Medical image anatomy coverage intelligent evaluation method and system

Abstract

The invention relates to the technical field of medical image data processing, in particular to a medical image anatomy coverage intelligent evaluation method and system, and the method comprises the steps: obtaining an original medical image, and carrying out the preprocessing of the original medical image, and obtaining a three-dimensional target medical image; according to preset segmentation mask labels, voxel-by-voxel classification is carried out on the target medical image to obtain a mask file, the mask file and the target medical image are the same in size, and each segmentation mask label corresponds to one anatomical structure; based on the segmentation mask labels, voxels of the target anatomical structure corresponding to the segmentation mask labels are screened, and the volume of the target anatomical structure is obtained; according to the target anatomical structure volume and the standard reference volume of the target anatomical structure, the target coverage rate is obtained, and the anatomical coverage evaluation efficiency and accuracy are improved.

Description

Technical Field

[0001] This application relates to the field of medical image data processing technology, and in particular to a method and system for intelligent assessment of anatomical coverage in medical images. Background Technology

[0002] With the rapid development of medical technology, medical imaging plays a crucial role in disease diagnosis. In clinical image quality control and research project management, assessing the anatomical coverage of computed tomography (CT) / magnetic resonance imaging (MRI) scans is a core element in ensuring medical quality, improving research effectiveness, and optimizing the allocation of medical resources.

[0003] Anatomical coverage assessment is a fundamental prerequisite for ensuring that imaging diagnosis is based on evidence and without omissions. In clinical diagnosis and treatment, doctors need to determine the location, extent, and relationship with surrounding tissues of lesions based on complete anatomical images. Incomplete scan coverage may directly lead to diagnostic errors.

[0004] Currently, anatomical coverage assessment mostly relies on radiologists reviewing images layer by layer to determine whether the target anatomical structure is covered. This method is time-consuming and highly subjective, and there is an urgent need to propose an efficient and reliable method for anatomical coverage assessment. Summary of the Invention

[0005] To improve the efficiency and accuracy of anatomical coverage assessment, this application provides an intelligent assessment method and system for anatomical coverage in medical images.

[0006] Firstly, this application provides an intelligent assessment method for anatomical coverage in medical images, employing the following technical solution:

[0007] A method for intelligent assessment of anatomical coverage in medical images, the method comprising:

[0008] The original medical images are acquired, and the original medical images are preprocessed to obtain a three-dimensional target medical image.

[0009] According to the preset segmentation mask labels, the target medical image is classified on a voxel-by-voxel basis to obtain a mask file, wherein the mask file and the target medical image have the same size, and each segmentation mask label corresponds to an anatomical structure;

[0010] Based on the segmentation mask label, the voxels of the target anatomical structure corresponding to the segmentation mask label are selected to obtain the volume of the target anatomical structure;

[0011] The target coverage is obtained based on the target anatomical structure volume and the standard reference volume of the target anatomical structure.

[0012] In one possible implementation, obtaining the target coverage based on the target anatomical structure volume and a standard reference volume of the target anatomical structure includes:

[0013] Determine whether the volume of the target anatomical structure is smaller than the standard reference volume corresponding to the target anatomical structure.

[0014] If so, calculate the ratio of the target anatomical structure volume to the standard reference volume of the target anatomical structure to obtain the target coverage.

[0015] If not, then the target coverage rate is set to 1.

[0016] In one possible implementation, obtaining the target coverage based on the target anatomical structure volume and a standard reference volume of the target anatomical structure includes:

[0017] Acquire multiple target medical images from multiple scan sequences of the same user, and extract the target anatomical structures and corresponding target coverage values;

[0018] Determine the number of target coverage values ​​for the same anatomical structure.

[0019] When there is only one target coverage value, the target coverage value is determined to be the comprehensive coverage value of the target's anatomical structure.

[0020] When the number of target coverage values ​​is greater than or equal to one, the maximum value among the multiple target coverage values ​​is determined as the comprehensive coverage value of the target anatomical structure.

[0021] In one possible implementation, determining the comprehensive coverage value includes:

[0022] Based on the mapping relationship between segmentation mask labels and region labels, the anatomical regions of the target anatomical structures in the target medical image are obtained, wherein each anatomical region includes at least one target anatomical structure.

[0023] Based on the user's preset inspection area and the anatomical structure to be inspected, determine whether the target anatomical structure and corresponding area labels of the user are completely covered, so as to generate a complete result;

[0024] Output a comprehensive report, wherein the comprehensive report includes at least one of user information, integrity results, region labels, target anatomical structures, and comprehensive coverage values.

[0025] In one possible implementation, acquiring the original medical image and preprocessing the original medical image to obtain a three-dimensional target medical image includes:

[0026] Acquire raw medical images, wherein the raw medical images include metadata of multi-layer slices, and the metadata includes, but is not limited to, pixel data;

[0027] The original medical image is converted into a first medical image in a preset format, wherein the first medical image is three-dimensional and has image spatial coordinate information;

[0028] The first medical image is resampled according to a preset resolution to obtain a second medical image;

[0029] Based on the preset window width and window level, the pixel values ​​of the second medical image are normalized to a preset range to obtain the target medical image.

