Focus assessment method based on multi-source data gray features

Abstract

The invention relates to the technical field of medical image processing, in particular to a focus assessment method based on multi-source data gray features, which comprises the following steps: S1, collecting and receiving multi-source image data of an object to be analyzed, and carrying out unified identification and classification on the data to form a multi-source image data set; s2, preprocessing the multi-source image data set, determining a normal tissue area and a focus area in an image, and extracting spatial position and form information of a focus; and S3, carrying out digital processing on the gray scale information of the lesion area to generate lesion gray scale feature data in a unified format. According to the invention, the gray scale information of the parenchyma part of the lesion in the image is mapped into 16 digits, so that the digitalization of gray scale data in image diagnosis is realized. By means of 16 gray scale levels which can be distinguished by human eyes, features such as the density, the signal intensity or the echo of the focus are accurately expressed, the data size in traditional image processing is reduced, and meanwhile the diagnosis accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, and in particular to a lesion evaluation method based on gray scale features of multi-source data. BACKGROUND

[0002] Imaging diagnosis is a widely used technology in modern medicine, which mainly analyzes image data to help doctors judge the internal pathological conditions of the human body. In imaging, the gray scale information of lesions such as density, signal intensity, and echo is an important diagnostic basis. Different imaging devices such as X-ray, CT, MRI, and B-US obtain lesion images through different imaging principles, and the gray scale information in different devices and technologies shows certain differences. Imaging physicians rely on these gray scale information to identify the type and nature of the lesion through visual observation and experience analysis, and usually divide the gray scale into several levels to help judge the benign and malignant nature of the lesion and other pathological characteristics. Different imaging techniques have different forms of lesion presentation, for example, CT can distinguish lesions by density, MRI relies on signal intensity for diagnosis, and B-US evaluates lesions based on echo intensity. These gray scale information of image data reflects the tissue structure and functional characteristics of the lesion to some extent, and is of great significance for auxiliary diagnosis and lesion evaluation.

[0003] In the prior art, imaging diagnosis relies on visual observation of the gray scale information in the image by imaging physicians, and the nature of the lesion is judged by intuitive description of different gray levels. Since the human eye can only distinguish a limited number of gray levels, and there are gray scale differences in the acquisition of images by different imaging devices, the subjectivity in the diagnosis process is strong, and the gray scale information between different imaging techniques is difficult to unify and standardize. This method cannot accurately capture the small gray scale differences of the lesion, thereby affecting the accuracy and consistency of the diagnosis. SUMMARY

[0004] To make up for the above shortcomings, the present application provides a lesion evaluation method based on gray scale features of multi-source data, aiming to improve the problems of large gray scale difference, non-uniform standard, and strong subjectivity in imaging diagnosis in the prior art.

[0005] In a first aspect, the present application provides the following technical solution: a lesion evaluation method based on gray scale features of multi-source data, comprising the following steps:

[0006] S1, acquiring and receiving multi-source image data of an object to be analyzed, and uniformly marking and classifying the data to form a multi-source image data set;

[0007] S2, preprocessing the multi-source image data set, determining the normal tissue region and lesion region in the image, and extracting the spatial position and morphological information of the lesion;

[0008] S3, digitize the gray scale information of the lesion area to generate a uniform format of lesion gray scale feature data;

[0009] S4, apply a region adaptive gray scale mapping mechanism to the lesion gray scale feature data, select or adjust the gray scale mapping strategy according to the characteristics of different internal and peripheral tissue regions of the lesion, and generate region differentiated gray scale mapping data;

[0010] S5, based on the physical optical characteristics of the image, apply a multi-dimensional gray scale mapping mechanism of layered optical response characteristics to the gray scale mapping data, analyze the reflection, transmission and scattering components of the image in layers, and generate multi-dimensional gray scale mapping data;

[0011] S6, standardize the multi-dimensional gray scale mapping data of each sequence and each tissue region in the order of the preset sequence and tissue region to form a lesion gray scale feature data set for subsequent lesion evaluation and analysis.

[0012] By adopting the above technical scheme, the gray scale information obtained by different devices and technologies in the image is standardized, and the small gray scale differences of the lesion are accurately captured through digitalization and region adaptive gray scale mapping mechanism. The problems of large gray scale difference, non-uniform standard and strong subjectivity in the prior art are solved, thereby improving the accuracy, consistency and automation level of image diagnosis.

[0013] Preferably, the forming of the multi-source image data set comprises: collecting multi-source image data of the object to be analyzed, including image data from different image modalities or different imaging devices; identifying and classifying the multi-source image data, associating each image data with its source information, acquisition time, sequence type and scanning parameters; performing format unification processing on the classified image data, including image resolution, pixel gray scale range and storage format standardization; correcting the spatial position and direction difference of the image according to the image acquisition conditions and the information of the subject to be examined, realizing the spatial registration of the multi-source image; performing integrity check and data completion on the spatially registered image data, including abnormal data identification and repair, to form a unified multi-source image data set.

[0014] Preferably, the determination of the normal tissue region and the lesion region in the image comprises: pre-processing the multi-source image data, including denoising, enhancing contrast and standardizing gray scale range; using image segmentation algorithm to divide the image data at pixel level or region level, and distinguishing different tissue regions and candidate lesion regions; classifying the segmented regions according to the tissue boundary, morphological characteristics and spatial position relationship, and marking them as normal tissue regions or lesion regions; extracting the spatial position, morphological characteristics and boundary information of the lesion region, and storing them in association with the multi-source image data set.

