. 15D and 15E are characteristic maps showing a histogram of fluorescence intensity measured after being determined to be abnormal. 15F and 15H are characteristic maps showing a QQ plot of fluorescence intensity measured after being determined to be abnormal.

[0009] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Note that like elements are given like numbers and their explanation may be omitted. Also, A to B in the text represent A or more and B or less.

[0010] (Embodiment 1) An inspection operation data system 1001 described in embodiment 1 is a system that uses a computer (server) to perform data shaping, storage, and analysis, and, as shown in FIG. 1 , includes a data input unit 11, a filtering unit 12, a data analysis and metadata generation unit 13, a metadata analysis unit 14, a metadata storage unit 15, a display unit 16, and an information path 17. Here, the metadata storage unit 15 is composed of storage devices such as temporary storage devices such as DRAM and SRAM, and long-term storage devices such as FLASH memory, FRAM (registered trademark), EPROM, hard disks, floppy disks, and magnetic tapes. The information path 17 is a facility that connects the data input unit 11, the filtering unit 12, the data analysis and metadata generation unit 13, the metadata analysis unit 14, the metadata storage unit 15, and the display unit 16 to transmit data and information, and is generally composed of wiring (optical interconnection is also acceptable).

[0011] The data input unit 11 (data input circuit 11) is configured to take in the test data 1. The filtering unit 12 (filtering circuit 12) is configured to select and take in data that is determined to be valid according to predetermined criteria from the data taken in by the data input unit 11, and to identify abnormal data.

[0012] The test data is not particularly limited, but examples thereof include attribute data such as personal attribute data and group attribute data, biomolecule fluorescence image data, image data such as X-ray, ultrasound, MRI and CT, blood test data, urine test data and genetic information data such as DNA. One of the features of the test operation data system 1001 is that it can handle large amounts of image data with further image analysis, as described below.

[0013] The data analysis / metadata generation unit 13 (data analysis / metadata generation circuit 13) is configured to analyze the imported data and then generate metadata, and the metadata analysis unit 14 (metadata analysis circuit 14) is configured to analyze the generated metadata to obtain analysis metadata. The metadata storage unit 15 (metadata storage circuit 15) is configured to store the metadata generated by the data analysis / metadata generation unit 13 and the analysis metadata obtained by the analysis by the metadata analysis unit 14. The metadata analysis performed by the metadata analysis unit 14 can be performed by statistical analysis, and the analysis can include one or more methods selected from the group consisting of time series analysis, regression analysis, principal component analysis, correlation analysis, factor analysis, cluster analysis, canonical analysis, t-analysis, analysis of variance, and random forest.

[0014] The display unit 16 (display circuit 16) is configured to have the function of extracting, organizing, visualizing, and displaying data stored in the metadata storage unit 15 for items selected by the user.

[0015] One feature of the filtering unit 12, data analysis / metadata generation unit 13, metadata analysis unit 14, and metadata storage unit 15 is that they save files in HDF5 (Hierarchical Data Format 5). Using the HDF5 file format enables lightweight hardware such as a PC to handle large volumes of image data, even for millions of users, at high processing speeds. The HDF5 format also allows existing data to be stored and utilized without being overwritten or erased by additional data, making it possible to aggregate data belonging to the same category by attribution. Meanwhile, the display unit 16 can use CSV files, text files, DAT files, and binary files as its file formats. CSV files, which are widely used and easy to handle, are particularly preferred. By using the appropriate file format for each purpose, the inspection operation data system 1001 is able to provide large amounts of data, including images, in a high-speed system with easy-to-use display capabilities.

[0016] Another feature of the test data operation system 1001 of the present invention is that it is equipped with a filtering unit 12. As a result, the test operation data system 1001 can manage abnormal data determined based on predetermined criteria for measured data, particularly image data, separately, and can format, store, and analyze data without including abnormal data.

[0017] Here, examples of abnormal data include, for example, image data in which the signal in only one to several pixels is significantly different from the signals in the surrounding pixels, the smoothness of the signal intensity distribution exceeds the allowable range, kinks occur, or the signal intensity exceeds the saturation level. Measured and input data 1 may contain measurement errors or abnormalities in the measurement sample, and there is a concern that incorporating abnormal data may reduce the reliability of the results of data analysis. Inspection operation data system 1001 is a system that manages abnormal data and is capable of analyzing data while preventing the incorporation of abnormal data.

