Tumor and blood vessel three-dimensional space relation quantitative analysis method based on enhanced CT image

Abstract

The invention relates to a tumor and blood vessel three-dimensional space relation quantitative analysis method based on an enhanced CT image, and the method comprises the following steps: obtaining a thin-layer enhanced CT image, and carrying out the segmentation processing of the thin-layer enhanced CT image, and obtaining a tumor and blood vessel segmentation result; extracting a blood vessel center line from the segmentation result to generate a local plane, and reconstructing to obtain two-dimensional slices corresponding to each point on the blood vessel center line, including a blood vessel slice and a tumor slice; and based on the blood vessel section and the tumor section, determining an interaction area and calculating a wrapping angle. Compared with the prior art, the method has the advantages that the three-dimensional space relation between the tumor and the main blood vessel can be efficiently and accurately quantified, especially the wrapping angle can be calculated, and therefore more accurate and quantitative data can be provided for follow-up operation resection evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method for quantitatively analyzing the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images. Background Art

[0002] Quantifying the three-dimensional spatial relationship between tumors and surrounding blood vessels is of great clinical significance in the clinical diagnosis and treatment planning of cancer. In particular, the extent to which tumors surround and invade multiple important blood vessels directly influences surgical resectability assessment and treatment selection. However, existing technologies in this area have significant shortcomings and limitations.

[0003] Currently, most clinical imaging diagnoses still rely on two-dimensional images (such as CT or MRI slices) to evaluate the relationship between tumors and blood vessels. Doctors usually examine the two-dimensional slices layer by layer and, combined with clinical experience, subjectively judge the invasion and envelopment of major blood vessels (such as the celiac trunk, superior mesenteric artery, portal vein, etc.) by the tumor [1]. This method is not only time-consuming and labor-intensive, but also has large subjective errors. Relying on two-dimensional images for visual judgment or simple morphological analysis, it is difficult to accurately quantify the degree of tumor envelopment of blood vessels. For example, when the boundary between the tumor and the blood vessels is blurred, it is easy to lead to misjudgment. It can be said that the limitation of two-dimensional imaging is that it cannot fully reflect the three-dimensional spatial relationship between the tumor and the blood vessels, especially when dealing with multiple blood vessels intertwined or complex branches, the accuracy is not high.

[0004] In recent years, some research and clinical practice have introduced simple three-dimensional modeling techniques, such as using surface reconstruction or bounding boxes to approximate the relative positions of tumors and blood vessels [2]. However, these models mostly evaluate the spatial relationship based on the volumetric overlap of tumors and blood vessels, which makes it difficult to handle situations such as irregular tumor morphology and complex vascular networks. In addition, most three-dimensional geometric models are based on simple shapes (such as ellipsoids and cylinders), which cannot accurately reflect the complex spatial relationship between tumors and blood vessels in actual anatomical structures, especially when the tumor surrounds multiple blood vessels.

[0005] Furthermore, traditional image analysis methods often require extensive manual intervention and repetitive operations when processing the complex spatial relationships between tumors and blood vessels, resulting in low efficiency and prone to errors. This is especially true when a tumor surrounds multiple major blood vessels, as manually quantifying the angles of contact between the tumor and each vessel is extremely tedious.

[0006] Currently, most studies related to the quantification of the spatial relationship between tumors and blood vessels focus on the analysis of the tumor microenvironment. For example, some studies use three-dimensional reconstruction technology to evaluate the changes and branching structure of blood vessels in the tumor microenvironment, but these studies mainly focus on the overall morphological changes of the vascular network rather than accurately quantifying the spatial relationship between the tumor and single or multiple blood vessels [3]. Other studies involve the use of digital pathology images to establish the spatial relationship of the tumor microenvironment, but these studies are limited to pathological analysis and are difficult to apply to the quantification of the relationship between tumors and blood vessels in imaging data [4].