[0015] Preferably, the digitization processing comprises: standardizing the gray scale values in the lesion area, unifying the gray scale range and eliminating the differences between different image sources; mapping the standardized gray scale values to digital representations according to a preset gray scale grading scheme to form lesion gray scale data in a unified format; and structuring the digital lesion gray scale data to form lesion gray scale feature data according to the spatial distribution of the lesion area and the image sequence order.

[0016] Preferably, the region-adaptive gray scale mapping mechanism comprises: dividing the lesion and the surrounding tissue area into a plurality of independent regions and identifying each region; and performing regional processing on the gray scale data of each independent region to generate gray scale mapping data of each region.

[0017] Preferably, the selection or adjustment of the gray scale mapping strategy comprises: classifying and identifying the lesion and the surrounding tissue area to define each region as an independent processing unit; establishing a candidate set of gray scale mapping parameters for each region, including a gray scale range, a gray scale distribution function, and region boundary information; selecting or adjusting the candidate parameters according to the structural features and gray scale distribution of the region to form a gray scale mapping strategy corresponding to each region; coordinating the gray scale mapping strategies of adjacent or connected regions to ensure the continuity and consistency of the gray scale data; and applying the gray scale mapping strategy of each region to the gray scale feature data to generate region-differentiated gray scale mapping data.

[0018] Preferably, the multi-dimensional gray scale mapping mechanism of the layered optical response features comprises: performing optical layering division on the image data of the lesion area and the surrounding tissue area to extract reflection, transmission, and scattering components; establishing an optical layering response matrix to parameterize the optical characteristics of each layer, including brightness, gray scale distribution, and spatial position information; generating a multi-dimensional gray scale mapping model according to the optical layering response matrix, each dimension corresponding to a different optical layering characteristic of the image; matching the multi-dimensional gray scale mapping model with the lesion gray scale feature data to map each layer of data and generate multi-dimensional gray scale data; and uniformly encoding and identifying each layer of mapping data to form a structured multi-dimensional gray scale mapping data set.

[0019] Preferably, the layered analysis comprises: performing optical channel separation on the input image data to extract data of reflection, transmission, and scattering components respectively; performing spatial filtering and feature extraction on each optical component to obtain gray scale distribution, brightness value, and region boundary information; establishing a layered data structure for the extracted optical component information to record the spatial position, gray scale range, and relative relationship of each layer; and uniformly identifying and archiving the layered data to generate a structured layered data set for use by the multi-dimensional gray scale mapping model.

[0020] Preferably, the preset sequence and tissue region order include: establishing a preset sequence order table and a tissue region order table according to the image type and the acquisition sequence; numbering and sorting the sequence image data according to the sequence order; and identifying and arranging the data of each tissue region where the lesion is located according to the tissue region order.

[0021] Preferably, the formation of the lesion gray scale feature data set includes: layering and integrating the multi-dimensional gray scale mapping data from different sequences and different tissue regions; establishing a data structure corresponding to the lesion region and the spatial position; and uniformly encoding and storing the integrated lesion gray scale feature data to form a standardized data set for subsequent analysis.

[0022] The present application has the following beneficial effects:

[0023] 1、In the present application, the gray scale information of the lesion in the image is mapped to 16-bit numbers, realizing the digitization of gray scale data in image diagnosis. By utilizing the 16 gray scale levels that the human eye can distinguish, the density, signal strength or echo characteristics of the lesion are accurately represented, reducing the data volume in traditional image processing and improving the accuracy of diagnosis. Through the digitized gray scale information, image doctors can more accurately identify different types of lesions, solving the problem of gray scale differences caused by different image devices and scanning methods, providing more efficient support for AI image diagnosis, and promoting the application development of whole-body image diagnosis.

[0024] 2、In the present application, by introducing a region adaptive adjustment model, the mapping strategy of different tissue regions is automatically selected and adjusted in the image gray scale mapping process, realizing the region-differentiated gray scale optimization. This method solves the problem of too large or too small gray scale difference in different tissue regions, and enhances the recognition ability and accuracy of the lesion region.

[0025] 3、In the present application, by introducing a multi-dimensional gray scale mapping mechanism of layered optical response characteristics, a multi-dimensional gray scale mapping model based on reflection, transmission and scattering components is established. The multi-dimensional gray scale information enhances the accuracy of traditional gray scale mapping, solves the problem that a single gray value cannot accurately represent complex lesion characteristics, and improves the delicacy and accuracy of lesion evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A flowchart of a lesion evaluation method based on multi-source data gray scale features is proposed for the present application;

[0027] Figure 2 A numerical value diagram measured by using gray scale scoring method for a lesion evaluation method based on multi-source data gray scale features is proposed for the present application;

[0028] Figure 3A 16-level gray diagram of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0029] Figure 4 An ovary teratoma diagram of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0030] Figure 5 A distal femur metaphyseal osteosarcoma diagram of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0031] Figure 6 A partial canceration diagram of a liver cirrhosis nodule of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0032] Figure 7 A basal ganglion type III astrocytoma diagram of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0033] Figure 8 A scatter diagram of four measurement points based on an ultrasound image of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0034] Figure 9 A polyline diagram of three measurement points based on a DR projection of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0035] Figure 10 A polyline diagram of three measurement points based on a CT scan of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0036] Figure 11 A polyline diagram of four measurement points based on an Mr sequence of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application;

[0037] Figure 12 A polyline diagram of data based on measurement point 1 generated in time sequence of a lesion evaluation method based on multi-source data gray scale characteristics according to the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0039] Embodiment one:

[0040] In a first embodiment of the present application, a lesion evaluation method based on multi-source data gray scale features is provided, as shown in the following steps: Figures 1-12

[0041] S1, collect and receive multi-source image data of the object to be analyzed, and uniformly identify and classify the data to form a multi-source image data set;

[0042] Further, forming the multi-source image data set includes: collecting multi-source image data of the object to be analyzed, including image data from different imaging modalities or different imaging devices; identifying and classifying the multi-source image data, associating each image data with its source information, acquisition time, sequence type and scanning parameters; performing format uniform processing on the classified image data, including image resolution, pixel gray scale range and storage format standardization; correcting the spatial position and direction difference of the image according to the image acquisition conditions and the information of the subject to be examined, realizing the spatial registration of the multi-source image; performing integrity check and data completion on the spatially registered image data, including abnormal data identification and repair, forming a unified multi-source image data set.

