Image processing method, image processing apparatus, and computer program product

Abstract

The present application relates to an image processing method, an image processing apparatus, and a computer program product, capable of improving the segmentation accuracy of a tubular structure. The image processing method comprises the steps of: analyzing the size of a lumen of a tubular structure in medical image data; distinguishing a plurality of regions of the tubular structure according to the size of the lumen; and respectively resampling the medical image data in the plurality of regions of the tubular structure and obtaining labels.

Description

Technical Field

[0001] This invention relates to an image processing method, an image processing apparatus, and a computer program product. Background Technology

[0002] In the field of medical image processing, accurate segmentation of tubular structures such as blood vessels and trachea is an important foundation for disease diagnosis (such as vascular stenosis and lung lesions) and treatment planning (such as interventional surgery navigation).

[0003] Existing image processing methods for segmenting tubular structures such as blood vessels and trachea are usually based on deep learning methods to automatically perform the segmentation task. However, when using deep learning methods to segment tubular structures such as blood vessels and trachea, segmentation errors are prone to occur. Summary of the Invention

[0004] The present invention was made in view of the above problems, and its purpose is to provide an image processing method, an image processing device, and a computer program product that can improve the segmentation accuracy of tubular structures.

[0005] The image processing method according to the present invention is characterized by comprising: a step of analyzing the size of the lumen of a tubular structure in medical image data; a step of distinguishing multiple regions of the tubular structure according to the size of the lumen; and a step of resampling the medical image data in the multiple regions of the tubular structure and obtaining labels.

[0006] Therefore, it is possible to improve the segmentation accuracy of tubular structures in medical image data segmentation tasks. In particular, for medical image data containing tubular structures with narrow lumens, it is possible to improve the connectivity and segmentation accuracy of the segmented tubular structures with narrow lumens.

[0007] According to the image processing method of the present invention, the step of analyzing the size of the lumen includes the steps of extracting the centerline of the tubular structure, performing distance transformation, and calculating the diameter or radius of the lumen.

[0008] According to the image processing method of the present invention, in the step of obtaining the label, for each region among the multiple regions of the tubular structure, different interpolation methods are used for resampling.

[0009] According to the image processing method of the present invention, in the step of distinguishing the multiple regions, multi-level labels are assigned according to the size of the lumen.

[0010] The image processing method according to the present invention is characterized by further comprising: a step of preprocessing the resampled medical image data; a step of training an inference model using the medical image data and the labels; and a step of performing a segmentation task on the medical image data using the inference model.

[0011] According to the image processing method of the present invention, in the step of obtaining the label, a functional relationship between the size and the threshold of the interpolation method is established, the threshold is applied to each pixel, the resampled medical image data is processed, and the label is obtained as a binary label.

[0012] Therefore, based on the functional relationship between the size of the tubular lumen and the threshold of the interpolation method used, the threshold of the interpolation method can be applied to each pixel in the medical image data. This improves the labeling accuracy in the medical image data used as training data, enabling accurate differentiation between tubular structure regions and background regions. Furthermore, it further enhances the segmentation accuracy of tubular structures in medical image data segmentation tasks.

[0013] According to the image processing method of the present invention, the medical image data is segmented using a learned inference model, wherein the inference model is learned by distinguishing multiple regions of the tubular structure in the medical image data according to the size of the lumen of the tubular structure, and resampling the medical image data in each of the multiple regions of the tubular structure and obtaining labels.

[0014] The image processing apparatus according to the present invention is characterized by comprising: an analysis unit for analyzing the size of the lumen of a tubular structure in medical image data; a differentiation unit for differentiating the tubular structure into multiple regions based on the size of the lumen; and an acquisition unit for resampling the medical image data in the multiple regions of the tubular structure and acquiring labels.

[0015] The computer program product according to the present invention includes a computer program, characterized in that, when executed, the computer program includes the following steps: analyzing the size of the lumen of a tubular structure in medical image data; distinguishing multiple regions of the tubular structure according to the size of the lumen; and resampling the medical image data in the multiple regions of the tubular structure respectively and obtaining labels.

[0016] According to the image processing method of the present invention, the inference result of the inference model is post-processed using the above-described image processing method.

[0017] Therefore, the image processing method of the present invention can be applied to the post-processing of the inference model. Consequently, the connectivity of the tubular structures in the segmentation results of the inference model can be improved.