[0030] In one possible implementation, based on the segmentation mask label, voxels corresponding to the target anatomical structure are filtered to obtain the volume of the target anatomical structure, including:

[0031] Obtain the voxels corresponding to the same segmentation mask label;

[0032] Based on the number and volume of voxels, the volume of the target anatomical structure corresponding to the segmentation mask label in the target medical image is calculated, wherein the volume of voxels is obtained according to the resolution of the target medical image.

[0033] In one possible implementation, obtaining the voxel corresponding to the same segmentation mask label includes:

[0034] Obtain the 3D coordinates of the voxels corresponding to the same segmentation mask label to obtain the bounding box range;

[0035] The number of voxels is obtained based on the bounding box range corresponding to the segmentation mask label.

[0036] Secondly, this application provides an intelligent assessment system for anatomical coverage in medical images, employing the following technical solution:

[0037] A medical imaging anatomical coverage intelligent assessment system, the system comprising:

[0038] An image preprocessing module is used to acquire raw medical images and preprocess the raw medical images to obtain a three-dimensional target medical image.

[0039] The mask classification module is used to classify the target medical image on a voxel-by-voxel basis according to a preset segmentation mask label to obtain a mask file, wherein the mask file and the target medical image have the same size, and each segmentation mask label corresponds to an anatomical structure;

[0040] The volume calculation module is used to filter the voxels of the target anatomical structure corresponding to the segmentation mask label based on the segmentation mask label, and obtain the volume of the target anatomical structure.

[0041] The coverage calculation module is used to obtain the target coverage based on the target anatomical structure volume and the standard reference volume of the target anatomical structure.

[0042] Thirdly, this application provides an intelligent assessment device for anatomical coverage in medical images, employing the following technical solution:

[0043] A medical imaging anatomical coverage intelligent assessment device includes: a memory and a processor;

[0044] The memory stores computer-executed instructions;

[0045] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0046] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0047] A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible embodiments of the first aspect as described above.

[0048] This application provides a method and system for intelligent assessment of anatomical coverage in medical images. By preprocessing the original medical images, a three-dimensional target medical image is obtained, eliminating image format differences and data heterogeneity caused by different devices and scanning parameters, thus improving the usability of the target medical image. The target medical image is then classified voxel-by-voxel using preset segmentation mask labels, where each segmentation mask label corresponds to a unique anatomical structure, achieving objective and accurate localization of the anatomical structure. The volume of the target anatomical structure is then calculated by filtering the voxels corresponding to the mask labels, and the target coverage is calculated by comparing it with a standard reference volume to obtain a quantitative result of the coverage degree. This method, through objective and accurate localization and quantitative results, reduces the subjective errors and low efficiency of manual identification, making the coverage assessment process more efficient and the assessment results more accurate. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating an intelligent assessment method for anatomical coverage in medical images, provided as an embodiment of this application.

[0050] Figure 2 This is a flowchart illustrating an intelligent assessment method for anatomical coverage in medical images, provided as an embodiment of this application.

[0051] Figure 3 This is a schematic diagram of the structure of a medical imaging anatomical coverage intelligent assessment system provided in one embodiment of this application.

[0052] Figure 4 This is a schematic diagram of the structure of a medical imaging anatomical coverage intelligent assessment device provided in one embodiment of this application. Detailed Implementation

[0053] To better understand the purpose, technical solutions, and advantages of this application, it has been described and illustrated below with reference to the accompanying drawings and embodiments. However, those skilled in the art should understand that this application can be implemented without these details. In some cases, to avoid obscuring various aspects of this application due to unnecessary description, well-known methods, processes, systems, components, and / or circuits already described at a higher level will not be elaborated upon. It will be apparent to those skilled in the art that various modifications can be made to the embodiments disclosed in this application, and the general principles defined in this application can be applied to other embodiments and application scenarios without departing from the principles and scope of this application. Therefore, this application is not limited to the illustrated embodiments, but conforms to the broadest scope consistent with the scope of protection claimed in this application.

[0054] It should be noted that the descriptions of these embodiments are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0055] It should be understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or any combination thereof. For hardware implementations, the processor may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to perform the functions described herein, or combinations thereof.

[0056] When an embodiment is implemented as software, firmware, middleware, or microcode, program code, or code segments, it may be stored in a machine-readable medium, such as a storage component. A code segment may represent a procedure, function, subroutine, program, routine, subroutine, module, software package, class, or any combination of instructions, data structures, or program statements. One code segment can be coupled to another code segment or hardware circuitry by passing and / or receiving information, data, arguments, parameters, or memory contents. Information, arguments, parameters, data, etc., can be passed, forwarded, or transmitted using any suitable means, including memory sharing, messaging, token passing, network transmission, etc.

[0057] For software implementations, the techniques described herein can be implemented using modules (e.g., programs, functions, etc.) that perform the functions described herein. The software code can be stored in memory units and executed by a processor. The memory units can be implemented within or outside the processor; in the latter case, the memory units can be communicatively coupled to the processor via various methods known in this art.

[0058] In the description of this application, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0059] In the description of this application, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples.