[0043] Specifically, multi-source image data of the object to be analyzed is collected and received. Multi-source image data refers to image data from different imaging modalities or different imaging devices, such as computed tomography (CT), magnetic resonance imaging (MRI), ultrasonic imaging (B-US), digital X-ray imaging (DR), etc. By using multiple image acquisition devices, image data with different characteristics can be obtained, thereby providing comprehensive information for lesion evaluation. In order to facilitate subsequent analysis and processing, these image data need to be uniformly identified and classified to ensure accurate association of each image data with its source information, acquisition time, sequence type and scanning parameters, etc.

[0044] In the identification and classification process, first, according to the information of the type of image, imaging device, acquisition time, etc., each image data is classified and identified. In this way, subsequent management and analysis are facilitated, and at the same time, the accurate traceability of the image data is ensured. In addition to including the basic information of the imaging modality, each image data also needs to be matched with specific imaging device parameters, including technical parameters during scanning, image resolution, sequence type of acquisition, etc. These information is crucial for subsequent image processing, especially when image fusion and comparison, ensuring consistency and comparability between different images.

[0045] ​For the classified image data, the next step is to perform format unification processing. The standardization of image resolution, pixel gray scale range, and storage format is the key content of this step. Since the image data from different devices may have different resolutions and gray scale ranges, it is necessary to perform uniform processing on image data from different sources. First, by unifying the resolution standard, the image data collected by different devices is adjusted to the same resolution, ensuring the accuracy of subsequent gray scale analysis and lesion evaluation. Then, the pixel gray scale range is standardized to eliminate the gray scale differences produced by different imaging devices and ensure that the image data is compared under the same standard. The formula for gray scale value standardization can be expressed as: where G norm is the standardized gray scale value, G is the original gray scale value, G min and G max are the minimum and maximum gray scale values in the original image, G new_min and G new_max are the minimum and maximum values of the target standardized gray scale range.

[0046] The uniform processing of storage format ensures that image data can be processed and stored on different software platforms without affecting subsequent data analysis due to inconsistent formats.

[0047] After completing the format unification, the image data needs to be spatially registered according to the image acquisition conditions and subject information. During image acquisition, due to differences in devices, body position differences of subjects, and different imaging modalities, the spatial position of the same lesion in different images may shift or differ in direction. Therefore, in the spatial registration step, common registration algorithms such as mutual information method or feature-based registration method are used to spatially align image data of different modalities, ensuring that different image data can be accurately mapped into the same spatial coordinate system. The implementation of spatial registration needs to consider the resolution of image data, the time point of acquisition, and the characteristics of imaging devices. The mathematical model of registration can be realized by minimizing the difference between images, and the commonly used registration objective function is: where F reg is the registration objective function, I1(x i ) and I2(x i ) are the pixel values of two images at position x i , and N is the total number of pixels in the image. By minimizing this objective function, the spatial registration of two images can be achieved.

[0048] The last step is to check the integrity of the registered image data and complete the data. The acquisition process of image data may be affected by various factors, such as equipment failure, poor acquisition conditions, etc., resulting in missing data or abnormal values. Therefore, the registered image data must be checked to identify abnormal data and repair it. The detection of abnormal data can be achieved by automated algorithms, such as image repair algorithms, edge compensation algorithms, etc., to fill in the missing parts and eliminate abnormal data. The completed image data will form a unified multi-source image data set, providing accurate data support for subsequent lesion gray scale analysis, region identification and lesion evaluation.

[0049] Through the above steps, the multi-source image data set provided by the embodiment can ensure the consistency, accuracy and integrity of the image data. After unified identification, format standardization and spatial registration processing of image data from different sources, the differences between data can be eliminated, providing a more reliable analysis basis. In addition, the data completion step can further improve the integrity of the image data, ensuring that subsequent analysis is not affected by data loss or abnormal values, thereby providing high-quality data support for accurate lesion evaluation.

[0050] The technical effect of this embodiment is that by performing unified identification and classification, format standardization, spatial registration and data completion on multi-source image data, the accuracy and reliability of the image data in subsequent lesion evaluation are ensured. These processing steps can effectively solve the problem of data differences between different devices and different modalities, ensuring accurate fusion and analysis of multi-source image data, and providing solid data support for lesion evaluation.

[0051] S2, pre-processing the multi-source image data set, determining the normal tissue region and lesion region in the image, and extracting the spatial position and morphology information of the lesion;

[0052] Further, determining the normal tissue region and lesion region in the image includes: pre-processing the multi-source image data, including denoising, enhancing contrast and standardizing gray scale range; using image segmentation algorithm to divide the image data at pixel level or region level, to distinguish different tissue regions and candidate lesion regions; according to the tissue boundary, morphological characteristics and spatial position relationship, classifying the segmented regions, and marking as normal tissue region or lesion region; extracting the spatial position, morphological characteristics and boundary information of the lesion region, and storing it in association with the multi-source image data set.