[0007] In summary, existing technologies have significant deficiencies in quantifying the three-dimensional spatial relationship between tumors and blood vessels. This is primarily due to the strong subjectivity of two-dimensional imaging methods, while simple three-dimensional modeling methods struggle to handle the complex morphologies of tumors and blood vessels. This is especially true when multiple blood vessels are simultaneously surrounded by tumors. Existing methods have limited quantification accuracy and low operational efficiency.

[0008] Prior art literature:

[0009] [1]Viviers,C.,Ramaekers,M.,Valiuddin,A., T.,Tasios,N.,vander Ven,J.,Jacobs,I.,Ewals,L.,Nederend,J.,de With,P.,Luyer,M.,&van derSommen,F.(2023).Segmentation-based assessment of tumor-vessel involvement forsurgical resectability prediction of pancreatic ductal adenocarcinoma.IEEE / CVF International Conference on Computer Vision Workshops(ICCVW),Paris,France,2023,pp.2413-2423.

[0010] [2] Groves, LA, VanBerlo, B., Veinberg, N., Alboog, A., Peters, TM, & Chen, ECS (2020). Automatic segmentation of the carotid artery and internal jugular vein from 2D ultrasound images for 3D vascular reconstruction. International Journal of Computer Assisted Radiology and Surgery, 15(12), 1835-1846.

[0011] [3]Silvestri, VL, Henriet, E., Linville, RM, Wong, AD, Searson, PC, & Ewald, AJ (2020). Atissue-engineered 3D microvessel model reveals the dynamics of mosaic vessel formation in breast cancer. Cancer Research, 80(19), 4288-4301.

[0012] [4] A., I., M., & et al. (2022). Deep learning-based tumor microenvironment segmentation is predictive of tumor mutations and patient survival in non-small-cell lung cancer. BMC Cancer, 22, 1001. Summary of the Invention

[0013] The purpose of the present invention is to overcome the defects of the above-mentioned existing technologies and provide a method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images, which can efficiently and accurately quantify the three-dimensional spatial relationship between tumors and major blood vessels.

[0014] The object of the present invention can be achieved by the following technical solution: A method for quantitatively analyzing the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images, comprising the following steps:

[0015] S1. Obtain thin-slice enhanced CT images and perform segmentation processing to obtain tumor and blood vessel segmentation results;

[0016] S2. Extracting the vascular centerline from the segmentation result to generate a local plane, and reconstructing the two-dimensional slices corresponding to each point on the vascular centerline, including vascular slices and tumor slices;

[0017] S3. Based on the vascular slices and tumor slices, the interaction area is determined and the wrapping angle is calculated.

[0018] Furthermore, the step S1 includes the following steps:

[0019] S11. Acquire thin-slice enhanced CT images and perform preprocessing, including noise reduction and normalization.

[0020] S12. For the pre-processed thin-layer enhanced CT images, a deep learning model is used to segment the tumor and blood vessels, and a morphological processing algorithm and a geometric consistency check are combined to obtain the tumor and blood vessel segmentation results.

[0021] S13. Save the segmentation result as a mask file, and superimpose the mask on the original CT image to generate a segmentation visualization image.

[0022] Furthermore, the step S11 specifically adopts non-local mean filtering and wavelet transform technology to perform noise reduction processing on the CT image, and converts the noise-reduced CT image into a standardized NIfTI format file, uniformly stores it and generates a processing log.

[0023] Furthermore, the step S12 specifically utilizes the nnUNet model to perform tumor and blood vessel segmentation processing;

[0024] The step S12 specifically adopts a morphological processing algorithm to clean up noise points and mis-segmented areas in the segmentation result through connected domain analysis and closing operations;

[0025] The step S12 specifically analyzes the continuity of the topological structure of the blood vessel through a geometric consistency check, and automatically repairs the broken or discontinuous areas of the blood vessel.