[0060] The traditional method, which relies on radiologists to examine CT / MRI images layer by layer to determine whether the target anatomical region is covered, is not only time-consuming and highly subjective, but also suffers from inconsistent standards for "complete coverage" among different doctors and a lack of quantitative indicators. Furthermore, existing automated methods can usually only determine whether a certain anatomical region is covered or not, and cannot refine the coverage of specific organs (such as the liver or spleen), resulting in inaccurate assessments of anatomical region coverage. Moreover, regardless of whether relying on radiologists or existing automated methods, incomplete coverage can only be detected after the patient has completed the scan, requiring the patient to undergo a re-examination, increasing medical costs and radiation exposure.

[0061] To address the aforementioned technical challenges, a smart assessment method for anatomical coverage in medical images is proposed. This method eliminates the need for manual coverage assessment, provides a unified evaluation standard, quantifies the assessment results, and improves the accuracy, objectivity, and efficiency of coverage assessment.

[0062] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0063] Figure 1 A flowchart illustrating an intelligent assessment method for anatomical coverage in medical images, provided as an embodiment of this application; Figure 1 As shown in the embodiments of this application, a method for intelligent assessment of anatomical coverage in medical images is disclosed. The method includes:

[0064] S101. Acquire the original medical images and preprocess them to obtain the target medical image in three dimensions.

[0065] Original medical images are medical images captured and output by medical imaging equipment, which typically includes CT scanners, MRI scanners, etc. Original medical images usually refer to CT or MRI images.

[0066] In clinical settings, raw medical images are typically in DICOM format (Digital Imaging and Communication Standard). DICOM is the default output format for medical imaging equipment. It contains pixel data from multiple slices, slice thickness, slice spacing, pixel spacing, patient information, scanning equipment parameters, and scanning time.

[0067] Anatomical structures refer to the collective term for the shape, location, and interrelationships of the various organs and tissues that make up the human body. They are the basic structural units of a normally functioning human body. Anatomical structures can be classified according to body parts, organ functions, and imaging recognition accuracy.

[0068] For example, based on morphological characteristics and core functions, anatomical structures can be divided into five major categories:

[0069] Skeletal system: cervical vertebrae (C1-C7), thoracic vertebrae (T1-T12), lumbar vertebrae (L1-L5), sacrum, pelvis, etc.;

[0070] Solid organs: liver, spleen, left kidney, right kidney, pancreas, gallbladder, etc.

[0071] Hollow organs: stomach, duodenum, small intestine, colon, bladder, etc.

[0072] Vascular system: aorta, inferior vena cava, portal vein, iliac vessels, etc.;

[0073] Other structures: heart, lungs, esophagus, trachea, etc.

[0074] Since most anatomical structures are three-dimensional, preprocessing such as format conversion and resampling of the original DICOM format medical images restores the original medical images to three-dimensional anatomical structures, reducing the structural misalignment caused by disordered slice order in the original medical images. Moreover, the scanning parameters of the same patient or different patients may vary due to differences in equipment models and examination purposes. After processing the original medical image data, a three-dimensional target medical image is obtained, which completely restores the spatial distribution relationship of the anatomical structures.

[0075] In this embodiment, the original medical image in DICOM format is preprocessed to obtain the target medical image in NIfTI format for subsequent processing.

[0076] S102. Based on the preset segmentation mask labels, classify the target medical image voxel by voxel to obtain the mask file.

[0077] TotalSegmentator deep learning model is a high-precision anatomical structure segmentation model in the field of medical imaging. It can automatically segment 104 anatomical structures in the human body, covering the key structures required for core clinical assessment scenarios.

[0078] The preprocessed NIfTI format target medical image is input into the TotalSegmentator deep learning model. By using preset segmentation mask labels, the target medical image can be classified on a voxel-by-voxel basis, and the output is a mask file of the target medical image after classification and labeling based on the preset segmentation label mask.

[0079] The mask file and the target medical image are the same size, and each segmentation mask label corresponds to an anatomical structure.

[0080] Regarding the preset category mask labels, for example,

[0081] Tag 1 = Liver;

[0082] Tag 2 = Spleen;

[0083] Tag 3 = Left kidney;

[0084] Tag 4 = Right kidney;

[0085] Tag 5 = Gallbladder;

[0086] Tag 6 = Stomach;

[0087] Tag 7 = Pancreas;

[0088] Tag 17 = Heart.

[0089] A total of 104 classification mask labels were set, covering 104 anatomical structures of the human body, with a value of 0 representing the background.

[0090] A voxel is a volume element in three-dimensional space. By classifying each voxel of a target medical image in three-dimensional form according to its corresponding segmentation mask label, the anatomical structure corresponding to each voxel in the three-dimensional matrix can be labeled. By extracting the classification mask label corresponding to each anatomical structure, the division and localization of the corresponding anatomical structure can be achieved.

[0091] S103. Based on the segmentation mask label, filter the voxels of the target anatomical structure corresponding to the segmentation mask label to obtain the volume of the target anatomical structure.