[0053] Specifically, for the preprocessing of multi-source image data sets, first, the data is denoised, the contrast is enhanced, and the standard gray scale range is processed to ensure that the image data used for subsequent analysis has higher quality and consistency. Denoising processing removes random noise in the image by applying filtering algorithms, such as using Gaussian filtering or median filtering algorithms, which can smooth the pixel values in the image while maintaining the edge features of the image unaffected. The process of enhancing contrast includes adjusting the brightness of the image to make the contrast between the lesion area and the background area more obvious, thereby facilitating subsequent region segmentation and classification.

[0054] The pre-processed image data is subjected to gray scale standardization processing to eliminate the gray scale differences produced by different devices, different image modalities or scanning conditions, ensuring the consistency of the data. Through standardization processing, the gray scale range of all images is unified, ensuring that different image data is compared and analyzed under the same standard.

[0055] After the image data is standardized, image segmentation algorithms are used for pixel-level or region-level division to distinguish normal tissue regions from lesion regions in the image. Common segmentation algorithms include threshold-based segmentation methods, region growing algorithms, and deep learning models, which can accurately divide different tissue types and lesion regions based on the differences in pixel gray values. Specifically, the threshold segmentation method sets a gray threshold, and regions with gray values above the threshold are determined as lesion regions, while regions with gray values below the threshold are normal tissue regions.

[0056] After segmentation, for the lesion region, the spatial position, morphological features and boundary information of the lesion are extracted through morphological analysis. Spatial position can be represented by coordinates in the coordinate system, while morphological features include lesion size, shape, edge smoothness, etc. Boundary information can be further analyzed by contour extraction algorithms such as Canny edge detection algorithm or Sobel operator. These information will provide the basis for subsequent lesion evaluation.

[0057] The extracted spatial position information, morphological features and boundary information of the lesion region are associated with the multi-source image data set and stored in the database for subsequent analysis. In this way, during lesion evaluation, multi-dimensional information of the lesion can be quickly retrieved and combined for diagnostic analysis. The data structure stored includes the coordinates of the lesion region, the boundary contour, the gray scale features and their relationship with the surrounding tissues, etc.

[0058] Through preprocessing, segmentation and feature extraction of image data, normal tissue and lesion area can be effectively distinguished, and foundation is laid for subsequent lesion gray scale feature extraction, analysis and evaluation. Through this method, the image data is standardized, denoised and enhanced, and the lesion area is accurately extracted and labeled, providing more accurate data support for subsequent AI diagnosis.

[0059] S3, digitally processing the gray scale information of the lesion area to generate lesion gray scale feature data in a unified format;

[0060] Further, the digital processing includes: standardizing the gray scale values in the lesion area, unifying the gray scale range and eliminating the differences between different image sources; mapping the standardized gray scale values to digital representations according to a preset gray scale grading scheme to form the lesion gray scale data in a unified format; and structuring the digital lesion gray scale data to form the lesion gray scale feature data according to the spatial distribution of the lesion area and the image sequence order.

[0061] Specifically, the gray scale information of the lesion area is digitally processed to generate lesion gray scale feature data in a unified format. In the specific implementation process, first, the gray scale values in the lesion area are standardized. The purpose of gray scale standardization is to unify the gray scale range and eliminate the gray scale differences between different image sources, to ensure the consistency of image data from different sources in subsequent processing. Each gray scale value can be mapped to a standard range of [0, 1], thereby eliminating the gray scale differences caused by equipment, modalities or other factors.

[0062] After gray scale standardization, the standardized gray scale values are digitally mapped according to a preset gray scale grading scheme to form the lesion gray scale data in a unified format. Specifically, the preset gray scale grading scheme maps the standardized gray scale values to discrete gray scale levels to form digital representations. Assuming that the grading scheme divides the gray scale values into N levels, the gray scale value G norm The discretization can be performed by the following mapping formula: where G digitized is the digital gray scale value, N is the preset gray scale grading number, represents the floor operation. Through the formula, the standardized gray scale values are mapped to discrete digital gray scale values to form the lesion gray scale data in a unified format.

[0063] The digitized grayscale data of lesions is structured. The purpose of this structuring is to systematically organize the grayscale data based on the spatial distribution of the lesion region and the sequence order of the images. The spatial distribution of the lesion region determines the spatial location of the grayscale data, while the image sequence order ensures that the data is ordered according to the sequence of image acquisition. Assuming the spatial location of the lesion region is represented by coordinates (x, y), and the sequence order of the images is represented by t, the grayscale feature data of the lesion can be structured in the following way: D feature ={(x i ,y i ,t j G digitized )};wherein, D feature This represents the structured grayscale feature data set of lesions, (x i ,y i ) represents the spatial coordinates of the lesion area, t j For the image sequence order, G digitized This represents the digital grayscale value corresponding to the lesion area.

[0064] Through this structured approach, grayscale feature data of lesions can be effectively organized according to the location and time series of the lesions, thereby providing structured input data for subsequent lesion classification, assessment and prediction.

[0065] By digitizing the grayscale information of lesion areas, we can not only effectively unify the differences between different image sources, but also provide a structured digital grayscale data format for lesions, facilitating subsequent analysis. This digitized grayscale data can be used to train machine learning models or for quantitative analysis, providing effective support for the qualitative and quantitative assessment of lesions.

[0066] S4. Apply a region-adaptive grayscale mapping mechanism to the grayscale feature data of lesions. Select or adjust the grayscale mapping strategy according to the characteristics of different lesion internal and surrounding tissue regions to generate region-differentiated grayscale mapping data.

[0067] Furthermore, the regional adaptive grayscale mapping mechanism includes: dividing the lesion and its surrounding tissue area into several independent regions and identifying each region; performing regionalization processing on the grayscale data of each independent region to generate grayscale mapping data for each region.

