Method and system for early diagnosis of lesions

Abstract

The present invention relates to a method and system for early diagnosis of lesions configured to enable early diagnosis of lesions by including functions for preprocessing, lesion detection, statistical analysis, visualization, and personalization adjustment based on Whole Slide Image (WSI) data. More specifically, the invention relates to a method for early diagnosis of lesions based on Whole Slide Image (WSI) data using an early lesion diagnosis system, comprising: a step of receiving and preprocessing said WSI data; and a step of detecting lesions by applying an artificial intelligence model to the preprocessed high-resolution image.

Description

Technology Field

[0001] The present invention relates to a technology for early diagnosis of lesions, and more specifically, to a method and system for early diagnosis of lesions configured to enable early diagnosis of lesions by including preprocessing, lesion detection, statistical analysis, visualization, and personalized adjustment functions based on Whole Slide Image (WSI) data. Background Technology

[0002] In pathological diagnosis, the technology of diagnosing lesions using Whole Slide Image (WSI) data, which is obtained by digitizing tissue slides, is attracting attention as a key means to enhance the accuracy and efficiency of medical imaging-based diagnosis. Compared to conventional microscopic observation methods, this WSI-based diagnostic technology has the advantage of being able to incorporate various image processing and artificial intelligence analysis techniques in a digital environment.

[0003] Conventionally, WSI data analysis was limited to manual methods or the application of limited artificial intelligence algorithms to determine the presence of lesions. However, high-resolution WSI data is large in volume and highly complex, making real-time analysis difficult. Furthermore, there were issues with the lack of capabilities to quantitatively display the location or reliability of lesions or to perform statistical analysis.

[0004] Furthermore, there were limitations in considering differences in lesion diagnostic reliability based on user characteristics (e.g., age, gender, medical history, etc.), and in intuitively providing the information necessary for the diagnosing physician's judgment even when lesions were automatically visualized. Accordingly, there is a growing need for technology that enables faster and more precise preprocessing, lesion detection, and visualization, while integrating user-customized lesion display and risk group comparison functions. Prior art literature

[0005] (0001) Korean Registered Patent No. 10-2781169 The problem to be solved

[0006] The technical problem that the present invention aims to solve is to provide a method and system for early diagnosis of lesions capable of rapidly analyzing high-volume medical images by improving the speed and efficiency of WSI (Whole Slide Image) data preprocessing.

[0007] Another technical problem that the present invention aims to solve is to provide a method and system for early diagnosis of lesions that can improve the accuracy of a diagnostician's lesion recognition by quantitatively exploring the location and reliability values ​​of lesions through an artificial intelligence model on high-resolution images and intuitively visualizing them.

[0008] Another technical problem that the present invention aims to solve is to provide a method and system for early diagnosis of lesions that can provide quantitative analysis information usable for diagnostic assistance by statistically analyzing the number and location of lesions based on lesion detection results and visualizing reliability values ​​by interval.

[0009] Another technical problem that the present invention aims to solve is to provide a method and system for early diagnosis of lesions that can intuitively support the diagnostician's judgment by controlling whether to display lesions according to reliability threshold values ​​and zoom-in conditions set by the user, and automatically applying a visualization method accordingly.

[0010] Another technical problem that the present invention aims to solve is to provide a method and system for early diagnosis of lesions that can provide diagnostic information corrected to suit user characteristics, even with the same reliability value, by applying weights to reliability values ​​according to user characteristics such as age, gender, medical history, and family history.

[0011] Another technical problem that the present invention aims to solve is to provide a method and system for early diagnosis of lesions that can increase the objectivity and reliability of diagnosis by providing a user interface that displays the current patient's lesion information in comparison with the average lesion information of patient groups classified into high-risk and low-risk groups.

[0012] The technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below. means of solving the problem

[0013] A method for early diagnosis of a lesion according to an embodiment of the present invention for achieving the above technical problem comprises, in a method for early diagnosis of a lesion based on Whole Slide Image (WSI) data using an early lesion diagnosis system, a step of receiving and preprocessing said WSI data; and a step of applying an artificial intelligence model to the preprocessed high-resolution image to detect the lesion.

[0014] In addition, the step of searching for the lesions is characterized by deriving location information and reliability values ​​for each searched lesion, and determining whether to display the lesion on the image based on the zoom-in degree and reliability standard value set by the user.

[0015] In addition, it is characterized by generating statistical information based on the number and location information of detected lesions, classifying confidence values ​​by interval, and visualizing them in a chart.

[0016] In addition, the preprocessing step includes a tile splitting step, a color normalization step, and an artifact removal step, and is characterized in that the algorithm used for processing each step is optimized based on parallel processing to process the entire WSI data within one minute.

[0017] In addition, the list of the aforementioned detected lesions is displayed on one side of the screen, and each lesion item is characterized by including location information and a reliability value of the corresponding lesion.

[0018] In addition, the color of each lesion in the lesion list and image is distinguished and displayed according to the above reliability standard value, wherein the first color is displayed when the reliability standard value is greater than or equal to the above reliability standard value, and the second color is displayed when the reliability standard value is less than the above reliability standard value.

[0019] In addition, it is characterized by applying weights to the same reliability value by reflecting at least one of the user's age, gender, medical history, and family history information, and determining whether to display the lesion based on the weighted corrected reliability value.

[0020] In addition, it is characterized by providing a user interface (UI) that classifies patients into high-risk and low-risk groups and compares average lesion information derived from pre-accumulated training data with the current patient's lesion information.