[0018] According to the image processing method of the present invention, the diameter of the tubular structure is measured in the medical image data containing the label using the above-described image processing method.

[0019] Therefore, it is possible to assist in the measurement of tubular structures in tubular structures with narrow lumens.

[0020] Invention Effects

[0021] The image processing method, image processing apparatus, and computer program product according to the present invention can improve the segmentation accuracy of tubular structures. Attached Figure Description

[0022] Figure 1 This is a block diagram illustrating an example of the configuration of the image processing system of the present invention.

[0023] Figure 2 This is a block diagram illustrating a configuration example of the image processing apparatus according to the first embodiment.

[0024] Figure 3 This is a schematic diagram illustrating the size of the lumen of tubular structures in medical image data analyzed by the analysis unit.

[0025] Figure 4 This is a flowchart of the first implementation method.

[0026] Figure 5 This is a schematic diagram illustrating the differences between preprocessed medical image data based on the first embodiment and the comparative example.

[0027] Figure 6 This is a schematic diagram illustrating the differences in segmentation results between the first implementation method and the comparative example based on the test data.

[0028] Figure 7 This is a block diagram illustrating a configuration example of the image processing apparatus according to the second embodiment.

[0029] Figure 8 This is a schematic diagram illustrating the functional relationship between the diameter of the lumen of a tubular structure and a threshold.

[0030] Figure 9 This is a flowchart of the second implementation method.

[0031] Figure 10 This is a schematic diagram illustrating the difference between the segmentation results of the second implementation method and the comparative example based on the test data.

[0032] Figure 11 This is a block diagram illustrating a configuration example of the image processing apparatus according to the third embodiment.

[0033] Explanation of reference numerals in the attached figures

[0034] 1 Image processing device, 11 Processing circuit, 110 Receiving unit, 111 Analysis unit, 112 Differentiation unit, 113 Tag acquisition unit, 114 Preprocessing unit, 115 Training unit, 116 Inference unit, 12 Memory, 13 Communication IF, 2 Operation terminal, 3 Data storage device, 1A Image processing device, 11A Processing circuit, 113A Tag acquisition unit, 1B Image processing device, 11B Processing circuit, 110B Receiving unit, 111B Analysis unit. Detailed Implementation

[0035] Hereinafter, embodiments of the image processing method, image processing apparatus, and computer program product of the present invention will be described with reference to the accompanying drawings. In the following embodiments, the same or equivalent parts will be given the same reference numerals for description.

[0036] Furthermore, in the following description, the image processing method, image processing apparatus, and computer program product of the present invention are used as examples for segmentation tasks of tubular structures such as blood vessels and trachea. However, the present invention is not limited to this, and can of course be applied to segmentation tasks of other tissues, organs, and structures with tubular structures.

[0037] (First Implementation)

[0038] Reference Figure 1 An example of the configuration of the image processing system 100 will be described. Figure 1 This is a block diagram illustrating an example of the configuration of the image processing system 100 of the present invention.

[0039] The image processing system 100 includes an image processing device 1, an operation terminal 2, and a data storage device 3. The devices are connected to each other in a manner that enables communication via a shared signal transmission line, such as a bus or network.

[0040] Image processing apparatus 1 is an apparatus for performing segmentation tasks on medical images. Image processing apparatus 1 functions as a server of image processing system 100, for example, and can be a workstation capable of performing high-speed information processing.

[0041] The operation terminal 2 is a terminal for user operation. The operation terminal 2 functions as a client of the image processing system 100. The operation terminal 2 can be a computer, laptop, smartphone, tablet computer, wearable device, mobile terminal, or handheld terminal, etc.

[0042] Additionally, the operating terminal 2 may also have an input interface for receiving various operations from the user. The input interface converts these operations into electrical signals and transmits them to the processing circuitry of the operating terminal 2. The input interface of the operating terminal 2 can be a mouse, keyboard, buttons, panel switches, slider switches, trackball, operating panel, or touch panel.

[0043] Additionally, the operating terminal 2 may also have a display for showing various data or information. The display may be a liquid crystal display, a plasma display, an organic EL display, or an LED display. The display may also be a touch panel display that also functions as an input interface.