[0068] Furthermore, the selection or adjustment of grayscale mapping strategies includes: classifying and labeling the internal and surrounding tissue regions of the lesion, defining each region as an independent processing unit; establishing a candidate set of grayscale mapping parameters for each region, including grayscale range, grayscale distribution function, and region boundary information; selecting or adjusting the candidate parameters based on the structural characteristics and grayscale distribution of the region to form a grayscale mapping strategy corresponding to each region; coordinating the grayscale mapping strategies of adjacent or connected regions to ensure the continuity and consistency of grayscale data; and applying the grayscale mapping strategy of each region to the grayscale feature data to generate region-differentiated grayscale mapping data.

[0069] Specifically, the lesion area and its surrounding tissue area are divided into several independent regions, and each region is labeled. This division can be achieved using image segmentation algorithms, such as threshold-based segmentation methods or semantic segmentation using deep learning techniques. In image data, lesion areas often differ significantly from surrounding normal tissues in grayscale values, morphological features, and spatial distribution. Therefore, it is essential to clearly distinguish these regions for subsequent processing.

[0070] After standardization, based on a pre-defined grayscale grading scheme, the standardized grayscale values ​​are mapped to a digital representation. A suitable grayscale distribution function can be selected for mapping based on the distribution characteristics of the grayscale values. Different types of grayscale distribution functions can be used depending on the characteristics of the region, such as linear mapping, gamma functions, or Gaussian functions. Assuming a linear mapping method is used, the grayscale mapping formula is: G mapped =a·G norm +b; where a and b are the coefficients of the linear mapping, G mapped These are the mapped grayscale values. This linear mapping process converts the standardized grayscale values ​​into a data format suitable for subsequent analysis.

[0071] In the selection and adjustment of grayscale mapping strategies, each region is first classified based on its grayscale distribution and regional structural characteristics. For example, the grayscale values ​​of lesion regions often have more concentrated or specific distribution characteristics, while the grayscale values ​​of normal tissues may be more uniform. Therefore, it is necessary to establish a set of candidate grayscale mapping parameters for each region, including: grayscale range (G min and G max ); gray-scale distribution function (e.g., Gaussian distribution, gamma distribution, etc.); region boundary information (e.g., spatial location, morphological characteristics, etc. of the region).

[0072] Based on this information, the grayscale mapping strategy will select or adjust the grayscale mapping function for each region, thereby ensuring that the processing results can accurately reflect the grayscale characteristics within the region.

[0073] For coordinating grayscale mapping strategies between adjacent regions, especially between lesion areas and surrounding normal tissue areas, the mapping strategy needs to ensure a smooth transition of grayscale data. To this end, a smoothing function is introduced to ensure the continuity of grayscale data at region boundaries. Assuming a smoothing coefficient α is introduced between adjacent regions, the smoothed grayscale value can be calculated using the following formula: G smooth =α·G mapped,1 +(1-α)·G mapped,2 Among them, G mapped,1 and G mapped,2 α represents the mapped grayscale values ​​of adjacent regions 1 and 2, respectively, and α is the smoothing coefficient, ranging from 0 to 1. This smoothing function ensures the smoothness of grayscale value changes at region boundaries, avoiding abrupt changes.

[0074] A grayscale mapping strategy for each region is applied to the grayscale feature data to generate the final region-specific grayscale mapping data. The grayscale mapping data for each region includes not only the grayscale value itself but also the spatial location information of that region. This data accurately reflects the grayscale characteristics of the lesion area and surrounding tissues, providing precise data support for subsequent analysis, identification, and diagnosis.

[0075] Through a region-adaptive grayscale mapping mechanism, this embodiment can flexibly select or adjust grayscale mapping strategies based on the grayscale characteristics of different regions, generating region-differentiated grayscale mapping data. This mechanism can effectively eliminate grayscale differences between different regions, ensuring the continuity and consistency of image data. In particular, different mapping strategies and grayscale smoothing processing are used between lesion areas and normal tissue areas to ensure smooth data transition and accurate mapping, thereby providing high-quality grayscale feature data for subsequent lesion analysis and feature extraction.

[0076] S5. Based on the physical optical properties of the image, a multi-dimensional gray-scale mapping mechanism with layered optical response characteristics is applied to the gray-scale mapping data to perform layered analysis of the reflection, transmission and scattering components of the image and generate multi-dimensional gray-scale mapping data.

[0077] Furthermore, the multidimensional grayscale mapping mechanism for layered optical response features includes: optically dividing the image data of the lesion area and surrounding tissue area, and extracting the reflection, transmission, and scattering components; establishing an optical layered response matrix to parameterize the optical characteristics of each layer, including brightness, grayscale distribution, and spatial location information; generating a multidimensional grayscale mapping model based on the optical layered response matrix, with each dimension corresponding to different optical layering characteristics of the image; matching the multidimensional grayscale mapping model with the grayscale feature data of the lesion, and mapping the data of each layer to generate multidimensional grayscale data; and uniformly encoding and identifying the mapping data of each layer to form a structured multidimensional grayscale mapping dataset.

[0078] Furthermore, the layered analysis includes: performing optical channel separation on the input image data to extract data of reflection, transmission, and scattering components respectively; performing spatial filtering and feature extraction on each optical component to obtain grayscale distribution, brightness values, and region boundary information; establishing a layered data structure for the extracted optical component information to record the spatial location, grayscale range, and relative relationship of each layer; and uniformly identifying and archiving the layered data to generate a structured layered dataset that can be used by a multidimensional grayscale mapping model.

[0079] Specifically, through the multi-dimensional grayscale mapping mechanism of layered optical response characteristics, the reflection, transmission and scattering components of an image can be analyzed layer by layer, and corresponding multi-dimensional grayscale mapping data can be generated.