[0026] Furthermore, step S2 includes the following steps:

[0027] S21. Based on the mask file obtained by segmentation, a skeleton extraction algorithm is used to generate the blood vessel centerline;

[0028] S22, determining whether the blood vessel centerline is located within the blood vessel mask through a geometric consistency check, and performing repair processing;

[0029] S23, storing the blood vessel centerline result as three-dimensional coordinates, and generating a local plane corresponding to each coordinate point by calculating the tangent vector, normal vector, and binormal vector corresponding to each coordinate point on the blood vessel centerline;

[0030] S24. Resampling the three-dimensional image on the generated local plane coordinates to extract two-dimensional slices, interpolating the original image data using the map_coordinates function, and generating corresponding blood vessel slices and tumor slices on the local plane.

[0031] Furthermore, the specific process of step S23 is as follows:

[0032] Arrange the coordinates of the blood vessel centerline in order, and calculate the tangent vector of each coordinate point by solving the difference between adjacent coordinate points one by one;

[0033] For the last point, simply repeat the tangent vector of the second-to-last point to ensure that the tangent vector corresponds one-to-one with the centerline coordinates;

[0034] According to the tangent vector, a fixed vector that is not parallel to it is selected as a reference, and the normal vector is generated by the cross product, and the binormal vector is generated by the cross product again;

[0035] Normalize all vectors to form an orthogonal local coordinate system, and check whether the tangent vector, normal vector, and binormal vector form an orthogonal relationship to ensure the validity of the coordinate system;

[0036] Based on the coordinate points of the blood vessel centerline and their corresponding orthogonal vectors, a two-dimensional plane perpendicular to the tangent line, ie, a local plane, is generated.

[0037] Furthermore, step S3 includes the following steps:

[0038] S31, marking the blood vessel region and the tumor region in the two-dimensional slice as binary masks, respectively, and extracting the interaction region between the tumor and the blood vessel through a logical AND operation;

[0039] S32, extracting the intersection of the tumor and the blood vessel boundary in the interaction region;

[0040] S33. Determine the surrounding conditions of the tumor and the blood vessels based on the interaction area and the intersection of the tumor and blood vessel boundaries, and calculate the surrounding angle.

[0041] Furthermore, in step S33, if the intersection of the blood vessel region and the tumor region is empty, it is determined that there is no contact between the tumor and the blood vessel;

[0042] If the vascular region is a subset of the tumor region, then the tumor and the blood vessels are completely surrounded;

[0043] If the intersection of the blood vessel region and the tumor region is not empty, and the blood vessel region is not a subset of the tumor region, it is determined that the tumor and the blood vessel are partially surrounded.

[0044] Furthermore, in step S33 , for the partial wrapping situation, the wrapping angle is calculated in combination with the intersection point of the blood vessel region and the tumor region, and the centroid of the blood vessel region.

[0045] Furthermore, the calculation process of the wrapping angle is as follows: the intersection points of the blood vessel region and the tumor region include point b and point c, the centroid of the blood vessel region is point a,

[0046] If the centroid a of the blood vessel region is outside the interaction region, the wrapping angle is the angle between line segment ab and line segment ac;

[0047] If the centroid a of the blood vessel region is within the interaction region, the wrapping angle is [360 degrees - (the angle between line segment ab and line segment ac)].

[0048] Compared with the prior art, the present invention has the following advantages:

[0049] The present invention first acquires thin-slice enhanced CT images and performs segmentation processing to obtain tumor and blood vessel segmentation results. The vascular centerline is then extracted from the segmentation results to generate local planes, and two-dimensional slices corresponding to each point on the vascular centerline are reconstructed, including vascular and tumor slices. Based on the vascular and tumor slices, the interaction region is determined and the wrapping angle is calculated. This allows for automatic and rapid processing of complex 3D image data without manual adjustments. Furthermore, it accurately segments blood vessels and tumors from enhanced CT images, calculates the extent of tumor wrapping around blood vessels, and calculates the 3D wrapping angle, thereby precisely quantifying the 3D spatial relationship between the tumor and major blood vessels.