[0021] Meanwhile, an early lesion diagnosis system according to one embodiment of the present invention is a system for early diagnosing lesions based on Whole Slide Image (WSI) data, comprising: a preprocessing module configured to receive said WSI data and perform preprocessing including tile segmentation, color normalization, and artifact removal steps, wherein each preprocessing step is configured to process the entire data within one minute by a parallel processing-based algorithm; a lesion detection module configured to automatically detect lesions by applying an artificial intelligence model to the preprocessed high-resolution image, derive location information and confidence values ​​for each detected lesion, and determine whether to display the lesion on the image according to the zoom-in degree and confidence threshold value set by the user; a statistical analysis module configured to generate statistical data based on the number and location information of the detected lesions, classify the confidence values ​​by intervals, and visualize them as a chart; and a visualization module configured to display a list of detected lesions on one side of the screen, include location information and confidence values ​​of the corresponding lesion in each lesion list item, and distinguish and display the lesion on the lesion list and the image in a first color or a second color according to the confidence threshold value. It is characterized by including: a personalization adjustment module configured to apply weights to the same reliability value by reflecting at least one of the user's age, gender, medical history, and family history information, and to determine whether to display a lesion based on the weighted corrected reliability value; and a risk group comparison module that classifies patients into high-risk and low-risk groups and provides the average lesion information of each group derived based on pre-accumulated training data to a user interface (UI) by comparing it with the current patient's lesion information.

[0022] The above embodiments of the present invention are merely some of the preferred embodiments of the present invention, and various embodiments reflecting the technical features of the present invention can be derived and understood by those skilled in the art based on the detailed description of the present invention to be described below. Effects of the invention

[0023] The present invention described above has the following effects.

[0024] First, the present invention can automate the diagnosis of lesions in high-resolution medical images and reduce the time and effort of the diagnostician by preprocessing WSI data and applying an artificial intelligence model to automatically detect lesions.

[0025] In addition, the present invention calculates location information and reliability values ​​for each lesion and controls whether to display the lesion according to the zoom-in degree and reliability standard value set by the user, thereby allowing the visualization of the lesion to be adjusted according to user settings, which can improve the precision and intuitiveness of diagnosis.

[0026] In addition, the present invention generates statistical information on the number and location of lesions and classifies reliability values ​​by interval and visualizes them in a chart, thereby enabling quantitative interpretation of lesion detection results and enhancing the objectivity and explanatory power of the diagnosis.

[0027] In addition, the present invention enables high-speed processing of WSI data within one minute through a parallel processing-based preprocessing algorithm including tile splitting, color normalization, and artifact removal steps, thereby enabling real-time analysis of large volume medical images.

[0028] In addition, the present invention displays a list of detected lesions on one side of the screen and includes location information and reliability values ​​for each lesion item, thereby allowing the location and reliability of the lesions to be intuitively grasped, making it effective for diagnostic assistance.

[0029] In addition, the present invention displays lesions in the lesion list and images in different colors based on reliability criteria values, thereby allowing the importance of lesions to be visually distinguished, which can induce the diagnostician's attention and improve diagnostic efficiency.

[0030] In addition, the present invention enables customized lesion display according to user characteristics and realizes personalization of diagnosis by applying weights to the same reliability value while reflecting the user's age, gender, medical history, and family history information.

[0031] In addition, the present invention provides a user interface that classifies patients into high-risk and low-risk groups and compares average lesion information, thereby allowing the current lesion status of the patient to be judged according to relative standards, thus ensuring objectivity and persuasiveness in diagnosis.

[0032] In addition, the present invention can provide a practical and highly reliable early lesion diagnosis system by integrally performing the entire lesion diagnosis process, from lesion detection to visualization, personalization, and comparative diagnosis, through a system composed of a preprocessing module, a lesion detection module, a statistical analysis module, a visualization module, a personalization adjustment module, and a risk group comparison module.

[0033] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0034] FIG. 1 is an overall conceptual diagram of the configuration according to one embodiment of the early lesion diagnosis system of the present invention. FIG. 2 is an overall flow conceptual diagram according to one embodiment of the method for early diagnosis of lesions according to the present invention. FIG. 3 is a conceptual diagram illustrating a lesion detection process according to an embodiment of the present invention. FIG. 4 is a conceptual diagram of statistical analysis and visualization according to an embodiment of the present invention, FIG. 5 is a conceptual diagram showing a preprocessing step according to an embodiment of the present invention, FIG. 6 is a conceptual diagram of a user interface configuration according to an embodiment of the present invention, FIG. 7 is a conceptual diagram of the user characteristic-based weighting application state according to an embodiment of the present invention, FIG. 8 is an example diagram of a comparative analysis between a high-risk group and a low-risk group according to an embodiment of the present invention. Specific details for implementing the invention

[0035] Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the embodiments of the present invention, if it is determined that a detailed description of related known components or functions would hinder understanding of the embodiments of the present invention, such detailed description is omitted.

[0036] In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the embodiments of the present invention. These terms are intended merely to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by these terms. Where it is stated that a component is "connected," "combined," or "joined" to another component, it should be understood that the component may be directly connected or joined to the other component, but that components may also be "connected," "combined," or "joined" between each component.