[0044] Data storage device 3 is a device for storing medical image data. Data storage device 3 can be a storage medium (e.g., magnetic storage medium, electromagnetic storage medium, optical storage medium, semiconductor memory) or a drive device for reading and writing information between itself and the storage medium.

[0045] Reference Figure 2 An example of the configuration of the image processing apparatus 1 will be described. Figure 2 This is a block diagram illustrating a configuration example of the image processing apparatus 1 according to the first embodiment.

[0046] The image processing apparatus 1 includes a processing circuit 11, a memory 12, and a communication IF 13. The components are communicatively connected to each other via a bus that serves as a shared signal transmission line.

[0047] The memory 12 is a device for storing various data or information. The memory 12 can be a storage medium that can be read by a processor (e.g., magnetic storage medium, electromagnetic storage medium, optical storage medium, semiconductor memory), or a drive device for reading and writing data or information between itself and the storage medium. The memory 12 stores computer programs that enable the processing circuit 11 to perform its various functions.

[0048] Communication IF13 is an interface for communicating various data or information with the various components included in the image processing system 100. Communication IF13 communicates various data or information with the operation terminal 2 or the data storage device 3.

[0049] The processing circuit 11 is a circuit that controls the overall operation of the image processing device 1. The processing circuit 11 includes at least one processor. A processor can be a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an Application Specific Integrated Circuit (ASIC), or a programmable logic device (such as a Simple Programmable Logic Device (SPLD), a Complex Programmable Logic Device (CPLD), or a Field Programmable Gate Array (FPGA)). When the processor is a CPU, the CPU implements each function by reading and executing the computer programs stored in the memory 12. When the processor is an ASIC, each function is directly embedded as a logic circuit within the ASIC's circuitry. The processor can be configured as a single circuit or by combining multiple independent circuits. The processing circuit 11 implements the functions of the functional units described later.

[0050] In the first embodiment, the processing circuit 11 includes a receiving unit 110, an analysis unit 111, a differentiation unit 112, a tag acquisition unit 113, a preprocessing unit 114, a training unit 115, and a reasoning unit 116.

[0051] The receiving unit 110 receives medical image data. Here, the medical image data includes, for example, a medical image and a label for a tubular structure in the medical image. The receiving unit 110 sends the received medical image data to the analysis unit 111.

[0052] Analysis unit 111 analyzes the size of the lumen of tubular structures in medical image data. (Refer to...) Figure 3 The size of the lumen of tubular structures in medical image data is explained based on the analysis of the analysis unit 111. Figure 3 This is a schematic diagram showing the size of the lumen of a tubular structure in medical image data analyzed by the analysis unit 111. Figure 3 The image above shows labels for tubular structures in medical image data. Figure 3 The diagram below shows the centerline of the tubular structure.

[0053] In the following explanation, we will use a tubular structure as an example to illustrate blood vessels. However, it is not limited to this; it can also be other tubular structures such as the trachea.

[0054] like Figure 3As shown, the analysis unit 111 extracts the centerline of the tubular structure from the medical image data. Specifically, the analysis unit 111, in the received medical image data, extracts the centerline of the tubular structure based on its label, such as... Figure 3 The centerline of the tubular structure is extracted as shown in the figure below.

[0055] Furthermore, the analysis unit 111 performs a distance transformation on the extracted centerline of the tubular structure to calculate the diameter or radius of the lumen of the tubular structure. Here, the analysis unit 111 calculates the diameter or radius of the lumen of the tubular structure as the size of the lumen. Moreover, the methods for extracting the centerline of the tubular structure, performing the distance transformation, and calculating the diameter or radius of the lumen can be known image processing methods.

[0056] The differentiation unit 112 distinguishes the tubular structure into multiple regions based on the size of the lumen of the extracted tubular structure. Specifically, the differentiation unit 112 assigns multi-level labels to the tubular structure based on the size of the lumen of the extracted tubular structure.

[0057] Here, as Figure 3 The following diagram illustrates a multi-level labeling system containing two levels of labels. In this case, the differentiation unit 112 assigns a "coarse lumen" label (the green label at the bottom of the diagram) to tubular structures with a larger diameter (diameter greater than a specified value), and a "fine lumen" label (the red label at the top of the diagram) to tubular structures with a smaller diameter (diameter less than a specified value). Thus, the differentiation unit 112 assigns multi-level labels and divides the tubular structure into multiple regions.