[0018] As shown in FIG. 2, the test data management system 1002 is based on the test data management system 1001 and further includes an abnormal data storage unit 22 (abnormal data storage circuit 22). The abnormal data storage unit 22 is composed of storage devices such as temporary storage devices such as DRAM and SRAM, and long-term storage devices such as FLASH memory, FRAM (registered trademark), EPROM, hard disks, floppy disks, and magnetic tapes, and stores data determined to be abnormal by the filtering unit 12. The stored information can be displayed and viewed on the display unit 16. Abnormal data is not necessarily generated by measurement errors. It may also lead to the discovery of new phenomena. The test data management system 1002 is a system that can contribute to the discovery of new phenomena.

[0019] As shown in FIG. 3, the test data management system 1003 is based on the test data management system 1002 and further includes an abnormal data analysis unit 21 (abnormal data analysis circuit 21). The abnormal data analysis unit 21 is arranged so as to be able to exchange information with the abnormal data storage unit 22. The abnormal data analysis unit 21 has a function for analyzing whether the data is abnormal data that should be eliminated or data that should be treated as data representing a new event, and in the case of a new event, has a function for displaying the details of the new event on the display unit. Here, the analysis items of the abnormal data analysis unit 21 include discontinuity, deviation, and non-uniformity, and the analysis methods include standard deviation analysis, singular value analysis, and correlation analysis.

[0020] The test data management systems 1001, 1002, and 1003 of the first embodiment provide a test data management system and a data management method therefor, which are lightweight systems capable of high-speed operation and capable of organizing, formatting, storing, analyzing, and visualizing large amounts of data obtained through biomolecular measurement devices and in vitro diagnostic tests. The data is converted into metadata and output, making it convenient and highly statistically applicable. An example of metadata created by the present invention is shown in FIG. 9 . "Bood" in the figure indicates that blood was tested. The test data management system of the present invention comprehensively analyzes health information from in vitro diagnostic tests and other sources, makes a diagnosis, and provides the results directly to the subject in a short period of time, thereby significantly improving convenience and medical care efficiency. Furthermore, the test data management system of the present invention can comprehensively manage information on diseases caused by genetic characteristics and acquired lifestyle habits, thereby enabling the provision of individually optimized health information not available in existing health checkups.

[0021] Second Embodiment In a second embodiment, a method for obtaining metadata related to biomolecules using a biomolecule fluorescence image as input data, using the test data management system 1001 described in the first embodiment, will be described.

[0022] First, a biomolecule fluorescence image is acquired using a biomolecule fluorescence inspection device. Although there are no particular limitations on the biomolecule fluorescence inspection device, it is preferable to use an inspection device equipped with an inspection chip having a metasurface, as disclosed in Patent Document 4, which is highly sensitive and capable of inspection with a small amount of sample.

[0023] The acquired biomolecule fluorescence image is input into the testing operation data system 1001 by the data input unit 11 and sent to the data analysis and metadata generation unit 13 via the filtering unit 12. The data analysis and metadata generation unit 13 automatically selects an analysis region from the sent image data in accordance with predetermined criteria based on the maximum value of the fluorescence intensity of the fluorescence image data, performs smoothing processing on the fluorescence image data within the analysis region to reduce mechanical and thermal noise originating from the measuring instrument, and determines the background intensity distribution within the analysis region.

[0024] The difference between the spatial distribution of fluorescence intensity of the smoothed fluorescence image data and the intensity distribution of the background is then calculated to obtain the spatial distribution of fluorescence intensity of the net fluorescence image data, and the mean intensity and full width at half maximum of the spatial distribution of fluorescence intensity are then calculated.

[0025] Through the above processes (steps), it is possible to obtain from the biomolecule fluorescence image a data group that packages one or more characteristic data selected from the mean intensity, full width at half maximum, and difference amount of the fluorescence intensity spatial distribution, linked to the biomolecule being measured and the biomolecule's attribution information. Note that, although the above example shows an example in which image processing is performed by the data analysis and metadata generation unit 13, it does not necessarily have to be performed by the data analysis and metadata generation unit 13; it may be performed by the filtering unit 12, or the work may be divided in advance and performed by both the filtering unit 12 and the data analysis and metadata generation unit 13.

[0026] Example 1 In Example 1, the test data management system 1003 described in the first embodiment was fabricated as a prototype, and the results of a fluorescence test using a biomolecule sample will be described.