[0080] First, the image data of the lesion area and its surrounding tissue area are optically layered. Specifically, by extracting the reflection, transmission, and scattering components of the image, each layer can be independently processed. By establishing an optical layer response matrix, the optical characteristics of each layer can be parametrically represented, including brightness, grayscale distribution, and spatial location information. The optical layer response matrix can be constructed using the following formula: R = {R1, R2, ..., R...} n}; where R is the optical hierarchical response matrix, R i This represents the optical characteristic matrix of the i-th layer, including brightness, grayscale distribution, spatial location information, etc.

[0081] A multidimensional grayscale mapping model is generated using an optical layered response matrix. Each dimension corresponds to different optical layering characteristics of the image. The goal of this process is to ensure that the data of each layer is independent and can effectively represent its corresponding grayscale information based on the characteristics of different optical layers. Through a mapping strategy, the grayscale data of each layer is mapped, and corresponding multidimensional grayscale data is generated based on the physical characteristics of the image.

[0082] During multidimensional grayscale mapping, all mapping data is matched with lesion grayscale feature data. Consistency and continuity of data are ensured during grayscale data mapping at each layer. Specifically, by applying grayscale mapping strategies for each layer, differentiated grayscale mapping data for each region can be generated. In this process, the grayscale mapping strategy must be dynamically adjusted according to the different characteristics of surrounding tissues and lesion areas.

[0083] The grayscale mapping strategy needs to be adjusted for different regions. By classifying and labeling the lesion area and surrounding tissue areas, each region is defined as an independent processing unit. These independent units are assigned candidate grayscale mapping parameters, including grayscale range, grayscale distribution function, and region boundary information. In this process, the structural characteristics and grayscale distribution properties of the region are the key factors determining the mapping strategy.

[0084] Based on these parameters, each grayscale data point within a region will be mapped to a uniform format. Simultaneously, grayscale mapping strategies for adjacent or connected regions will be coordinated to ensure the continuity and consistency of the overall image data. Ultimately, the grayscale mapping strategies for all regions will be applied to the grayscale feature data, generating a regionally differentiated grayscale mapping dataset.

[0085] To ensure the efficiency of this process, all layered grayscale data and mapping information will be uniformly encoded and labeled to generate a structured multidimensional grayscale mapping dataset. This dataset will effectively support subsequent image analysis and lesion detection.

[0086] In the specific implementation process, optical channel separation is performed on the image data to independently extract the reflection, transmission, and scattering components. The data for each optical component undergoes spatial filtering and feature extraction to obtain its grayscale distribution, brightness values, and region boundary information. This information is then used to establish a hierarchical data structure, recording the spatial location, grayscale range, and relative relationships of each layer. Finally, this hierarchical data is uniformly identified and archived, forming a structured hierarchical dataset that can be used by multidimensional grayscale mapping models.

[0087] Through the above steps, this embodiment can accurately restore the physical and optical characteristics of the image during the multi-level grayscale mapping process, ensuring that the grayscale mapping data can fully reflect the grayscale characteristics of the lesion area and its surrounding tissues, and providing accurate data support for subsequent lesion identification and detection.

[0088] S6. Standardize the multidimensional grayscale mapping data of each sequence and each tissue region according to the preset sequence and tissue region order to form a grayscale feature dataset of lesions for subsequent lesion evaluation and analysis.

[0089] Furthermore, the preset sequence and tissue region order include: establishing a preset sequence order table and tissue region order table based on the image type and acquisition sequence; numbering and sorting each sequence image data according to the sequence order; and identifying and arranging the tissue region data where the lesion is located according to the tissue region order.

[0090] Furthermore, the formation of the lesion grayscale feature dataset includes: hierarchical integration of multidimensional grayscale mapping data from different sequences and different tissue regions; establishment of a data structure corresponding to the lesion region and spatial location; and unified encoding and storage of the integrated lesion grayscale feature data to form a standardized dataset that can be used for subsequent analysis.

[0091] Specifically, preset sequence order tables and tissue region order tables were established for the image sequences and tissue region sequences. The preset sequence order tables are set according to the image type and acquisition order to ensure the temporal and logical order of the image data; while the tissue region order tables are set according to the structural characteristics of the lesion and surrounding tissues, facilitating the arrangement and identification of image data from different tissue regions in a specific order. Based on this, each sequence of image data is numbered and sorted to ensure that the order of the image data is consistent with its acquisition time, location, and other information; at the same time, for the data of each tissue region, it is identified and arranged according to its position and relative relationship in the image to ensure the accuracy and structure of the data of the lesion area.

[0092] Multidimensional grayscale mapping data from different sequences and tissue regions are hierarchically integrated. Specifically, a complete grayscale data structure is established by combining the grayscale data of each sequence with the grayscale data of its corresponding tissue region. This structure reflects the grayscale characteristics and spatial relationship between the lesion area and surrounding tissues. During this process, all integrated data are standardized through unified encoding and storage to form a unified lesion grayscale feature dataset. This dataset will provide stable and reliable basic data for subsequent analysis and lesion assessment.

[0093] By establishing a data structure corresponding to the lesion area and its spatial location, the spatial consistency of grayscale data is ensured. This data structure not only records the values ​​of the grayscale data itself, but also includes the location, boundaries, and order of each data point with its corresponding tissue region in the image sequence. This structured data ensures that the grayscale characteristics of the lesion area and its related tissues can be accurately located during subsequent analysis, thereby improving the accuracy of the analysis.

[0094] Through the above steps, the resulting grayscale feature dataset for lesions not only includes multidimensional grayscale data from different sequences and tissue regions, but also achieves data consistency, completeness, and standardization. After processing, this dataset can be used for subsequent lesion assessment and analysis, providing precise grayscale feature support, making lesion localization and assessment more reliable and accurate.