[0050] The present invention performs preprocessing after acquiring thin-layer enhanced CT images, and adopts non-local mean filtering and wavelet transform technology to perform noise reduction on the CT images. By accurately removing artifacts and random noise in the images, it ensures the generation of clear image data, provides higher accuracy for subsequent segmentation and analysis, and optimizes noise reduction for complex backgrounds. Even in tumor-adjacent areas or low-contrast environments, the integrity of image details can be maintained, which is conducive to subsequent segmentation processing.

[0051] This invention utilizes the nnUNet model, which supports complex segmentation tasks and adapts to segmentation requirements in various scenarios. During the segmentation process, it maintains extremely high segmentation accuracy, particularly for regions with irregular morphology or strong local invasiveness. Furthermore, the invention employs an automated morphological processing algorithm to automatically clean up noise points and mis-segmented regions in the segmentation results through connected domain analysis and closing operations. A geometric consistency check analyzes the continuity of vascular topology and automatically repairs broken or discontinuous regions, ensuring the reliability of the segmentation results.

[0052] The present invention extracts the vascular centerline from the segmentation results, then generates local planes and reconstructs two-dimensional slices. Based on the two-dimensional slices, the interaction area between the tumor and the blood vessels is analyzed to determine the surrounding situation and calculate the three-dimensional surrounding angle for the partial surrounding situation. The surrounding angle can be effectively quantified, which is conducive to the subsequent evaluation of the degree of tumor invasion on the blood vessels. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 Schematic diagram of the method flow of the present invention;

[0054] Figure 2 Schematic diagram of the application process of the embodiment;

[0055] Figure 3 Schematic diagram of the blood vessel centerline superimposed on the blood vessel structure in the embodiment

[0056] Figure 4 Schematic diagram of the surrounding situation of tumor and blood vessels;

[0057] Figure 5 Schematic diagram of the case where the tumor is partially surrounded by blood vessels and the centroid of the blood vessels is located outside the interaction area;

[0058] Figure 6 Schematic diagram of the case where the tumor is partially surrounded by blood vessels and the blood vessel centroid is located within the interaction area. DETAILED DESCRIPTION

[0059] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] Example

[0061] like Figure 1 As shown, a method for quantitatively analyzing the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images includes the following steps:

[0062] S1. Obtain thin-slice enhanced CT images and perform segmentation processing to obtain tumor and blood vessel segmentation results;

[0063] S2. Extracting the vascular centerline from the segmentation result to generate a local plane, and reconstructing the two-dimensional slices corresponding to each point on the vascular centerline, including vascular slices and tumor slices;

[0064] S3. Based on the vascular slices and tumor slices, the interaction area is determined and the wrapping angle is calculated.

[0065] This embodiment applies the above solution, such as Figure 2 As shown, the main contents are:

[0066] Step 1. Image data acquisition and segmentation

[0067] Realize automatic collection and preprocessing of image data to ensure that data quality meets the requirements of subsequent analysis.

[0068] 1. Automatic image loading

[0069] This embodiment builds a quantitative analysis system for the three-dimensional spatial relationship between tumors and blood vessels. The system is seamlessly connected to the interface of the PACS (picture archiving and communication system) of the hospital. It can automatically extract the enhanced CT image data of the patient based on the patient information retrieval, and simultaneously capture the complete scan sequence of the arterial and venous phases.

[0070] Automatically identify image type, slice thickness, and scan sequence using DICOM image metadata to avoid data mismatch.

[0071] For imaging data acquired from external sources (such as CT scan files imported via USB drives, network shares, or cloud storage), the system performs an integrity check on each file to ensure it has not been corrupted during transmission. This check includes file format, metadata integrity, and the continuity of the scan sequence. Missing or corrupted files are automatically flagged and the user is notified, allowing for timely data addition or repair.