[0037] FIG. 1 is an overall configuration conceptual diagram according to an embodiment of the lesion early diagnosis system of the present invention, FIG. 2 is an overall flow conceptual diagram according to an embodiment of the lesion early diagnosis method of the present invention, FIG. 3 is a conceptual diagram showing the lesion search process according to an embodiment of the present invention, FIG. 4 is a conceptual diagram of statistical analysis and visualization according to an embodiment of the present invention, FIG. 5 is a conceptual diagram showing the preprocessing step according to an embodiment of the present invention, FIG. 6 is a conceptual diagram of the user interface configuration according to an embodiment of the present invention, FIG. 7 is a conceptual diagram of the user characteristic-based weight application state according to an embodiment of the present invention, and FIG. 8 is an example diagram of comparative analysis between a high-risk group and a low-risk group according to an embodiment of the present invention.

[0038] Hereinafter, a method and system for early diagnosis of lesions according to an embodiment of the present invention will be described with reference to FIGS. 1 to 8.

[0039] First, referring to FIG. 1, an early lesion diagnosis system according to one embodiment of the present invention is described as follows.

[0040] A lesion early diagnosis system according to an embodiment of the present invention is a system for early diagnosis of lesions based on Whole Slide Image (WSI) data, comprising: a preprocessing module configured to receive said WSI data and perform preprocessing including tile segmentation, color normalization, and artifact removal steps, wherein each preprocessing step is configured to process the entire data within one minute by a parallel processing-based algorithm; a lesion detection module configured to automatically detect lesions by applying an artificial intelligence model to the preprocessed high-resolution image, derive location information and confidence values ​​for each detected lesion, and determine whether to display the lesion on the image according to a zoom-in degree and confidence threshold value set by a user; a statistical analysis module configured to generate statistical data based on the number and location information of the detected lesions, classify the confidence values ​​by intervals, and visualize them as a chart; and a visualization module configured to display a list of detected lesions on one side of a screen, include location information and confidence values ​​of the corresponding lesion in each lesion list item, and distinguish and display the lesion on the lesion list and the image in a first color or a second color according to the confidence threshold value. It may be configured to include a personalization adjustment module configured to apply weights to the same reliability value by reflecting at least one of the user's age, gender, medical history, and family history information, and to determine whether to display the lesion based on the weighted corrected reliability value; and a risk group comparison module configured to classify patients into high-risk and low-risk groups and provide the average lesion information of each group derived based on pre-accumulated training data to a user interface (UI) by comparing it with the current patient's lesion information.

[0041] That is, the lesion early diagnosis system according to one embodiment of the present invention can be configured to automatically detect and visualize lesions based on WSI (Whole Slide Image) data, and to adjust the display of lesions according to user characteristics and comparison group information.

[0042] The preprocessing module (10) is a module that receives WSI data to be diagnosed for lesion diagnosis and performs preprocessing by applying a parallel processing-based algorithm for high-speed processing. That is, by dividing the WSI data into several small data and simultaneously processing the divided data in parallel, the preprocessing work can be completed in a shorter time.

[0043] Specifically, high-resolution images suitable for artificial intelligence analysis can be generated by dividing WSI data into tile units and then performing color normalization and artifact removal on each tile image. At this time, it is preferable that the preprocessing module (10) be configured to process the entire data within one minute.

[0044] The lesion detection module (20) can receive a preprocessed high-resolution image and apply an artificial intelligence model to automatically detect lesions.

[0045] The lesion search module (20) calculates location information and reliability values ​​for each lesion and may include logic to determine whether the lesion is displayed on the screen according to the zoom-in level and reliability standard value set by the user. Through this, it can be configured to enable not only automatic search of lesions but also user-customized display adjustment.

[0046] The statistical analysis module (30) is a module that generates statistical information by collecting the number and location information of lesions detected by the lesion detection module (20).

[0047] The statistical analysis module (30) can divide the reliability value into intervals and visualize the distribution of lesions in each interval in the form of a chart, thereby allowing the diagnostician to quantitatively determine the results of the lesion search.

[0048] The statistical analysis module (30) can be configured to transmit the analyzed information to a server or user terminal so that it is reflected on the diagnostic assistance screen.

[0049] The visualization module (40) is a module that displays a list of discovered lesions on one side of the screen and provides the user with intuitive information about the location and reliability of the lesions in each list item.

[0050] Additionally, the visualization module (40) can distinguish lesions in the list of lesions and images based on a reliability threshold value and display them in color. For example, it is preferable to distinguish and display them in a first color such as blue when the threshold value is greater than or equal to the threshold value, and in a second color such as red when the threshold value is less than or equal to the threshold value.

[0051] The personalization adjustment module (50) may include a function to apply weights to a reliability value by utilizing at least one of the user's age, gender, medical history, and family history, and to calculate a corrected reliability value by correcting the existing reliability value.

[0052] By determining whether to display lesions based on corrected reliability values, customized lesion visualization reflecting the individual characteristics of the user can be realized.

[0053] The personalization adjustment module (50) is linked with the lesion search module (20) and the visualization module (40) to adjust the lesion search and display results in real time.

[0054] The risk group comparison module (60) can store the average lesion information of pre-classified high-risk and low-risk patient groups and perform the function of providing it visually by comparing it with the lesion information of the current patient.

[0055] Comparative information is provided through a user interface (UI) and can support the diagnostician's judgment by visually presenting which group the current patient's lesion distribution or reliability approximates.

[0056] It is desirable that the risk group comparison module (60) be configured to be linked with the statistical analysis module (30) so that average values ​​can be calculated and distributions can be compared.