[0027] <Prototype Test Data Operation System> First, the details of the configuration of the prototype test operation data system 1003 will be described below.

[0028] The hardware used was a notebook PC (manufactured by HP Japan) as an analysis server that handles the data input unit 11, filtering unit 12, data analysis / metadata generation unit 13, metadata analysis unit 14, abnormal data analysis unit 21, abnormal data storage unit 22, and display unit 16, and storage as a database server for the metadata storage unit 15.

[0029] The overall structure of the file system consists of Folder 1 (101) and Folder 2 (102), as shown in Figure 4. Folder 1 (101) consists of five functions: Data / H5py (111) responsible for creating H5py, H5py (112) which is the data file storage location, Image (113) which performs image analysis, Make_CSV (114) which creates CSV files, and Visual (115) which performs visualization.

[0030] Data / H5py Reading.ipynb (111) has the function of creating CSV files containing factor names, attribute data, test sample data, etc. in a predetermined format and importing them into H5py. H5py (112) is a directory that stores aggregated data in a predetermined H5py format. Image / CH Image Reading.ipynb (113) is a directory that performs image analysis and consists of two functions: one for detecting anomalies contained in the image and the other for analyzing fluorescence intensity. The analysis results are sent to Folder 2 (102). Make CSV / Make Files.ipynb (114) is a function that aggregates the data obtained from Data / H5py (111) and Image (113) and outputs it in CSV format. Visual / Visual.ipynb (115) has the function of reading H5py (112) files, performing statistical analysis, and visualizing the results.

[0031] Folder 2 (102) consists of Analysis (116), 10h40m31s (117) which stores background data in Tiff format, 11h14m26s (118) which stores foreground data in Tiff format, and Input ref. csv (119) which is an input reference data file. Analysis (116) outputs the analysis results in JPEG and CSV format in parallel with the data of the directory in which the target image is stored.

[0032] The data creation steps were performed as follows (see Figure 5). The entire analysis process was performed on an in-house PC using the Python development environment (Anaconda) on the PC (Step S101). The data to be uploaded to the H5py file was created in a predetermined format, with personal ID, factor name, and test date used as keys (Step S102). Next, an H5py file was created (Step S103), and CSV factor list data was imported into the dataset within the H5py file (Step S104). Subsequently, CSV personal attribute data, CSV time-series data, and automatically generated calendar data were imported into the dataset within the H5py file (Steps S105-S107). A program was then launched to aggregate the data stored in Steps S104-S107 into the H5py file, and the aggregated data was managed (Step S108). Finally, the aggregated data was output to CSV in a predetermined format (Step S109).

[0033] The filtered image analysis was performed according to the following procedure (see Figure 6). The analysis was performed on a PC in the institution using the Python development environment (Anaconda) on the PC (Step S201). First, Tiff format image data from the inspection device was stored in a database server (Step S202). Next, the image analysis program was launched (Step S203). The directory in which the program was stored was Image / CH Image Reading.ipynb (113) in Folder 1 (101).

[0034] Next, image data was read from the database server (step S204), image cell division was performed (step S205), and each cell's image data outlier evaluation was performed (step S206). Specifically, image data reading in step S204 was performed using a Python built-in function compatible with 16-bit images, and image cell division in step S205 divided the 1040 x 1392 pixel image into 100 x 100 pixel square cells. In calculating the image data outliers for each cell in step S206, three types of outliers, a, b, and c, were calculated. Here, outlier a was defined as the number of inflection points excluding the maximum extreme value in the intensity distribution, outlier b was defined as the number of pixels in each cell (100 x 100 pixels) whose intensity exceeded 60,000, and outlier c was defined as the number of pixels in each cell (100 x 100 pixels) whose intensity exceeded the mean value + 3.5 x σ (standard deviation). The mean and standard deviation here were calculated using pixels in the image with an intensity greater than 500 (intensity > 500).

[0035] Thereafter, an error determination was made for the file based on the abnormal values ​​calculated in step S206 (step S207). Specifically, an abnormal value was determined to be abnormal if it met one or more of the following conditions: for abnormal value a, there were two or more abnormal values; for abnormal value b, the maximum value of all calculated values ​​for all cells exceeded 100; and for abnormal value c, the maximum value of all calculated values ​​for all cells exceeded 100. The statistical analysis results were then saved as an analysis result CSV in the data directory (step S208), and image data determined to be normal was stored in an H5py file (step S209).