[0092] Since each voxel is labeled with a segmentation mask based on the anatomical structure, by filtering voxels of the target anatomical structure, the target anatomical structure can be separated from the target medical image, and the abstract image structure can be transformed into computable volume data to obtain the volume of the target anatomical structure.

[0093] S104. Based on the target anatomical structure volume and the standard reference volume of the target anatomical structure, obtain the target coverage.

[0094] The standard reference volume is a representative normal volume benchmark value determined based on an anatomical database and extensive statistical analysis of the physical volume of specific organs in healthy individuals.

[0095] Since the standard reference volume already covers the differences in organ volume among people of different heights, weights, and young people, using the standard reference volume as a benchmark to calculate the target coverage of the target anatomical structure volume can objectively quantify the coverage assessment results.

[0096] For example, if the standard reference volume of the liver is 1200-1600 ml, we can define the target coverage of the liver as 1350 ml when the liver volume detected in the image is 1350 ml, with the average standard reference volume of 1500 ml as the baseline.

[0097] This application provides an intelligent assessment method for anatomical coverage of medical images. It preprocesses the original medical images to obtain target medical images with unified standards and formats. The method then uses a TotalSegmentator deep learning model and pre-set segmentation mask labels to classify the target medical images voxels individually, labeling the anatomical structures corresponding to each voxel. By extracting the segmentation mask labels corresponding to the target anatomical structures, the volume of the target anatomical structures is divided and located. By comparing the volume of the target anatomical structures with the standard reference volume, a quantitative result of the target coverage can be obtained, reducing subjective human intervention and achieving accuracy and objectivity in coverage assessment based on unified evaluation standards.

[0098] Figure 2 A flowchart illustrating an intelligent assessment method for anatomical coverage in medical images, as provided in one embodiment of this application, is shown below. Figure 2 As shown, based on the above embodiments, this embodiment includes the following method:

[0099] S201. Obtain the original medical images.

[0100] S202. Convert the original medical image into a first medical image in a preset format.

[0101] Tools such as MONAI and SimpleITK were used to convert raw medical images in DICOM format into first-order medical images in NIfTI format. The first-order medical images are three-dimensional and contain spatial coordinate information, which facilitates subsequent voxel-level calculations.

[0102] S203. Resample the first medical image according to the preset resolution to obtain the second medical image.

[0103] The default standard resolution is equidistant resolution, meaning that the length, width, and height of the voxels are the same.

[0104] For example, if the default standard resolution is set to 1.5mm, then the volume of a 1.5mm isometric voxel = 1.5. 3 =3.375m 3 .

[0105] First, the first medical image needs to be read and its original spatial information extracted, that is, the original resolution and original size of the first medical image need to be obtained.

[0106] Then, the target size after resampling is calculated, which is the number of voxels in the new image. The physical spatial range of the image remains unchanged, such as the total length in the x-direction = original resolution * original size. Only the voxel size (resolution) and the number of voxels (size) are adjusted. The formula for calculating the target size is:

[0107] Target size = Original physical size / Preset standard resolution;

[0108] Wherein, the original physical range = original resolution × original size (each dimension is calculated separately), and the result needs to be rounded down, that is, the number of voxels is an integer.

[0109] For example, if the first medical image is an NIfTI file of an abdominal CT scan,

[0110] Original resolution = (0.8, 0.8, 2.0) (x = 0.8mm, y = 0.8mm, z = 2.0mm, i.e., planar resolution 0.8mm, layer thickness 2mm, non-equidistant).

[0111] Original size = (512, 512, 300) (512 voxels in the x direction, 512 voxels in the y direction, and 300 slices in the z direction).

[0112] The default standard resolution is 1.5mm equidistant.

[0113] Calculate the original physical dimensions (x, y, z):

[0114] The physical range in the x-direction is 0.8mm × 512 = 409.6mm;

[0115] The physical range in the y-direction is 0.8mm × 512 = 409.6mm.

[0116] The physical range in the z-direction is 2.0mm × 300 = 600mm;

[0117] Calculate the target size:

[0118] Target size x = 409.6 ÷ 1.5 ≈ 273 (rounded down);

[0119] Target size y = 409.6 ÷ 1.5 ≈ 273 (rounded down);

[0120] Target size z = 600 ÷ 1.5 = 400 (rounded down);

[0121] Therefore, the target size is (273, 273, 400), which means the number of voxels in the second medical image is 273 × 273 × 400.

[0122] Then, by calling the resampling function and inputting parameters such as the first medical image, preset standard resolution, target size, and interpolation algorithm, a second medical image with uniform spatial scale is generated. The interpolation algorithm is determined according to the image type.

[0123] The position of the voxels is adjusted by a spatial transformation matrix, and then the pixel value of the new voxels is calculated by an interpolation algorithm to ensure that the shape and relative position of the anatomical structure are not distorted.

[0124] Finally, the blank areas are filled with preset background fill values, such as -1024 commonly used in CT images, which corresponds to the pixel value of air and conforms to clinical imaging logic.

[0125] Resampling the first medical image according to the preset resolution can eliminate the scale heterogeneity caused by different scanning devices and parameters. By adjusting the spatial size (length, width, and height) of the image voxels, the first medical image is unified to the preset standard resolution, resulting in a second medical image with consistent structural proportions and uniform voxel volume, providing a standardized spatial scale for subsequent segmentation and volume calculation.