[0057] The early lesion diagnosis system configured in this way can simultaneously improve the efficiency and precision of diagnosis by rapidly preprocessing large volumes of WSI data and providing reliability-based visualization and user-customized judgment criteria for automatically detected lesions.

[0058] A method for early diagnosis of a lesion according to one embodiment of the present invention may be configured to include the step of receiving and preprocessing WSI (Whole Slide Image) data based on WSI data using an early diagnosis of a lesion system, and the step of applying an artificial intelligence model to the preprocessed high-resolution image to detect the lesion.

[0059] That is, the invention relates to a method for early diagnosis of lesions based on WSI data using the aforementioned early lesion diagnosis system, which can be implemented by organically linking with the components of the early lesion diagnosis system.

[0060] Referring to FIG. 2 together with FIG. 1, a method for early diagnosis of a lesion according to one embodiment of the present invention may include a step of receiving and preprocessing WSI data. This step may be performed through a preprocessing module (10).

[0061] WSI (Whole Slide Image) data typically consists of large-capacity pathological image data ranging from hundreds of MB to several GB, and preprocessing must be performed prior to diagnosis to make it suitable for AI-based analysis.

[0062] The preprocessing module (10) can divide the input WSI data into tile units of a certain size and then sequentially apply color normalization and artifact removal to each tile.

[0063] In addition, it is desirable to optimize the preprocessing process based on parallel processing so that the entire data can be processed within one minute.

[0064] Through such preprocessing, the accuracy of lesion detection can be improved by removing unbalanced color tones or imaging noise, and the data to be analyzed can be refined into a consistent format.

[0065] Additionally, referring further to FIG. 3, a method for early diagnosis of lesions according to one embodiment of the present invention may include a step of applying an artificial intelligence model to a preprocessed high-resolution image to search for lesions. This step may be performed in conjunction with a lesion search module (20).

[0066] Based on preprocessed high-resolution image data, the lesion detection module (20) can automatically detect lesion candidate regions by applying a trained artificial intelligence model, such as a CNN or a Transformer-based model.

[0067] Each detected lesion has unique location information (coordinates, etc.), and the confidence value of the corresponding area can be calculated together.

[0068] The exploration results are utilized for visualization and statistical processing in subsequent stages, and the artificial intelligence model can be trained on various pathology datasets.

[0069] Lesion detection can automatically determine the presence and distribution of lesions by analyzing the size, color, boundary information, etc., of the lesions based on the preprocessing results.

[0070] The early lesion diagnosis method configured in this way can automatically analyze large volumes of pathology images to quickly and precisely detect lesions, thereby reducing the burden on the diagnostician and improving diagnostic accuracy.

[0071] Additionally, referring further to FIG. 4, the step of searching for the lesion according to one embodiment of the present invention may derive location information and reliability values ​​for each searched lesion, and determine whether to display the lesion on the image according to the zoom-in degree and reliability standard value set by the user. This can be implemented centering on the lesion search module (20) and the visualization module (40).

[0072] The lesion detection module (20) can automatically detect lesions through an artificial intelligence model, calculate location information such as the center coordinates or boundary coordinates of each lesion, and simultaneously output a confidence value that quantifies the probability of being classified as a lesion.

[0073] This confidence value can be expressed as a continuous value between 0 and 1, and a higher value may indicate a higher likelihood of a lesion.

[0074] The lesion search module (20) can map the location information and reliability values ​​of each lesion to the lesion and output them in the form of structured data.

[0075] In the subsequent step, whether to display lesions on the image can be determined based on the zoom level and reliability threshold set by the user.

[0076] For example, if the user has set the display of lesions only at high magnification levels (zoom-in degree), the display of lesion search results can be restricted to only those lesions within the user's current screen range.

[0077] In addition, if the user sets a confidence threshold value such as 0.5, they can control it so that only lesions satisfying that threshold value or higher are visually displayed on the image.

[0078] This operation can be implemented through a filtering logic that compares the lesion information detected by the lesion search module (20) with the user setting value, and the result can be displayed on the screen through the visualization module (40).

[0079] The visualization module (40) displays only filtered lesions on the screen, and the displayed lesions are marked at the exact location on the image based on location information and can subsequently be linked in the form of colors or lists.

[0080] The early lesion diagnosis method configured in this way enables flexible lesion visualization tailored to the diagnostician's preferences and circumstances by determining whether to display automatically detected lesions based on user criteria. This improves the accuracy and selectivity of lesion display and can direct the diagnostician's attention to the key areas of the lesions.

[0081] In addition, according to one embodiment of the present invention, statistical information can be generated based on the number and location information of the detected lesions, and reliability values ​​can be classified by interval and visualized as a chart. This is a process of generating and visualizing statistical information based on the lesion detection results, and this process can be performed primarily by the statistical analysis module (30) of the system.

[0082] The aforementioned lesion search module (20) can automatically search for multiple lesions and then transmit location information and reliability values ​​of each lesion to the statistical analysis module (30).

[0083] It is desirable for the statistical analysis module (30) to generate the following statistical information based on the received lesion data.

[0084] - Total number of lesions

[0085] - Distribution of lesions by location (e.g., tissue quadrants, specific anatomical zones, etc.)

[0086] - Distribution of distances between lesions

[0087] - Lesion density per patient, etc.

[0088] This statistical information can be used to intuitively grasp the risk distribution of lesions or to compare with existing patient groups.