[0036] Image analysis of normal images was performed using the following procedure (see Figure 7). The analysis was performed on a PC within the institution using the Python development environment (Anaconda) on the PC (Step S301). First, Tiff format image data from the inspection device was stored in a database server (Step S302). Next, the image analysis program was launched (Step S303). The directory in which the program was stored was Image / CH Image Reading.ipynb (113) in Folder 1 (101).

[0037] The image data was then read from the database server (step S304), and Gaussian smoothing was performed on the high fluorescence intensity observed at all coordinates (X, Y) within the image (step S305). The built-in function was ndi.gaussian.filter with sigma set to 50 pixels. The maximum (extreme) value of the smoothed fluorescence intensity was then calculated for the Gaussian smoothing results, and its coordinate position (X, Y) within the image was identified (step S306). Subsequently, smoothing was performed to increase the kurtosis of the distribution for visualization (step S307). The built-in function was ndi.uniform.filter with size set to 5 (simple average within a 5 x 5 pixel cell). Thereafter, the Gaussian mean value was calculated for the fluorescence intensity around the extreme value calculated in step S307 at radius = 150 pixels and sigma = 50 pixels (step S308).

[0038] Next, the ID and concentration reference value were read from the database server, and the fluorescence intensity was read from the results of the above steps (step S309), and the calculated concentration was calculated using the Newton method from the average value obtained in step S308 (step S310). Finally, the histogram and statistical values ​​of the fluorescence intensity were output as a CVS file (step S311), and the fluorescence intensity and test value (concentration) were stored in H5py (step S312), completing the series of normal image analyses.

[0039] Visualization was performed using the following procedure (see Figure 8). The entire analysis process was performed on an in-house PC using the Python development environment (Anaconda) (Step S401). Image data in Tiff format, already imported into the database server, was used (Step S402). Next, a visualization program was launched (Step S403). This visualization program accesses the test metadata and aggregates the relationships between factors. Specifically, it uses regression analysis to identify correlations between each data series and visualizes the factor effects over time and across sections. Data was then loaded from the database server (Step S404), and a histogram (Step S405), correlation matrix (Step S406), two-dimensional factor plot (Step S407), three-dimensional factor plot (Step S408), and fluorescence intensity factor analysis (Step S409) were sequentially performed. Here, in step S405, a histogram was created for each factor, and in step S406, a correlation matrix was created, creating a matrix of correlation coefficients between each factor. In step S407, a two-dimensional factor plot was created, creating a two-dimensional plot between factors with high correlation over time. In step S408, a three-dimensional factor plot was created, creating a three-dimensional plot between factors with high correlation over time. In step S409, a factor analysis of fluorescence intensity was performed using machine learning. Finally, a JPEG file was output to complete the visualization (step S410).

[0040] <Evaluation> Next, the biomolecules were subjected to fluorescence measurement using the biomolecule testing device 2001 shown in FIG. 10, and the measurement results were input into the testing operation data system 1003 for analysis.

[0041] The biomolecule testing device 2001 is a biomolecule detection device that detects and measures biomolecules by irradiating a test chip with a microchannel with light and measuring the fluorescence emitted from a test sample captured on the test chip's test surface. Its main components are shown in Figure 10. The test chip's test surface is a metasurface, and the device achieves extremely high test sensitivity by utilizing the resonance phenomenon caused by the metasurface. The biomolecule testing device 2001 primarily comprises a test chip holder 250 that houses the test chip (not shown), a sample holder 212 that houses a microtube containing a sample, a mounting jig 202 that holds an array of sample collection pins 201 for collecting sample liquid, a pump 205 that delivers the sample to the test chip, a sample holder transfer mechanism 215, a first motorized stage 203, a second motorized stage 260, and a third motorized stage 220 that precisely position the fluorescent test surface to obtain a fluorescent image, and a detection mechanism 214 that obtains a fluorescent image using a confocal optical system. The test chip here is manufactured in accordance with the one disclosed in Patent Document 4.

[0042] The cancer marker PSA was used as a sample. Light with a peak wavelength of 527 nm and a wavelength width of 10 nm was irradiated onto this sample, and fluorescence from the sample with wavelengths of 580 to 650 nm was detected by the fluorescence detection optical system of the biomolecule testing device 2001.