[0072] 2. Automated data preprocessing

[0073] This embodiment uses non-local mean filtering and wavelet transform techniques to reduce noise in loaded CT images. By precisely removing artifacts and random noise from the images, it ensures clear image data, enhancing the accuracy of subsequent segmentation and analysis. The noise reduction algorithm is optimized for complex backgrounds, preserving the integrity of image details even in areas adjacent to tumors or in low-contrast environments.

[0074] After preprocessing, this embodiment also saves the results in a standardized NIfTI format for unified storage and subsequent access. Simultaneously, the system generates a detailed processing log, recording the parameter settings and execution results of each step. This log file provides a reliable basis for clinical audits and result replication, facilitating user monitoring and optimization of the entire process.

[0075] Step 2. Tumor and blood vessel segmentation

[0076] Automatic segmentation of tumors and blood vessels is achieved through advanced deep learning models, outputting high-precision segmentation results.

[0077] 1. Automatic loading of segmentation model

[0078] The system's built-in nnUNet model supports complex segmentation tasks. After loading an image, the system automatically analyzes the image's metadata and matches the most appropriate model parameters and pre-trained weights based on the image's characteristics, ensuring the model can adapt to segmentation needs in various scenarios.

[0079] During the segmentation process, the system supports multi-task segmentation of tumors and multiple blood vessels (such as the celiac trunk, superior mesenteric artery, common hepatic artery, portal vein, and superior mesenteric vein) at the same time, and can maintain extremely high segmentation accuracy, especially for areas with irregular morphology or strong local invasiveness.

[0080] 2. Automation of post-segmentation processing

[0081] The system uses an automated morphological processing algorithm to automatically clean up noise points and mis-segmented areas in the segmentation results through connected domain analysis and closing operations.

[0082] Through geometric consistency checking, the continuity of the vascular topology is analyzed, and the broken or discontinuous areas of the blood vessels are automatically repaired.

[0083] 3. Automatic saving and monitoring of results

[0084] The system saves the segmentation results as a mask file in real time and superimposes the mask on the original image to generate a segmentation visualization image, thereby intuitively displaying the shape and range of the segmented area, making it convenient for doctor users to quickly review and verify the accuracy of the segmentation results.

[0085] In this embodiment, an alert is automatically triggered when an abnormal segmentation result is detected (for example, the mask volume is lower than the clinical reference value or the morphology does not conform to anatomical conventions). The alert details the cause of the problem and provides a visual screenshot of the segmented area, allowing doctors to further confirm and adjust the results.

[0086] Step 3. Blood vessel centerline extraction

[0087] The three-dimensional centerline is extracted from the blood vessel segmentation results, providing a basis for the subsequent accurate calculation of the wrapping angle and spatial relationship.

[0088] 1. Centerline extraction

[0089] This embodiment generates a centerline based on the segmented blood vessel mask using a skeleton extraction algorithm.

[0090] The coordinates and tangent vectors of the centerline points are automatically calculated by the system, and the normal vector and binormal vector of each point are generated at the same time, providing a mathematical basis for subsequent two-dimensional slice generation.

[0091] 2. Automatic adjustment of complex structures

[0092] For complex scenes with vascular bifurcations or tumor encapsulation, parameters (such as skeleton resolution and filtering threshold) are automatically optimized to ensure the continuity and accuracy of the centerline.

[0093] This embodiment supports automatic adaptation to different blood vessel types, avoiding errors caused by morphological differences.

[0094] 3. Automatic calibration of extraction results

[0095] Through geometric consistency checking, the centerline of the blood vessel is analyzed to see if it is within the blood vessel mask, and the extraction results are automatically verified. When an anomaly is detected, such as when the centerline deviates from the blood vessel area, the correction program is automatically triggered for repair.

[0096] The resulting centerline results are stored as 3D coordinates and visualized as line drawings superimposed on the vascular structure (e.g. Figure 3 shown).

[0097] Step 4. Calculate the 3D wrap angle

[0098] The spatial relationship between the tumor and major blood vessels was calculated, and the wrapping angle and contact depth were quantified.