[0089] In addition, the statistical analysis module (30) can classify the reliability values ​​calculated for each lesion into certain intervals, such as 0.0 to 0.3, 0.3 to 0.6, 0.6 to 1.0, etc., and then display the number of lesions corresponding to each interval in a chart.

[0090] It is desirable that the chart be implemented in the following way.

[0091] - Visually display the distribution of lesions by confidence interval using bar graphs or pie charts

[0092] - Visualize confidence areas where lesions are concentrated using color or highlight effects

[0093] - Display side-by-side comparison of the patient lesion reliability distribution and the average distribution of the high-risk group

[0094] These charts can be implemented by linking the internal data analysis function of the statistical analysis module (30) and the output function of the visualization module (40).

[0095] The early lesion diagnosis method constructed in this manner can provide a statistical-based analysis that allows for the visual assessment of the distribution and risk level of lesions, going beyond simply determining their presence. Through this, diagnosticians can make more quantitative and reliability-based clinical judgments by considering both numerical and visual information, and the method can also be utilized for comparative learning or performance evaluation with diagnostic AI in the future.

[0096] Additionally, referring further to FIG. 5, the preprocessing step according to one embodiment of the present invention includes a tile splitting step, a color normalization step, and an artifact removal step, and the algorithm used for processing each step may be optimized based on parallel processing to process the entire WSI data within one minute.

[0097] The tile splitting step, color normalization step, and artifact removal step can be configured as follows.

[0098] - Tile division step

[0099] Since WSI data is generally an ultra-high resolution image of hundreds of millions of pixels or more, processing the entire image at once may be inefficient or impossible. Accordingly, it is desirable for the preprocessing module (10) to divide the input WSI data into multiple tile units.

[0100] The tile size can be set to, for example, 256×256 or 512×512 pixels, and may include or exclude overlap.

[0101] By applying parallel processing-based algorithms at this stage, the task of dividing into thousands or more tiles can be performed at high speed using multithreading or GPU parallel computing.

[0102] - Color Normalization Step

[0103] Since the colors of medical images can vary significantly depending on the slide scanner manufacturer, staining conditions, etc., color normalization per tile is necessary.

[0104] The preprocessing module (10) can apply a color normalization algorithm based on a predefined standard color model, such as HED or LAB, for each tile.

[0105] Representative algorithms such as Reinhard normalization and Macenko normalization can be applied in parallel, and this is also optimized based on parallel computation to be executed quickly on all tiles.

[0106] - Artifact Removal Step

[0107] Slide edges, dust, air bubbles, scanner noise, and other artifacts can interfere with lesion detection, so a preprocessing step to remove them is necessary.

[0108] The preprocessing module (10) can be configured to remove artifacts in the following ways. That is, unnecessary areas can be removed or masked using background separation algorithms, boundary detection algorithms, etc., and artifact removal algorithms can be performed in parallel for each tile to remove abnormal elements from the entire WSI image.

[0109] Meanwhile, all of the above preprocessing steps are implemented using parallel processing-based algorithms, and it is desirable to perform them in parallel in a multi-core CPU and GPU-based computing environment.

[0110] Based on a parallel operation optimization design, it can be configured to process the entire WSI data based on an average pathology slide within one minute, and the limitation of processing time can be very important in clinical real-time diagnosis or high-volume lesion analysis environments.

[0111] The preprocessing method configured in this way enables high-speed, accurate, and standardized preprocessing of WSI data, and can significantly improve the accuracy and processing efficiency of the subsequent lesion detection stage.

[0112] In addition, the performance of the entire system can be optimized by strengthening the connectivity between the preprocessing module (10) and the lesion detection module (20).

[0113] In addition, as illustrated in FIG. 6, according to one embodiment of the present invention, a list of the detected lesions is displayed on one side of a screen, and each lesion item may include location information and a reliability value of the corresponding lesion. This relates to an early diagnosis method for lesions configured to display a list of detected lesions on one side of a screen and to include location information and a reliability value of the corresponding lesion in each lesion item. To this end, a visualization module (40) may serve as a major component.

[0114] Information on lesions detected by the lesion detection module (20) can be displayed in a list form on one side of the screen, for example, on the right side of the screen or in the bottom area of ​​the screen, so that the user can intuitively check it. The list can be implemented in the form of a scrollable table or cards.

[0115] Each lesion item may include an individual index (ID), coordinate values ​​(x, y) of the lesion, an enlarged thumbnail image, etc.

[0116] Each lesion item may include location information of the lesion and a confidence score value calculated by the lesion search module (20).

[0117] Location information is provided based on the image coordinate system and can indicate where the lesion is located within the WSI image.

[0118] The confidence value may be provided in the form of a probability value between 0.0 and 1.0 or a percentage (%), for example, and the confidence value may be displayed with a color or level to help the user interpret it.

[0119] When a lesion item included in the list is clicked or selected, the visualization module (40) can be configured to automatically zoom in and display the area where the lesion is located. The zoomed-in location can be variably adjusted according to the zoom-in level set by the user, thereby enabling interactive operation between the lesion list and the overall image viewer.

[0120] The visualization module (40) may also include the following functions to facilitate navigation of the lesion list.

[0121] - Confidence-based filtering function (e.g., Show only confidence levels of 80% or higher)

[0122] - Alignment based on lesion location (e.g., top to bottom)

[0123] - Descending confidence or importance-based sorting function

[0124] These functions can be useful for determining diagnostic priorities based on the severity of the lesion.