[0043] The measured detection image (fluorescence image) is shown in Figure 11A, and the background image acquired with the same sensor is shown in Figure 11B. The bright vertical band near the center is the sensor area. Temporally, Figure 11B was acquired first, and then the target biomolecule was fixed to the sensor and Figure 11A was acquired. The two images have the same brightness and contrast, and Figure 11A is relatively brighter because it contains a fluorescent signal. Quantifying the data requires analysis of the two images, which was quantified using the inspection operation data system 1003 of the present application.

[0044] Figure 12 shows the image evaluation results of the fluorescence image measurement of Figure 11A, visualizing the absence of abnormalities and proper data. Figure 12A shows the intensity distribution, Figure 12B shows that there are no values ​​outside the 16-bit signal range, and Figure 12C shows that there are no unnatural jumps in the intensity distribution (the evaluation condition is that the value does not exceed 3.5 times the standard deviation). Figure 13A shows the intensity distribution near the maximum value of Figure 11A, which is approximately Gaussian. By approximating this distribution with a Gaussian distribution, the mean value and distribution width (standard deviation) can be quantified. Figure 13B shows the deviation from the Gaussian distribution approximation using a QQ plot. Here, a linear characteristic curve indicates a match with a Gaussian distribution. A match within a ±3 range from the center confirms a nearly Gaussian distribution. Through the above, the difference in mean intensity values ​​from the images of Figures 11A and B could be quantified, and the value was determined to be 945. Since this is the net fluorescence signal intensity, the detected biomolecules could be quantified using a previously measured detection curve.

[0045] On the other hand, Figure 14 shows an example of a fluorescence measurement in which the fluorescence image on the sensor exhibits an uneven distribution (unevenness) due to an experimental issue. In such cases, the data should be rejected because the detection was not performed properly. Figure 15 shows the results of image evaluation performed on the measurement image in Figure 14. Figure 15A shows the spatial distribution abnormality corresponding to the abnormal value a, and Figure 15C, corresponding to the abnormal value c, visualizes the intensity jumps within the image. Figure 15B, corresponding to the abnormal value b, shows no abnormality within the 16-bit signal range. Figure 16 shows an example corresponding to the abnormal value b, in which an abnormality was detected within the 16-bit signal range. In this way, abnormalities in the fluorescence image are identified based on multiple criteria, ensuring the reliability of the data analysis. For reference, Figure 15 shows the characteristic diagram corresponding to the abnormal value b, and Figures 17A and 17B show the distribution diagram and its QQ plot near the maximum intensity value of an actual measurement example that deviates from the Gaussian distribution, respectively.

[0046] Manual image analysis without automation requires at least an hour to process 10 images, even with extensive experience. Furthermore, data shaping is not possible at this stage. In contrast, the present invention, using the inspection operation data system 1003, takes approximately 10 seconds to obtain equivalent analysis results, and data shaping is completed in approximately 5 seconds. The improved processing capacity enables large amounts of data processing, demonstrating the significant technical effect of enabling analysis of volumes that would be impossible to process manually. Furthermore, the present invention is useful because it simultaneously detects abnormalities in image data. There are also significant economic benefits, such as reduced labor costs due to faster and shorter processing times.

[0047] In order to widely provide high-quality, advanced medical care, a test data management system is needed that can quickly read and write data while also organizing, formatting, storing, analyzing, and visualizing large amounts of data obtained through biomolecular measurement devices and in vitro diagnostic tests. The test data management system of the present invention has such characteristics. For this reason, we believe that the system of the present invention will make a significant contribution to society, particularly in the field of medical care.

[0048] 1 Input data 11 Data input unit 12 Filtering unit 13 Data analysis and metadata generation unit 14 Metadata analysis unit 15 Metadata storage unit 16 Display unit 17 Information path (information path) 21 Abnormal data analysis unit 22 Abnormal data storage unit 101 First holder 102 Second holder 111 Data directory (H5py) 112 Data directory H5py (for aggregated data) 113 Image directory (directory for image analysis) 114 Directory for creating CSV data 115 Visual directory (directory for statistical analysis and visualization) 116 Analysis directory 117 First input image data directory (for background) 118 Second input image data directory (for foreground) 119 Input reference file 201 Liquid collection pin 202 Mounting jig 203 First motorized stage 205 Pump 212 Sample holder 214 Detection mechanism 215 Sample holder transfer mechanism 220 Third motorized stage 250 Test chip holder 260 Second motorized stage 1001 Test operation data system 1002 Test operation data system 1003 Test operation data system 2001 Biomolecule testing device