[0099] 1. Tangent vector calculation

[0100] The centerline coordinates are arranged in order, and the tangent vector of each point is calculated by solving the difference between adjacent coordinate points one by one.

[0101] For the last point, simply repeat the tangent vector of the second-to-last point, ensuring that the tangent vector corresponds one-to-one with the centerline coordinate.

[0102] 2. Normal vector and binormal vector generation

[0103] According to the tangent vector, a fixed vector that is not parallel to it is selected as a reference, and the normal vector is generated by the cross product;

[0104] Again through the cross product, generate the binormal vector;

[0105] All vectors are normalized to form an orthogonal local coordinate system (tangent, normal vector, binormal vector).

[0106] Check whether the vectors form an orthogonal relationship (tangent, normal, and binormal vectors are perpendicular to each other) to ensure the validity of the coordinate system.

[0107] 3. Automatic generation of local planes

[0108] Based on the centerline point and its corresponding orthogonal vectors (tangent, normal, binormal), a two-dimensional plane perpendicular to the tangent is generated.

[0109] Define the plane grid range, and calculate the actual coordinates of each point on the plane in three-dimensional space through the plane grid coordinates, combined with the normal vector and binormal vector.

[0110] 4. Slice Reconstruction and Data Extraction

[0111] The 3D image (tumor and vascular data) is resampled on the generated local plane coordinates to extract 2D slices. The original image data is interpolated using the map_coordinates function to generate corresponding vascular and tumor slices on the 2D plane.

[0112] The vascular area and tumor area in the slice are marked as binary masks respectively, and the interaction area between the tumor and the blood vessel is extracted through logical AND operation.

[0113] 5. Boundary point detection

[0114] In the interaction region, the intersection points of the tumor and blood vessel boundaries are extracted;

[0115] Checking whether these intersection points form a continuous interaction area ensures the basic reliability of the wrap angle calculation.

[0116] 6. Angle calculation

[0117] like Figure 4 As shown, when the intersection of vessel and tumor is empty, it is defined as no contact ( Figure 4 (indicated by C in the middle);

[0118] When the vessel is a subset of the tumor, it is defined as completely enclosed ( Figure 4 A in the middle);

[0119] When the intersection of vessel and tumor is not empty, and vessel is not a subset of tumor, it is defined as partial envelopment ( Figure 4 (indicated by B in the middle).

[0120] In the case of partial encapsulation, there are two intersection points between the blood vessel and the tumor: b and c, and there are two scenarios:

[0121] 1) The centroid a of the blood vessel is outside the overlap interaction area (e.g. Figure 5 As shown), the wrapping angle is the angle between line segments ab and ac;

[0122] 2) The centroid a of the blood vessel is within the overlap interaction region (e.g. Figure 6 As shown), the wrapping angle is [360deg-(the angle between ab and ac)].

[0123] 7. Multi-slice integration and result statistics

[0124] The system repeats the above calculations for each slice generated at each point on the vessel centerline, recording the wrap angle and interaction area for each slice.

[0125] The maximum wrapping angle is then extracted to assess the most severe degree of tumor invasion of blood vessels. While recording the angle, the classification of the wrapping situation (complete wrapping, partial wrapping, slight contact) is statistically analyzed to generate a quantitative report.

[0126] Step 5. Report generation and data storage

[0127] This embodiment is designed to generate standardized analysis reports and visualization results for quick reference by physician users.

[0128] 1. Quantitative analysis report

[0129] Generates a report containing the following information: 1) the wrap angle and contact depth of each vessel; 2) quantitative indicators of the comprehensive spatial relationship between the tumor and multiple vessels

[0130] Supports PDF and Excel format output, making it convenient for doctors to share and archive.

[0131] 2. Visualization result generation

[0132] Automatically generate a 3D visualization model, display the 3D interaction between tumors and blood vessels, and present a 2D slice overlay.

[0133] The visualization results can dynamically adjust the perspective or level, supporting doctor users to view and modify them in real time.