[0126] S204. Based on the preset window width and window level, normalize the pixel values ​​of the second medical image to a preset range to obtain the target medical image.

[0127] Window Width (WW): refers to the range of pixel values ​​displayed in an image, which determines the image contrast (the smaller the window width, the higher the contrast and the more prominent the details).

[0128] Window Level (WL): This refers to the center pixel value of the window width, which determines the brightness of the image (the higher the window level, the brighter the overall image).

[0129] The preset window width and window level are the standard for clinical images. By normalizing the pixel values ​​of the second medical image after resampling at standard resolution to the preset range, extreme pixel values ​​can be truncated, extreme interference can be eliminated, and key details can be highlighted.

[0130] Based on the standards for clinical imaging window width and level, the lower limit and upper limit of pixel values ​​can be obtained. Pixel values ​​outside the range are considered extreme values. The formulas for the lower limit and upper limit of pixel values ​​are as follows:

[0131] Lower limit of pixel value = window level - window width / 2;

[0132] Maximum pixel value = window level + window width / 2.

[0133] For example, if the preset window level and width for CT images are window level = 40 and window width = 400, then:

[0134] Lower limit of pixel value = window level - window width / 2 = 40 - 400 / 2 = 40 - 200 = -160;

[0135] Maximum pixel value = window level + window width / 2 = 40 + 400 / 2 = 40 + 200 = 240;

[0136] That is, only signals with pixel values ​​in the range of [-160, 240] are retained. Pixel values ​​outside this range (such as -1000 for air and 1000 for bones) will be truncated to -160 or 240. Values ​​below the lower limit of pixel value are truncated to the lower limit of pixel value, and values ​​above the upper limit of pixel value are truncated to the upper limit of pixel value.

[0137] The truncated pixel values ​​are mapped to a preset normalized range, commonly [0,1] or [0,255] in clinical practice. The specific formula is as follows:

[0138] Normalized pixel value = (truncated pixel value - lower limit of pixel value) / (upper limit of pixel value - lower limit of pixel value)

[0139] If the target range is [0, 255], then multiply the above formula by 255, that is:

[0140] Normalized pixel value = (truncated pixel value - lower limit of pixel value) / (upper limit of pixel value - lower limit of pixel value) × 255.

[0141] Example: Still using window level = 40 and window width = 400 as an example, when the preset range is [0,1]:

[0142] For the truncated pixel value x, the normalized value is (x-(-160)) / (240-(-160)) = (x+160) / 400.

[0143] S205. Based on the preset segmentation mask labels, classify the target medical image voxel by voxel to obtain the mask file.

[0144] The description of step S205 can be found in the description of step S102.

[0145] S206. Obtain the voxel corresponding to the same segmentation mask label.

[0146] Since each segmentation mask label corresponds to a unique anatomical structure, by filtering the same segmentation mask label, the complete spatial morphology of the target anatomical structure can be directly extracted from complex influences.

[0147] S207. Obtain the three-dimensional coordinates of the voxels corresponding to the same segmentation mask label to obtain the bounding box range.

[0148] After filtering out all voxels corresponding to the target anatomical structure from the mask file, record the spatial coordinates of these voxels in the 3D image. They are usually represented by the world coordinate system or voxel index coordinate system of the image, such as (x,y,z), where x / y corresponds to the row / column in the slice and z corresponds to the slice layer number.

[0149] By statistically analyzing the voxel coordinates of the target anatomical structure, the minimum and maximum boundaries of the bounding box in the x, y, and z dimensions are determined, i.e., the coordinates of the diagonal vertices of the bounding box. The formula is as follows:

[0150] The minimum bounding box coordinates (min_x, min_y, min_z) are the minimum values ​​of the x-coordinate, y-coordinate, and z-coordinate of all target voxels, respectively.

[0151] The maximum bounding box coordinates (max_x, max_y, max_z) are the maximum values ​​of the x-coordinate, y-coordinate, and z-coordinate of all target voxels, respectively.

[0152] Bounding box range: (max_x-min_x+1,max_y-min_y+1,max_z-min_z+1);

[0153] The x / y / z+1 is because the voxel coordinates are integer indices and must include the boundary voxels.

[0154] S208. Obtain the number of voxels based on the bounding box range corresponding to the segmentation mask label.

[0155] By using the bounding box range, voxels are traversed only within the space of min_z~max_z, min_y~max_y, and min_x~max_x to filter voxels of the target anatomical structure and count their number, thereby improving the efficiency of voxel count and avoiding traversing the entire 3D image, thus reducing computational redundancy.

[0156] S209. Based on the number and volume of voxels, calculate the volume of the target anatomical structure corresponding to the segmentation mask label in the target medical image.

[0157] Target anatomical structure volume = number of voxels × voxel volume;

[0158] The number of voxels is the number of voxels in the target anatomical structure, and the voxel volume is obtained based on the resolution of the target medical image.

[0159] S210. Determine whether the volume of the target anatomical structure is smaller than the standard reference volume corresponding to the target anatomical structure.