[0125] In this way, by providing users (doctors or researchers) with an interface environment that allows them to structurally verify and interact with discovered lesions, the accuracy and efficiency of diagnosis can be significantly improved.

[0126] In addition, by clearly presenting the lesion information derived from the lesion search module (20) through the visualization module (40), the reliability and user experience of the early lesion diagnosis system can be simultaneously enhanced.

[0127] In addition, according to one embodiment of the present invention, the color of each lesion in the lesion list and image may be distinguished and displayed according to the reliability standard value, wherein a first color is displayed when the reliability standard value is greater than or equal to the reliability standard value, and a second color is displayed when the reliability standard value is less than the reliability standard value. This may be advantageous for increasing the efficiency of diagnosis by providing the user with a visual distinction based on the importance or diagnostic reliability of the lesion, and can be implemented mainly through the linkage of the lesion search module (20) and the visualization module (40).

[0128] The lesion search module (20) can calculate location information and confidence scores for each lesion derived based on WSI (Whole Slide Image).

[0129] The visualization module (40) can determine the display color of the lesion based on the reliability value and the reliability reference value, which is a comparison standard.

[0130] It is desirable to configure the confidence threshold value so that it can be set by a doctor or analyst in the user interface (UI). For example, if the threshold value is set to 0.7, lesions with a confidence level of 0.7 or higher may be considered "suspected lesions."

[0131] The visualization module (40) can display the lesion in different colors depending on whether the reliability value of the lesion is higher or lower than the reference value.

[0132] In other words, if the value is above the threshold, it can be displayed in a first color, such as red or orange, to attract the user's attention.

[0133] In addition, if the value is below the threshold, it can be visually distinguished by displaying it in a second color, such as gray or green, which is a color with a low level of attention.

[0134] The corresponding color distinction can be applied to both of the following two screen elements.

[0135] - Border, background, or text color of items in the lesion list

[0136] - Markers, dots, circles, or highlighted areas marked at the lesion location on the WSI image

[0137] In addition, when the user changes the reliability threshold value, it is desirable to configure the visualization module (40) to automatically update by synchronizing the lesion list and the display color on the image in real time, so that the user can intuitively check the classification status of the lesions while adjusting the threshold value.

[0138] The visualization module (40) is designed to take into account users who are colorblind or colorblind, and provides various visual distinction means such as icons or border thickness in addition to colors, for example, the first color area can be replaced with a thick solid line border, the second color with a dotted line border, etc.

[0139] In this way, by providing an interface environment that enhances intuitiveness by visually distinguishing the diagnostic reliability of lesions, it is possible to contribute to improving the speed and accuracy of judgment in the actual pathology diagnosis process.

[0140] In addition, through the organic linkage of the lesion search module (20) and the visualization module (40), visual feedback can be immediately reflected according to the user's settings, allowing for flexible response to various clinical environments.

[0141] In addition, as illustrated in FIG. 7, according to one embodiment of the present invention, weights are applied to the same reliability value by reflecting at least one of the user's age, gender, medical history, and family history information, and the indication of a lesion can be determined based on the weighted corrected reliability value. This is intended to improve the precision and customization of diagnosis by allowing the risk of a lesion to be judged differently according to the user's health history or biological characteristics, even with the same image reading result, and can be implemented mainly through the interaction of a lesion search module (20), a visualization module (40), and a personalization adjustment module (50).

[0142] The lesion detection module (20) can calculate a basic confidence value for each lesion detected in the WSI data.

[0143] The personalization adjustment module (50) can calculate a weight for the risk of a lesion by referring to the user's demographic information (age, gender, etc.) and medical history information, and reflect this in the existing reliability.

[0144] The visualization module (40) can control the display or highlighting of lesions when the corrected reliability value exceeds a set threshold value.

[0145] The personalization adjustment module (50) can operate based on information input by the user or linked from an EHR (electronic health record) system. For example, if a male over 50 years of age has a family history, it can be implemented by evaluating the risk of a specific lesion as high and applying a weight of 1.2 times to the existing reliability value.

[0146] Conversely, for users with low risk factors, excessive display can be prevented by applying a reduction or a weight less than 1.

[0147] It can be configured so that a lesion is included in the list or visually displayed in the image only when the weighted correction confidence value is greater than or equal to the set confidence threshold.

[0148] This method is advantageous for realizing personalized medicine because the display of the same lesion can vary depending on the user.

[0149] The personalization adjustment module (50) can evolve into a dynamic weighting model by reflecting the user's accumulated medical history or previous test results over time, and this can also be applied based on machine learning.

[0150] For example, it is desirable to continuously correct the weighting algorithm by learning from data on discrepancies between past lesion detection results and actual clinical outcomes.

[0151] In this way, the limitations of the lesion display method based on static criteria can be overcome, and a user-centered, flexible diagnostic environment can be provided. Furthermore, by configuring the personalization adjustment module (50) not to function independently but to be linked in real-time with the lesion search module (20) and the visualization module (40), it is possible to enable user-customized lesion judgment within a standardized system.

[0152] In addition, as illustrated in FIG. 8, according to one embodiment of the present invention, a user interface (UI) can be provided to classify patients into high-risk and low-risk groups and compare average lesion information derived based on previously accumulated learning data with the current patient's lesion information. This reflects that the focus of diagnosis should differ depending on the patient's risk group even if the same lesion is detected, and can support clinicians or users in interpreting the meaning of the lesion more precisely through comparison-based visualization, and the statistical analysis module (30) and the risk group comparison module (60) can primarily play a central role.