[0134] 3. Automatic storage and backup

[0135] Reports and visualization results are automatically stored on the server, and support data encryption and permission management to ensure information security.

[0136] In summary, this solution provides a technology that can accurately segment blood vessels and tumors from enhanced CT images through the nnUNet model and centerline extraction algorithm, and further calculates the three-dimensional angle of the tumor surrounding the blood vessels. This method significantly improves the quantitative accuracy of the spatial relationship between blood vessels and tumors in CT imaging diagnosis, and can provide a reliable imaging basis for the formulation of subsequent treatment plans. In addition, this solution can automatically calculate the surrounding angle and depth of each major blood vessel and tumor, and can provide consistent quantitative results for different doctors, effectively reducing the impact of subjective factors on the evaluation results, and improving the accuracy and consistency of tumor surgical resectability assessment, thereby providing a scientific basis for the formulation of personalized treatment plans. By introducing automated vascular centerline extraction and angle calculation algorithms, this solution can quickly and efficiently complete the quantitative analysis of the spatial relationship between tumors and multiple blood vessels. Compared with traditional methods, the processing time is significantly shortened and the quantitative accuracy is significantly improved, providing a convenient solution for batch data analysis in clinical practice.

[0137] Compared with the existing technology, the advantages of this solution are mainly reflected in the following aspects:

[0138] 1) Innovative spatial relationship quantification methods

[0139] Compared to existing methods that rely primarily on subjective judgment or simple geometric modeling, this approach uses automated quantitative algorithms to accurately calculate the spatial relationship between tumors and multiple blood vessels. This approach not only improves assessment accuracy but also eliminates human error, making surgical and treatment planning more scientific and data-driven.

[0140] 2) Automation and efficiency

[0141] Traditional methods often require extensive manual work and multiple steps of manual adjustment. This solution introduces a fully automated analysis process that can rapidly process complex 3D imaging data, significantly reducing the workload for physicians. By automatically detecting and quantifying the spatial interaction between tumors and blood vessels, it can accelerate assessments and improve clinical efficiency.

[0142] 3) High accuracy and reliability in clinical applications

[0143] Because this solution provides more precise data when calculating the relationship between tumor and blood vessels, especially in complex situations with multiple intertwined blood vessels, it ensures the reliability and consistency of surgical decision-making. This technology provides doctors with a clear and quantitative basis for decision-making in surgical planning and radiotherapy, significantly reducing the risk of postoperative complications and improving treatment outcomes.

[0144] 4) Flexibility to adapt to complex cases

[0145] This solution can handle diverse tumor morphologies and complex vascular structures, making it suitable for different types of cases. Its high flexibility enables it to adapt to various clinical scenarios, not only improving treatment success rates but also providing strong support for the development of personalized treatment plans.

[0146] Through these innovations, this solution not only fills the gaps in existing technologies in quantifying the spatial relationship between tumors and blood vessels, but also greatly enhances the application value of medical imaging in clinical decision-making.

Claims

1. A method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images, characterized in that: The following steps are involved: S1. Obtain thin-slice enhanced CT images and perform segmentation processing to obtain tumor and blood vessel segmentation results; S2. Extracting the vascular centerline from the segmentation result to generate a local plane, and reconstructing the two-dimensional slices corresponding to each point on the vascular centerline, including vascular slices and tumor slices; S3. Based on the vascular slices and tumor slices, the interaction area is determined and the wrapping angle is calculated.

2. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 1, characterized in that: The step S1 comprises the following steps: S11. Acquire thin-slice enhanced CT images and perform preprocessing, including noise reduction and normalization. S12. For the pre-processed thin-layer enhanced CT images, a deep learning model is used to segment the tumor and blood vessels, and a morphological processing algorithm and a geometric consistency check are combined to obtain the tumor and blood vessel segmentation results. S13. Save the segmentation result as a mask file, and superimpose the mask on the original CT image to generate a segmentation visualization image.

3. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 2, characterized in that: The step S11 specifically uses non-local mean filtering and wavelet transform technology to perform noise reduction processing on the CT image, and converts the noise-reduced CT image into a standardized NIfTI format file, uniformly stores it and generates a processing log.

4. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 2, characterized in that: The step S12 specifically uses the nnUNet model to perform tumor and blood vessel segmentation processing; The step S12 specifically adopts a morphological processing algorithm to clean up noise points and mis-segmented areas in the segmentation result through connected domain analysis and closing operations; The step S12 specifically analyzes the continuity of the topological structure of the blood vessel through a geometric consistency check, and automatically repairs the broken or discontinuous areas of the blood vessel.

5. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 2, characterized in that: The step S2 comprises the following steps: S21. Based on the mask file obtained by segmentation, a skeleton extraction algorithm is used to generate the blood vessel centerline; S22, determining whether the blood vessel centerline is located within the blood vessel mask through a geometric consistency check, and performing repair processing; S23, storing the blood vessel centerline result as three-dimensional coordinates, and generating a local plane corresponding to each coordinate point by calculating the tangent vector, normal vector, and binormal vector corresponding to each coordinate point on the blood vessel centerline; S24. Resampling the three-dimensional image on the generated local plane coordinates to extract two-dimensional slices, interpolating the original image data using the map_coordinates function, and generating corresponding blood vessel slices and tumor slices on the local plane.

6. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 5, characterized in that: The specific process of step S23 is as follows: Arrange the coordinates of the blood vessel centerline in order, and calculate the tangent vector of each coordinate point by solving the difference between adjacent coordinate points one by one; For the last point, simply repeat the tangent vector of the second-to-last point to ensure that the tangent vector corresponds one-to-one with the centerline coordinates; According to the tangent vector, a fixed vector that is not parallel to it is selected as a reference, and the normal vector is generated by the cross product, and the binormal vector is generated by the cross product again; Normalize all vectors to form an orthogonal local coordinate system, and check whether the tangent vector, normal vector, and binormal vector form an orthogonal relationship to ensure the validity of the coordinate system; Based on the coordinate points of the blood vessel centerline and their corresponding orthogonal vectors, a two-dimensional plane perpendicular to the tangent line, ie, a local plane, is generated.

7. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 5, characterized in that: The step S3 comprises the following steps: S31, marking the blood vessel region and the tumor region in the two-dimensional slice as binary masks, respectively, and extracting the interaction region between the tumor and the blood vessel through a logical AND operation; S32, extracting the intersection of the tumor and the blood vessel boundary in the interaction region; S33. Determine the surrounding conditions of the tumor and the blood vessels based on the interaction area and the intersection of the tumor and blood vessel boundaries, and calculate the surrounding angle.

8. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 7, characterized in that: In step S33, if the intersection of the blood vessel region and the tumor region is empty, it is determined that there is no contact between the tumor and the blood vessel; If the vascular region is a subset of the tumor region, then the tumor and the blood vessels are completely surrounded; If the intersection of the blood vessel region and the tumor region is not empty, and the blood vessel region is not a subset of the tumor region, it is determined that the tumor and the blood vessel are partially surrounded.

9. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 8, characterized in that: In step S33 , the wrapping angle is calculated for the partial wrapping situation by combining the intersection point between the blood vessel region and the tumor region and the centroid of the blood vessel region.

10. The method for quantitative analysis of the three-dimensional spatial relationship between tumors and blood vessels based on enhanced CT images according to claim 9, characterized in that: The calculation process of the wrapping angle is as follows: the intersection points of the blood vessel region and the tumor region include point b and point c, the centroid of the blood vessel region is point a, If the centroid a of the blood vessel region is outside the interaction region, the wrapping angle is the angle between line segment ab and line segment ac; If the centroid a of the blood vessel region is within the interaction region, the wrapping angle is [360 degrees - (the angle between line segment ab and line segment ac)].

Images