[0058] The above explanation uses the example of a multi-level tag containing two levels of tags. However, it is not limited to this; a multi-level tag can also have two or more other numbers of levels.

[0059] The tag acquisition unit 113 resamples the medical image data in multiple regions of the tubular structure and acquires tags accordingly. Specifically, the tag acquisition unit 113 resamples each region of the multiple regions of the tubular structure using a different interpolation method.

[0060] Here, "interpolation method" refers to a method of estimating the value of unknown data points using known data points, that is, interpolating missing values ​​between discrete data to achieve data continuity. Interpolation methods can directly affect the quality of labels in medical image data, especially in the tubular structure segmentation task involved in this invention, where appropriate interpolation methods can be used to eliminate structural errors in medical image data.

[0061] Additionally, "resampling" refers to adjusting the resolution, size, or grid structure of medical image data to suit subsequent processing. Resampling methods typically map the original data (such as pixels or voxels) from one space or resolution to another while preserving or reconstructing key information.

[0062] In the field of medical image data processing, interpolation methods such as Nearest Neighbor Interpolation, Linear Interpolation, and Cubic Spline Interpolation can be cited.

[0063] In addition, the labels obtained by the label acquisition unit 113 are the labels used for training data.

[0064] In this invention, the tag acquisition unit 113, for example, uses linear interpolation and a lower threshold binarization of the tag for the region where the segmentation unit 112 has assigned a tag to the thin lumen, to obtain an enhanced tag for the tubular structure of the thin lumen.

[0065] Specifically, "linear interpolation" refers to interpolating values ​​between two adjacent known data points using a linear function (a first-order polynomial). By using linear interpolation and lower thresholding for binarization, more consecutive pixels can be preserved, improving the connectivity of the tubular structure of the thin lumen and thus obtaining enhanced labels for the tubular structure. This provides the model with more complete structural information.

[0066] In addition, the label acquisition unit 113, for example, uses nearest neighbor interpolation to obtain an enhanced label for the tubular structure of the coarse lumen for the region where the label for the coarse lumen has been assigned by the differentiation unit 112.

[0067] Specifically, "nearest neighbor interpolation" refers to setting the value of an unknown data point to the value of the known data point whose location is closest to it. Because the tubular structure of a thick-walled cavity is large and its connectivity is not easily disrupted, using nearest neighbor interpolation can improve calculation speed and reduce computational cost, making it suitable for real-time processing.

[0068] The preprocessing unit 114 preprocesses the resampled medical image data.

[0069] Here, "preprocessing" refers to processes such as normalization of resampled medical image data. However, it is not limited to this; other preprocessing methods can also be applied to resampled medical image data.

[0070] The training unit 115 uses preprocessed medical image data and labels (thin lumen labels and thick lumen labels) obtained by the label acquisition unit 113 as a training dataset to train the inference model. This results in a trained inference model used for segmenting tubular structures in medical image data.

[0071] The inference unit 116 uses the trained inference model to perform a segmentation task on medical image data. This allows for the segmentation of tubular structures within the medical image data.

[0072] Therefore, according to the first embodiment of the present invention, the segmentation accuracy of tubular structures in medical image data segmentation tasks can be improved. In particular, for medical image data containing tubular structures with narrow lumens, the connectivity and segmentation accuracy of the segmented tubular structures with narrow lumens can be improved.

[0073] (Flowchart of the first implementation method)

[0074] The following is for reference Figure 4 The process of the first embodiment will be described. Figure 4 This is a flowchart of the first implementation method.

[0075] In step S10, the receiving unit 110 receives medical image data. Then, the process proceeds to step S11.

[0076] In step S11, the analysis unit 111 extracts the centerline of the tubular structure from the medical image data received by the receiving unit 110. Then, the process proceeds to step S12.

[0077] In step S12, the analysis unit 111 performs a distance transformation on the centerline of the extracted tubular structure to calculate the diameter or radius of the lumen of the tubular structure. The analysis unit 111 sends the medical image data and the calculated size of the lumen of the tubular structure to the differentiation unit 112. Then, the process proceeds to step S13.

[0078] In step S13, the differentiation unit 112 assigns multi-level labels to the tubular structure based on the size of the lumen of the tubular structure calculated by the analysis unit 111, thus differentiating the tubular structure into multiple regions. The differentiation unit 112 sends medical image data of the tubular structure, which is divided into multiple regions, to the label acquisition unit 113. Then, the process proceeds to step S14.