[0160] S211. If so, calculate the ratio of the target anatomical structure volume to the standard reference volume of the target anatomical structure to obtain the target coverage.

[0161] S212. If not, then the target coverage rate is determined to be 1.

[0162] If the volume of the target anatomical structure is greater than or equal to the standard reference volume corresponding to the target anatomical structure, from the perspective of volume, it can be determined that the core area affecting the covered target anatomical structure is determined, and the target coverage rate is determined to be 1.

[0163] If the volume of the target anatomical structure is smaller than the standard reference volume corresponding to that target anatomical structure, it means that the image does not cover the entire area of ​​the target anatomical structure, resulting in an insufficient number of voxels and an underestimation of the target volume. In this case, the ratio of the target anatomical structure volume to the standard reference volume of the target anatomical structure is calculated to obtain the target coverage rate, i.e.:

[0164] ;

[0165] in, Indicates the target coverage rate. Indicates the volume of the target anatomical structure. The standard reference volume representing the target anatomical structure.

[0166] S213. Obtain multiple target medical images from multiple scan sequences of the same user, and extract the target anatomical structures and corresponding target coverage values.

[0167] For the same user (i.e., patient), multiple scan sequences may be generated during a single examination due to various examination purposes, limited scanning range, and differences in equipment parameters. Examples include arterial / venous phase sequences for abdominal CT and T1-weighted / T2-weighted sequences for brain MRI. The target medical image of each scan sequence may only cover a portion of the structure. By extracting multiple target medical images from multiple scan sequences of the same user, the insufficient coverage of a single sequence can be compensated for, thereby improving the accuracy of the assessment.

[0168] Each user has a unique identifier, such as an ID card number, hospital number, or examination number. By obtaining all scan sequences within the same examination period for that user, such as plain scan sequences, arterial phase enhanced sequences, and venous phase enhanced sequences, each sequence corresponds to a set of DICOM files, i.e., raw medical images. After preprocessing multiple raw medical images, multiple target medical images are obtained. By extracting the target anatomical structures and corresponding target coverage values ​​from each target medical image, the insufficient coverage of a single sequence can be compensated for, and the accuracy of the assessment can be improved.

[0169] S214. Determine the number of target coverage values ​​for the same target anatomical structure.

[0170] When there is only one target coverage value, the target coverage value is determined to be the comprehensive coverage value of the target's anatomical structure.

[0171] When the number of target coverage values ​​is greater than or equal to one, the maximum value among the multiple target coverage values ​​is determined as the comprehensive coverage value of the target anatomical structure.

[0172] Multiple scan sequences for the same user, such as plain / arterial / venous phases of abdominal CT, may have different target coverage values ​​for the same anatomical structure (such as the liver) due to differences in scan range, image quality (such as artifacts), and segmentation accuracy. For example, the plain scan sequence might show 90%, the arterial phase 100%, and the venous phase 85%. Directly using multiple target coverage values ​​would lead to inconsistent evaluation results and make it impossible to determine whether the coverage is complete.

[0173] When there is only one target coverage value, the target coverage value is determined as the comprehensive coverage value of the corresponding target anatomical structure. When the target anatomical structure has multiple target coverage values, the maximum value is taken as the comprehensive coverage value. This transforms multi-dimensional data into a unique comprehensive coverage value, which avoids interference from multiple values ​​and ensures the uniformity of the evaluation benchmark.

[0174] For example, suppose a user undergoes an abdominal CT scan for liver cancer screening, generating three scan sequences. The target coverage values ​​of the liver are extracted as follows:

[0175] Sequence 1: Plain scan sequence → Liver target coverage = 90%;

[0176] Sequence 2: Arterial phase enhanced sequence → Liver target coverage = 100%;

[0177] Sequence 3: Venous phase enhanced sequence → Liver target coverage = 85%;

[0178] Therefore, the overall coverage value of the liver is determined to be 100%.

[0179] Furthermore, by obtaining the comprehensive coverage value, it is possible to classify the levels according to the structure of the comprehensive coverage value, and corresponding disposal suggestions can be provided, such as:

[0180] Excellent (Grade A): Overall coverage value ≥90%, complete coverage of the core anatomical structure area, imaging information meets the clinical diagnosis / screening needs, no key structures are missing, no need for additional scanning, and can be directly used for subsequent diagnosis and treatment;

[0181] Good (Grade B): 70% ≤ overall coverage value < 90%, major anatomical structures are covered, only marginal or non-critical areas are not covered, basically meeting clinical needs. If it is a routine examination, no additional scan is required; if it is for precision diagnosis and treatment, it is recommended to make a judgment on whether to perform additional scan based on clinical findings.

[0182] Partial (Grade C): 50≤overall coverage value<70%, more than half of the anatomical structure is covered, but key functional areas may be missing, and the imaging information is incomplete. It is recommended to scan the missing areas to avoid diagnostic errors due to insufficient information.

[0183] Insufficient (Grade D): Overall coverage <50%, less than half of the anatomical structure is covered, core structures are missing, the imaging cannot meet clinical needs, and additional scans must be performed, otherwise it may lead to missed diagnosis / misdiagnosis.