[0153] The statistical analysis module (30) can generate quantitative lesion information based on the number of lesions, locations, and reliability values ​​of the current patient.

[0154] The risk group comparison module (60) can store the average lesion distribution and characteristics of high-risk and low-risk groups based on previously accumulated learning data and perform the function of comparing this with the current patient's lesion data.

[0155] The classification of high-risk and low-risk groups can be determined through a pre-trained model, and factors such as age, gender, medical history, family history, and history of previous lesions can be used as key judgment elements.

[0156] The risk group classification can also be linked with the personalization adjustment module (50), and it is desirable to use the classification information as a prerequisite for lesion interpretation.

[0157] The risk group comparison module (60) can visually provide reference information, such as “the average number of lesions in the high-risk group is 3.2, and the average confidence level is 0.78,” and the current patient’s number of lesions and confidence level distribution side by side.

[0158] The comparison information can be displayed in the form of graphs, tables, heatmaps, etc. on the UI, and can help to intuitively determine whether there are any abnormal findings.

[0159] In addition, it is desirable for the comparison results to assist medical staff in their response decisions through risk warning messages, color coding, and the like.

[0160] The results of the risk group comparison can be integrated with the visualization module (40) and provided in real-time on the UI along with the distribution of lesions.

[0161] For example, a configuration that visually compares the average location of high-risk groups with the patient's lesion location by overlaying them, or highlights cases where statistical indicators exceed thresholds, can be effective.

[0162] It can be utilized to strengthen early monitoring in high-risk patients by highlighting lesions with relatively low reliability, and to reduce unnecessary additional diagnoses in low-risk patients by selectively highlighting only lesions with high reliability.

[0163] In this way, by enabling the interpretation of individual patient lesion information based on group standards, the reliability of the diagnosis results and the basis for clinical judgment can be strengthened, and by organically linking the risk group comparison module (60) with the statistical analysis module (30) and the visualization module (40), the system can be developed to provide meaning-based interpretation functions beyond simple exploration of lesion data.

[0164] An example of the usage state of the lesion early diagnosis system and method according to one embodiment of the present invention described above is as follows.

[0165] 1. Perform WSI data input and preprocessing

[0166] Whole Slide Image (WSI) data generated by digitizing patient tissue slides through a scanner can be input into an early lesion diagnosis system.

[0167] The input WSI data is transmitted to a preprocessing module (10), and preprocessing can be performed in the following steps.

[0168] - Tile Division Step: Divides WSI data into tile units of a fixed size to increase processing efficiency.

[0169] - Color Normalization Step: Ensures model consistency by removing color differences in pathological images.

[0170] - Artifact removal step: Remove non-pathological areas such as dust, stains, and poor focus.

[0171] Each preprocessing step can be configured to process the entire data within one minute by applying a parallel processing-based algorithm, and such preprocessing serves as a foundation for improving AI diagnostic accuracy and can be considered an essential component for realizing high-speed diagnostic services.

[0172] 2. Perform lesion search

[0173] The preprocessed high-resolution image is transmitted to the lesion detection module (20), and an AI model can be applied to automatically detect lesions.

[0174] The AI ​​determines the presence or absence of a lesion for each tile, and if a lesion is detected, it can calculate the corresponding location information and confidence value together; the location coordinates, tile ID, and confidence score of the detected lesion can be stored separately.

[0175] Users can adjust the zoom level and reliability threshold values ​​in the system settings UI, and whether lesions are displayed on the screen can be determined based on the set criteria.

[0176] This can provide the flexibility to dynamically set judgment criteria to suit the user's diagnostic purpose, such as high-sensitivity detection or specificity enhancement.

[0177] 3. Visualization of lesion list and interface display

[0178] Once the lesion detection is complete, the detected lesion information can be visually provided through the visualization module (40) in the following manner.

[0179] The list of lesions is displayed in a list format on the right or bottom of the screen, and each item may include the following information.

[0180] - Lesion location coordinates

[0181] - Confidence value

[0182] - Shortcut to lesion area (click to zoom in on the area)

[0183] Lesions exceeding the confidence threshold are displayed in the first color, while those below it are displayed in the second color, allowing users to make a judgment at a glance. This UI configuration provides visual clarity for lesion assessment and can serve as a useful interface for reviews and report writing.

[0184] 4. Provision of statistical analysis and visual feedback

[0185] The statistical analysis module (30) can analyze and provide the following information.

[0186] - Total number of lesions, mean confidence level, distribution by location

[0187] - Divide confidence values ​​into multiple intervals and visualize them in the form of a histogram or bar graph.

[0188] - Provides location-centered lesion heatmap

[0189] This enables the identification of quantitative characteristics of lesion distribution and can also be utilized for analyzing changes during follow-up examinations.

[0190] 5. Application of Personalized Lesion Weights

[0191] Information such as age, gender, medical history, and family history entered by the user into the system can be reflected in the lesion interpretation criteria by the personalization adjustment module (50).

[0192] Even with the same confidence level, lesions may be displayed with higher weights applied to users with many risk factors. For example, in the case of an elderly patient with a family history, lesions with a confidence level of 0.6 may be adjusted to be displayed on the screen.

[0193] This feature can optimize the balance of diagnostic sensitivity and accuracy through user-customized lesion indication criteria settings.

[0194] 6. Provision of risk group comparison results

[0195] The risk group comparison module (60) can provide the following comparison UI based on accumulated training data.