[0095] By standardizing grayscale mapping data, the effective integration of multi-sequence and multi-region grayscale data was ensured, providing accurate and structured data support for subsequent analysis. This standardization method solves the problem of variability in grayscale data from different sequences and regions, and improves the accuracy of analysis results while maintaining data integrity.

[0096] Example 2:

[0097] Lesion grayscale information typically relies on visual observation by radiologists. Furthermore, there is often a lack of standardized grayscale processing methods for image data acquired from different image types (such as CT, MR, B-US, etc.), leading to inaccurate lesion assessment and difficulty in handling large-scale image analysis needs. To address these issues, this invention provides a lesion assessment method based on multi-source data grayscale features, the structure of which is as follows: Figures 1-12 As shown. The specific implementation process of this method is as follows:

[0098] in: Figure 4 Right ovarian teratoma. A 16-level grayscale scoring method was used to score the three different echo intensities in descending order from "①" to "③", where "M" represents normal musculoskeletal tissue;

[0099] Figure 5 Osteosarcoma of the distal metaphysis of the right femur. Two different densities were arranged in descending order from "①" to "②" using a grading system, where "M" represents normal skeletal muscle.

[0100] Figure 6 Partial malignant transformation of cirrhotic nodules. In the arterial phase scan, the two different densities are scored in descending order from "①" to "②". The scoring criteria for the same area in the portal venous phase, delayed phase and plain scan are the same. Normal muscle tissue is represented by "M" in each phase.

[0101] Figure 7 Type III astrocytoma of the left basal ganglia. A scoring method was used to score three different signs in the enhanced scan images in descending order from "①" to "③". The corresponding regions in the other six sequences were scored consistently. All sequences were represented by "M" to indicate normal muscle tissue.

[0102] Specifically, by simulating the human eye's perception of grayscale in lesions, this method uses sixteen grayscale values ​​(0-15) to represent the grayscale information of different lesion components. Specifically, the darkest grayscale value corresponds to 0, the lightest to 15, and the intermediate grayscale values ​​are distributed proportionally. The core advantage of this method lies in its ability to convert image grayscale information into easily quantifiable digital data, avoiding the vague descriptions in traditional image analysis and thus achieving a more accurate assessment of the lesion's physical components.

[0103] In practical implementation, the first step is to standardize image data acquired from different imaging types, such as CT, MR, and B-US. Because images from different imaging devices have significantly different grayscale resolutions and features, directly performing grayscale analysis on image data from different sources may lead to errors. Therefore, this invention establishes a unified grayscale standard, enabling all image data to be converted at the same grayscale scale. Through this standardization process, all data from different imaging devices and different scanning sequences are unified into grayscale numbers from 0 to 15, ensuring the consistency and comparability of different image data.

[0104] In the process of lesion analysis, the first step is to extract image data of the lesion area and digitize the grayscale information within that area. Different types of lesions, such as lipomas, lymphomas, and meningiomas, exhibit different grayscale characteristics. By converting different lesion components, such as myxoid degeneration, calcification, and hemorrhage, into digital grayscale values, the structural and compositional information of the lesion can be accurately captured. During processing, this invention simulates the grayscale resolution capability of the human eye, dividing the lesion portion in the image into 16 grayscale levels. This not only effectively records subtle grayscale differences in the image but also improves the distinguishability between lesions and pathological components. Especially for lesions with small grayscale differences but representing different pathological components, digital processing effectively reduces the subjective errors that may occur in traditional visual observation.

[0105] Because muscle tissue is widely distributed in images and its grayscale changes are relatively stable, muscle grayscale can be used as a standard for adjusting digital errors. This adjustment mechanism can eliminate grayscale differences caused by different devices and scanning settings, thereby ensuring the accuracy of lesion assessment.

[0106] The grayscale digitization method of this invention is also adaptable to the analysis of multiple lesions. When multiple lesions exist in an image, the system can select three lesions (large, medium, and small) for grayscale analysis as needed, further enhancing the flexibility and adaptability of the method. In the case of multiple lesions, the system can accurately analyze the grayscale components of each lesion and provide a reference for subsequent pathological diagnosis.

[0107] By uniformly processing grayscale data from different image types, effective comparison of image data from different sources becomes possible. For example, DR images have low grayscale resolution, typically only able to distinguish tissue components such as gas, fat, soft tissue, bone, or calcification; therefore, lesion diagnosis mainly relies on the morphological characteristics of the lesion. B-US images, on the other hand, can better distinguish organs and their functional morphology, the softness of lesions, and blood flow direction, exhibiting relatively strong grayscale resolution. CT images can differentiate components such as liquids, gases, fat, and calcification, and reveal grayscale changes of different components within the lesion through contrast-enhanced scanning. MR images possess the strongest tissue resolution, especially when subjected to contrast-enhanced scanning with different sequences, providing richer diagnostic information. By converting the grayscale of image data into a unified 0-15 digit, this invention overcomes the problem of grayscale inconsistency, ensuring comparability and uniformity between different image data.

[0108] Especially when using MR images, recording grayscale changes at different scanning phases can effectively reflect changes in the solid components of lesions. For example, by digitizing the grayscale results of enhanced scans at different stages such as the arterial, portal venous, and venous phases, the dynamic changes of lesions can be revealed, providing more detailed diagnostic information. Furthermore, the system can automatically compare this grayscale information with known pathological image databases, enabling AI-assisted diagnosis and providing doctors with more accurate lesion assessment results.