[0184] S215. Based on the mapping relationship between the segmentation mask label and the region label, the anatomical region of the target anatomical structure in the target medical image is obtained.

[0185] Each anatomical region includes at least one target anatomical structure.

[0186] The segmentation mask labels correspond to individual anatomical structures, such as label 1 = liver, label 2 = spleen, and label 3 = left kidney, while clinical examinations are usually conducted in units of human body regions, such as abdominal examinations and chest examinations.

[0187] Based on medical physiological structures, mapping relationships are established, such as liver → abdominal region and heart → chest region. Individual anatomical structures are integrated into a hierarchical relationship of anatomical region-anatomical structure. For example, abdominal structures such as the liver, spleen, kidneys, and pancreas are uniformly classified into the abdominal region. This upgrades the assessment from whether a single anatomical structure is covered to whether all structures in the entire anatomical region are covered, which is in line with the actual clinical scenario of regional examination.

[0188] S216. Based on the user's preset inspection area and the anatomical structure to be inspected, determine whether the target anatomical structure and the corresponding area label of the user are completely covered in order to generate a complete result.

[0189] In clinical practice, different examinations have clearly defined pre-defined examination areas and anatomical structures to be examined. For example, the pre-defined examination area for an abdominal CT scan is the abdomen, and the pre-defined anatomical structures to be examined are the liver, spleen, kidneys, and pancreas. Simply looking at the overall coverage value of a single anatomical structure, such as 100% for the liver and 80% for the spleen, does not determine whether it meets the overall requirements for an abdominal examination. By comparing the target anatomical structure and its corresponding area label with the pre-defined examination area and the structure to be examined, a more accurate assessment can be made.

[0190] If all anatomical structures are completely covered, a complete coverage result will be generated, and no additional scan is required.

[0191] If an anatomical structure is not covered during an examination, such as a pancreatic comprehensive coverage value of 0%, a "incomplete coverage" result will be generated, clearly identifying the missing item and avoiding diagnostic errors due to "missing key structures".

[0192] S217, Output a comprehensive report.

[0193] The comprehensive report includes at least one of the following: user information, integrity results, region labels, target anatomy, and comprehensive coverage values.

[0194] Example:

[0195] User: P001-V01

[0196] Scanning sequences: 3 (plain scan, arterial phase, portal venous phase)

[0197] √ Area coverage: Complete (neck, chest, abdomen, pelvis)

[0198] Anatomical structure coverage details:

[0199] √Liver: 98% (Excellent)

[0200] √ Spleen: 92% (Excellent)

[0201] Left kidney: 68% (Warning: Coverage is slightly low, repeat examination recommended)

[0202] × Right kidney: 45% (Severe: Insufficient coverage, additional scan required)

[0203] Overall rating: Grade C (Insufficient coverage of some organs)

[0204] Recommendation: It is recommended to perform a supplementary scan of the right kidney area.

[0205] This application provides an intelligent assessment method for anatomical coverage of medical images. It acquires raw medical images such as CT and MRI, preprocesses them to obtain three-dimensional target medical images, then classifies the target medical images voxels according to preset segmentation mask labels to generate mask files. Each segmentation mask label corresponds to a unique anatomical structure. Subsequently, it filters the voxels corresponding to the segmentation mask labels, obtains three-dimensional coordinates to determine the bounding box range, and counts the number of voxels. It calculates the volume of the target anatomical structure based on the voxel volume, then compares this volume with a standard reference volume to obtain the target coverage rate. Finally, it acquires the target coverage values ​​of multiple scan sequences from the same user; a single value is directly used as the comprehensive coverage value, and the maximum value among multiple values ​​is taken. The anatomical region is determined according to the mapping relationship between segmentation mask labels and region labels. The coverage integrity is judged by combining preset examination areas and anatomical structures. Finally, it outputs a comprehensive report containing user information and integrity results, achieving automated and standardized coverage assessment, improving efficiency and accuracy.

[0206] Figure 3 A schematic diagram of the structure of a medical imaging anatomical coverage intelligent assessment system provided in one embodiment of this application is shown below. Figure 3 As shown, the intelligent assessment system 30 for medical image anatomical coverage provided in this embodiment includes:

[0207] Image preprocessing module 301 is used to acquire raw medical images and preprocess the raw medical images to obtain a three-dimensional target medical image.

[0208] The mask classification module 302 is used to classify the target medical image on a voxel-by-voxel basis according to the preset segmentation mask labels to obtain a mask file. The mask file and the target medical image have the same size, and each segmentation mask label corresponds to an anatomical structure.

[0209] The volume calculation module 303 is used to filter the voxels of the target anatomical structure corresponding to the segmentation mask label based on the segmentation mask label, and obtain the volume of the target anatomical structure.

[0210] The coverage calculation module 304 is used to obtain the target coverage based on the target anatomical structure volume and the standard reference volume of the target anatomical structure.

[0211] This embodiment provides a medical imaging anatomical coverage intelligent assessment system that can execute the methods provided in the above-described method embodiments. Its implementation principle and technical effects are similar, and will not be described in detail here.