Claims

1. A computer-based data formatting, storage and analysis system for managing test data, the system comprising a data input unit, a filtering unit, a data analysis and metadata generation unit, a metadata analysis unit, a metadata storage unit and a display unit, the data input unit is configured to import test data, the filtering unit is configured to select and import data determined to be valid according to predetermined criteria from the data imported by the data input unit, and identify abnormal data, the data analysis and metadata generation unit is configured to analyze the imported data and generate metadata, the metadata analysis unit is configured to analyze the generated metadata to obtain analysis metadata, the metadata storage unit is configured to store the metadata generated by the data analysis and metadata generation unit and the analysis metadata obtained by analyzing the metadata analysis unit, the display unit is configured to have a function of extracting, organizing, visualizing and displaying data stored in the metadata storage unit for an item selected by a user, and the filtering unit, the data analysis and metadata generation unit, the metadata analysis unit and the metadata storage unit store files in HDF5 format.

2. The inspection data management system of claim 1, further comprising an abnormal data storage unit for storing the abnormal data, and configured to have a function for displaying the abnormal data stored in the abnormal data unit by the display unit.

3. The inspection data management system of claim 2, further comprising an abnormal data analysis unit which analyzes whether the abnormal data is abnormal data that should be eliminated or data that should be treated as data representing a new event, and which is configured to have a function of displaying the contents of the new event on the display unit in the case of a new event.

4. The inspection data management system according to any one of claims 1 to 3, wherein the metadata analysis means is statistical analysis.

5. The inspection data management system of any one of claims 1 to 4, wherein the metadata analysis is one or more selected from the group consisting of time series analysis, regression analysis, principal component analysis, correlation analysis, factor analysis, cluster analysis, canonical analysis, t-analysis, analysis of variance and random forest.

6. The inspection data management system according to any one of claims 1 to 5, wherein the file format of the display unit is a CSV file.

7. A test data management system according to any one of claims 1 to 6, wherein the test data includes personal attribute data.

8. An inspection data management system according to any one of claims 1 to 7, wherein the inspection data includes image data.

9. The inspection data management system according to claim 8, wherein the image data includes biomolecule fluorescence image data.

10. The inspection data management system of claim 9, wherein the inspection data is biomolecule fluorescence image data, and at least one of the filtering unit and the data analysis / metadata generation unit has the following functions: a function to automatically select an analysis area according to predetermined criteria based on the maximum value of the fluorescence intensity of the biomolecule fluorescence image data; a function to perform a smoothing process on the fluorescence image data within the analysis area; a function to obtain a background intensity distribution within the analysis area; a function to calculate the difference between the fluorescence intensity spatial distribution of the fluorescence image data after the smoothing process and the intensity distribution of the background to obtain a net fluorescence intensity spatial distribution of the fluorescence image data; and a function to calculate an average intensity and a full width at half maximum of the net fluorescence intensity spatial distribution.

11. The inspection data management system of claim 10, wherein the metadata includes one or more selected from the group consisting of the average intensity, full width at half maximum, difference amount, the biomolecule being measured, and the attribution information of the biomolecule of the fluorescence intensity spatial distribution.

12. A method for automatically formatting, storing and analyzing test data using a computer, comprising the steps of: data input, filtering, data analysis / metadata generation, metadata analysis, metadata storage and display; the data input step comprises inputting test data into the computer; the filtering step comprises selecting data determined to be valid according to a predetermined standard from the data obtained by the data input step and inputting the data into the computer, and identifying abnormal data; the data analysis / metadata generation step comprises analyzing the input data and generating metadata; the metadata analysis step comprises analyzing the generated metadata and generating analysis metadata; the metadata storage step comprises storing the metadata generated by the data analysis / metadata generation step and the analysis metadata obtained by analyzing the metadata analysis means in the computer; and the display step comprises extracting, organizing, visualizing and displaying the metadata-stored data for an item selected by a user; and the filtering step comprises using HDF5 format as a file format for the filtering step, the data analysis / metadata generation step, the metadata analysis step and the metadata storage step.

Citation Information

Patent Citations

  • Device for genotypic analysis and method for genotypic analysis

    JP2014117222A

  • Abnormality determination method and abnormality determination device

    JP2021174456A

  • System and method of managing large data files

    US20140164444A1

  • Gene-mutation analysis device, gene-mutation analysis system, and gene-mutation analysis method

    WO2015015585A1

  • Image processing apparatus and image processing program

    WO2016093090A1