[0109] In summary, by converting grayscale information from different image types into a unified digital form, not only is the accuracy and consistency of lesion assessment improved, but it also effectively processes large-scale image data, adapting to the complex analytical needs of multi-lesion and multi-source data. This method not only draws on the experience of traditional imaging diagnosis but also combines modern digital technology, promoting the transformation of imaging diagnosis towards AI-driven intelligent development.

[0110] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A lesion assessment method based on grayscale features of multi-source data, characterized in that, Includes the following steps: S1. Collect and receive multi-source image data of the object to be analyzed, and uniformly identify and classify the data to form a multi-source image dataset; S2. Preprocess the multi-source image dataset to determine the normal tissue area and lesion area in the image, and extract the spatial location and morphological information of the lesion. S3. The grayscale information of the lesion area is digitally processed to generate grayscale feature data of the lesion in a unified format; S4. Apply a region-adaptive grayscale mapping mechanism to the grayscale feature data of the lesions, select or adjust the grayscale mapping strategy according to the characteristics of different lesion internal and surrounding tissue regions, and generate region-differentiated grayscale mapping data. S5. Based on the physical optical characteristics of the image, a multi-dimensional gray-scale mapping mechanism with layered optical response characteristics is applied to the gray-scale mapping data to perform layered analysis of the reflection, transmission and scattering components of the image and generate multi-dimensional gray-scale mapping data. S6. Standardize the multidimensional grayscale mapping data of each sequence and each tissue region according to the preset sequence and tissue region order to form a grayscale feature dataset of lesions for subsequent lesion evaluation and analysis.

2. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The formation of the multi-source image dataset includes: Collect multi-source image data of the object to be analyzed, including image data from different image modalities or different imaging devices; The multi-source image data is identified and classified, and each image data is associated with its source information, acquisition time, sequence type and scanning parameters; The classified image data is processed to unify the format, including standardization of image resolution, pixel grayscale range and storage format; Based on the image acquisition conditions and the information of the inspected object, the spatial position and orientation differences of the images are corrected to achieve spatial registration of multi-source images; The spatially registered image data undergoes integrity checks and data completion, including anomaly identification and repair, to form a unified multi-source image dataset.

3. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The determination of normal tissue areas and lesion areas in the image includes: The multi-source image data is preprocessed, including denoising, contrast enhancement, and standardization of grayscale range; The image data is divided into pixel-level or region-level segments using image segmentation algorithms to distinguish different tissue regions and candidate lesion regions; Based on tissue boundaries, morphological characteristics, and spatial relationships, the segmented areas are classified and marked as normal tissue areas or lesion areas. Extract the spatial location, morphological features, and boundary information of the lesion area, and associate and store them with the multi-source image dataset.

4. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The digital processing includes: The grayscale values ​​within the lesion area are standardized to unify the grayscale range and eliminate differences between different image sources. According to the preset grayscale grading scheme, the standardized grayscale values ​​are mapped to digital representations to form grayscale data of lesions in a unified format. The digitized grayscale data of lesions is structured and organized to form grayscale feature data of lesions according to the spatial distribution of lesion areas and the order of image sequences.

5. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The region adaptive grayscale mapping mechanism includes: The lesion and its surrounding tissue area were divided into several independent regions, and each region was labeled. The grayscale data of each independent region is processed regionally to generate grayscale mapping data for each region.

6. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The strategy for selecting or adjusting grayscale mapping includes: The lesion's internal and surrounding tissue areas are classified and labeled, with each area defined as an independent treatment unit; Establish a candidate set of grayscale mapping parameters for each region, including grayscale range, grayscale distribution function and region boundary information; Based on the structural characteristics and grayscale distribution of the region, candidate parameters are selected or adjusted to form a grayscale mapping strategy for each region. Coordinate the grayscale mapping strategies for adjacent or connected regions to ensure the continuity and consistency of grayscale data; The grayscale mapping strategy for each region is applied to the grayscale feature data to generate region-differentiated grayscale mapping data.

7. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The multidimensional grayscale mapping mechanism of the hierarchical optical response characteristics includes: Optical layering was performed on the imaging data of the lesion area and surrounding tissue areas to extract the reflection, transmission and scattering components. An optical layer response matrix is ​​established, and the optical characteristics of each layer are parameterized, including brightness, grayscale distribution and spatial location information. A multidimensional grayscale mapping model is generated based on the optical layer response matrix, with each dimension corresponding to different optical layer characteristics of the image. The multidimensional grayscale mapping model is matched with the grayscale feature data of the lesions, and the data of each layer is mapped to generate multidimensional grayscale data. The mapping data of each layer are uniformly encoded and identified to form a structured multidimensional grayscale mapping dataset.

8. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The hierarchical parsing includes: Optical channel separation is performed on the input image data to extract data of reflection, transmission and scattering components respectively; Spatial filtering and feature extraction are performed on each optical component to obtain grayscale distribution, brightness value and region boundary information; A hierarchical data structure is established for the extracted optical component information, and the spatial location, grayscale range and relative relationship of each layer are recorded. The hierarchical data is uniformly identified and archived to generate a structured hierarchical dataset that can be used by a multidimensional grayscale mapping model.

9. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The preset sequence and organization region order include: Establish a preset sequence order table and tissue region order table based on image type and acquisition sequence; Each sequence of image data is numbered and sorted according to the sequence order; Data on the tissue regions where the lesions are located are identified and arranged according to the order of the tissue regions.

10. The lesion assessment method based on multi-source data grayscale features according to claim 1, characterized in that, The dataset containing grayscale features of the lesions includes: Hierarchical integration of multidimensional grayscale mapping data from different sequences and different tissue regions; Establish a data structure corresponding to the lesion area and its spatial location; The integrated grayscale feature data of lesions are uniformly encoded and stored to form a standardized dataset that can be used for subsequent analysis.

Images