[0212] Figure 4 A schematic diagram of the structure of a medical imaging anatomical coverage intelligent assessment device provided in one embodiment of this application is shown below. Figure 4 As shown, the intelligent medical imaging anatomical coverage assessment device 40 provided in this embodiment includes:

[0213] The device 40 includes at least one processor 401 and a memory 402. Optionally, the device 40 also includes a communication component 403. The processor 401, memory 402, and communication component 403 are connected via a bus 404.

[0214] In a specific implementation, at least one processor 401 executes computer execution instructions stored in memory 402, causing at least one processor 401 to perform the above-described method.

[0215] The specific implementation process of processor 401 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0216] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0217] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0218] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0219] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0220] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0221] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0222] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0223] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0224] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0225] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0226] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0227] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0228] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for intelligent evaluation of anatomical coverage of medical images, characterized in that, The method includes: The original medical images are acquired, and the original medical images are preprocessed to obtain a three-dimensional target medical image. According to the preset segmentation mask labels, the target medical image is classified on a voxel-by-voxel basis to obtain a mask file, wherein the mask file and the target medical image have the same size, and each segmentation mask label corresponds to an anatomical structure; Based on the segmentation mask label, the voxels of the target anatomical structure corresponding to the segmentation mask label are selected to obtain the volume of the target anatomical structure; The target coverage is obtained based on the target anatomical structure volume and the standard reference volume of the target anatomical structure.

2. The method of claim 1, wherein, The process of obtaining target coverage based on the target anatomical structure volume and the standard reference volume of the target anatomical structure includes: Determine whether the volume of the target anatomical structure is smaller than the standard reference volume corresponding to the target anatomical structure. If so, calculate the ratio of the target anatomical structure volume to the standard reference volume of the target anatomical structure to obtain the target coverage. If not, then the target coverage rate is set to 1.

3. The method of claim 2, wherein, After obtaining the target coverage based on the target anatomical structure volume and the standard reference volume of the target anatomical structure, the process includes: Acquire multiple target medical images from multiple scan sequences of the same user, and extract the target anatomical structures and corresponding target coverage values; Determine the number of target coverage values ​​for the same anatomical structure. When there is only one target coverage value, the target coverage value is determined to be the comprehensive coverage value of the target's anatomical structure. When the number of target coverage values ​​is greater than or equal to one, the maximum value among the multiple target coverage values ​​is determined as the comprehensive coverage value of the target anatomical structure.

4. The method according to claim 3, characterized in that, The determination of the comprehensive coverage value includes: Based on the mapping relationship between segmentation mask labels and region labels, the anatomical regions of the target anatomical structures in the target medical image are obtained, wherein each anatomical region includes at least one target anatomical structure. Based on the user's preset inspection area and the anatomical structure to be inspected, determine whether the target anatomical structure and corresponding area labels of the user are completely covered, so as to generate a complete result; Output a comprehensive report, wherein the comprehensive report includes at least one of the following: user information, integrity results, region labels, target anatomical structures, and comprehensive coverage values.

5. The method according to claim 1, characterized in that, The process of acquiring the original medical image and preprocessing the original medical image to obtain a three-dimensional target medical image includes: Acquire raw medical images, wherein the raw medical images include metadata of multi-layer slices, and the metadata includes, but is not limited to, pixel data; The original medical image is converted into a first medical image in a preset format, wherein the first medical image is three-dimensional and has image spatial coordinate information; The first medical image is resampled according to a preset resolution to obtain a second medical image; Based on the preset window width and window level, the pixel values ​​of the second medical image are normalized to a preset range to obtain the target medical image.

6. The method according to claim 1, characterized in that, Based on the segmentation mask label, the voxels corresponding to the target anatomical structure are selected to obtain the volume of the target anatomical structure, including: Obtain the voxels corresponding to the same segmentation mask label; Based on the number and volume of voxels, the volume of the target anatomical structure corresponding to the segmentation mask label in the target medical image is calculated, wherein the volume of voxels is obtained according to the resolution of the target medical image.

7. The method according to claim 6, characterized in that, After obtaining the voxels corresponding to the same segmentation mask label, the process includes: Obtain the 3D coordinates of the voxels corresponding to the same segmentation mask label to obtain the bounding box range; The number of voxels is obtained based on the bounding box range corresponding to the segmentation mask label.

8. A medical imaging anatomical coverage intelligent assessment system, characterized in that, The system includes: An image preprocessing module is used to acquire raw medical images and preprocess the raw medical images to obtain a three-dimensional target medical image. The mask classification module is used to classify the target medical image on a voxel-by-voxel basis according to a preset segmentation mask label to obtain a mask file, wherein the mask file and the target medical image have the same size, and each segmentation mask label corresponds to an anatomical structure; The volume calculation module is used to filter the voxels of the target anatomical structure corresponding to the segmentation mask label based on the segmentation mask label, and obtain the volume of the target anatomical structure. The coverage calculation module is used to obtain the target coverage based on the target anatomical structure volume and the standard reference volume of the target anatomical structure.

9. A medical imaging anatomical coverage intelligent assessment device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images