[0196] - Classify whether the current patient is in the high-risk or low-risk group

[0197] - Compare information such as the average number of lesions, reliability distribution, and major locations for each group with the current patient.

[0198] - Example screen: Visual comparison display such as "Average number of lesions in high-risk group: 3.8 / Current patient: 5 → High-risk or higher warning"

[0199] Such functions can be utilized as a basis to enhance the reliability of the clinical interpretation of diagnostic results.

[0200] Meanwhile, the preprocessing module (10) and the lesion detection module (20) can be configured to be operated on a cloud basis so that they can be utilized simultaneously by multiple hospitals or diagnostic institutions.

[0201] In addition, among the detected lesions, those exhibiting atypical features (irregular boundaries, high heterogeneity, etc.) can be separately classified and provided as notifications, which can be useful for setting priorities for subsequent biopsies.

[0202] In addition, a function to determine progression by comparing past slide information and current lesion information of the same patient can be added, and an example of providing a UI that allows lesions to be explored and checked on a tablet or mobile device can also be implemented by simplifying the visualization module (40).

[0203] According to the usage state of one embodiment of the present invention described above, the process can be performed sequentially as 'data input → preprocessing → AI exploration → visualization and statistical analysis → personalization adjustment → risk group comparison', and can be utilized as a practical technology that can simultaneously ensure the speed and precision of pathology diagnosis in the medical field.

[0204] In the foregoing, although all components constituting an embodiment of the present invention have been described as being combined or operating in combination, the present invention is not necessarily limited to such embodiments. That is, within the scope of the purpose of the present invention, all components may be selectively combined in one or more ways to operate. Furthermore, terms such as "include," "constitute," or "have" described above, unless specifically stated otherwise, mean that the relevant component may be inherent; thus, they should be interpreted as allowing for the inclusion of additional components rather than excluding other components. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains, unless otherwise defined. Terms commonly used, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and, unless explicitly defined in the present invention, should not be interpreted in an ideal or overly formal sense.

[0205] The foregoing description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications and variations within the scope of the essential characteristics of the present invention. Accordingly, the embodiments disclosed in this invention are intended to explain, not limit, the technical concept of the present invention, and the scope of the technical concept of the present invention is not limited by such embodiments. The scope of protection of the present invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present invention. Explanation of the symbols

[0206] 10: Preprocessing Module 20: Lesion Detection Module 30: Statistical Analysis Module 40: Visualization Module 50: Personalized Adjustment Module 60: Risk Group Comparison Module

Claims

Claim 1 A method for early diagnosis of lesions based on Whole Slide Image (WSI) data using an early lesion diagnosis system, comprising: receiving said WSI data and performing preprocessing including a tile segmentation step, a color normalization step, and an artifact removal step, wherein the algorithm used for processing each step is optimized based on parallel processing to process the entire WSI data within one minute; Early diagnosis of lesions characterized by including the step of applying an artificial intelligence model to preprocessed high-resolution images to detect lesions, deriving location information and confidence values ​​for each detected lesion, determining whether to display lesions on the image based on a zoom level and confidence threshold set by the user, displaying a list of detected lesions on one side of the screen, wherein each lesion item includes location information and confidence values ​​of the corresponding lesion, and distinguishing and displaying the colors of each lesion list and lesion on the image according to the confidence threshold value, wherein a first color is displayed if the confidence threshold value is greater than or equal to the confidence threshold value, and a second color is displayed if the confidence threshold value is less than the confidence threshold value, applying weights to identical confidence values ​​by reflecting at least one of the user's age, gender, medical history, and family history information, determining whether to display lesions based on weighted corrected confidence values, generating statistical information based on the number and location information of detected lesions, classifying confidence values ​​by intervals and visualizing them in a chart, classifying patients into high-risk and low-risk groups, and providing a user interface (UI) that compares average lesion information derived based on pre-accumulated training data with the current patient's lesion information. method. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 A system for early diagnosis of lesions based on Whole Slide Image (WSI) data comprises: a preprocessing module configured to receive the WSI data as input, perform preprocessing including tile segmentation, color normalization, and artifact removal steps, wherein each preprocessing step processes the entire data within one minute using a parallel processing-based algorithm; a lesion detection module configured to automatically detect lesions by applying an artificial intelligence model to the preprocessed high-resolution image, derive location information and confidence values ​​for each detected lesion, and determine whether to display the lesion on the image according to the zoom level and confidence threshold value set by the user; a statistical analysis module configured to generate statistical data based on the number and location information of the detected lesions, classify the confidence values ​​by intervals, and visualize them in a chart; a visualization module configured to display a list of detected lesions on one side of the screen, include location information and confidence values ​​for each lesion list item, and distinguish and display the lesions on the lesion list and image in a first color or a second color according to the confidence threshold value; and a system configured to apply weights to the same confidence value by reflecting at least one of the user's age, gender, medical history, and family history information, wherein the weights An early lesion diagnosis system characterized by comprising: a personalization adjustment module configured to determine whether to display a lesion based on an applied correction reliability value; and a risk group comparison module that classifies patients into high-risk and low-risk groups and provides average lesion information of each group derived based on pre-accumulated learning data, compared with the current patient's lesion information, through a user interface (UI); wherein the lesion detection module, statistical analysis module, visualization module, personalization adjustment module, and risk group comparison module are organically linked to integrally perform the entire lesion diagnosis process from lesion detection to visualization, personalization, and comparative diagnosis.

Images