[0079] In step S14, the tag acquisition unit 113 resamples each region of the tubular structure, which is divided by the differentiation unit 112, using different interpolation methods to acquire tags. The tag acquisition unit 113 sends the medical image data and the acquired tags to the preprocessing unit 114. Then, the process proceeds to step S15.

[0080] In step S15, the preprocessing unit 114 preprocesses the resampled medical image data. The preprocessing unit 114 then sends the preprocessed medical image data to the training unit 115. Next, the process proceeds to step S16.

[0081] In step S16, the training unit 115 uses the medical image data preprocessed by the preprocessing unit 114 and the labels obtained by the label acquisition unit 113 as a training dataset to train the inference model. The training unit 115 outputs the trained inference model to the inference unit 116. Then, the process proceeds to step S17.

[0082] In step S17, the inference unit 116 uses the inference model trained by the training unit 115 to perform a segmentation task on the medical image data. This segments the tubular structures within the medical image data. The process then ends.

[0083] (Comparative Example)

[0084] Reference Figures 5-6 The differences between the first embodiment and the comparative example will be explained. Figure 5 This is a schematic diagram illustrating the differences between preprocessed medical image data based on the first embodiment and the comparative example. Figure 5 The image above represents the original image of the medical image data. Figure 5 The middle image represents preprocessed medical image data based on a comparative example (without using the image processing method of the present invention). Figure 5 The figure below shows preprocessed medical image data based on the first embodiment. Figure 6 This is a schematic diagram illustrating the differences in segmentation results between the first implementation method and the comparative example based on the test data. Figure 6 The image above represents the ground truth of tubular structures in medical image data. Figure 6 The middle figure shows the segmentation results of the tubular structure based on the comparative example (prior art). Figure 6 The figure below shows the segmentation result of the tubular structure based on the first embodiment.

[0085] like Figure 5 As shown, in the preprocessed medical image data based on the first embodiment, compared with the preprocessed medical image data based on the comparative example, as indicated by the arrow, the tubular structure region of the thin lumen is labeled more accurately.

[0086] like Figure 6 As shown, in the segmentation results of the tubular structure based on the first embodiment, compared with the segmentation results of the tubular structure based on the comparative example, as indicated by the arrow, a better segmentation result was obtained in the tubular structure region of the thin lumen, and the connectivity of the tubular structure region of the thin lumen was improved.

[0087] Furthermore, according to the first embodiment of the present invention, in a test set of medical image data containing 14 cases, compared with the comparative example, the Dice coefficient of the segmentation result of the tubular structure based on the first embodiment is improved, indicating that the segmentation result of the tubular structure is not thickened, and the number of connected component numbers in the segmentation result of the tubular structure is reduced, indicating that the connectivity of the segmentation result of the tubular structure is significantly improved.

[0088] (Second Implementation)

[0089] The second embodiment of the present invention will now be described. In the following description, only the differences between the second embodiment and the first embodiment will be described, and the similarities will be omitted.

[0090] Reference Figure 7 An example of the configuration of the image processing apparatus 1A according to the second embodiment will be described. Figure 7 This is a block diagram illustrating a configuration example of the image processing apparatus 1A according to the second embodiment.

[0091] In the second embodiment, the processing circuit 11A includes a receiving unit 110, an analysis unit 111, a differentiation unit 112, a tag acquisition unit 113A, a preprocessing unit 114, a training unit 115, and a reasoning unit 116.

[0092] The tag acquisition unit 113A establishes a functional relationship between the size of the lumen of the tubular structure calculated by the analysis unit 111 and the threshold of the interpolation method used. By applying the threshold of the interpolation method to each pixel in the medical image data, the resampled medical image data is processed and a tag as a binary tag is obtained.

[0093] Reference Figure 8 The functional relationship between the diameter of the lumen of the tubular structure and the threshold is explained. Figure 8 This is a schematic diagram illustrating the functional relationship between the diameter of the lumen of a tubular structure and a threshold.

[0094] Specifically, the tag acquisition unit 113A uses the following formula (1) to calculate the functional relationship between the size of the tubular structure's lumen and the threshold of the interpolation method used.

[0095]

[0096] Where R represents the diameter of the lumen of the tubular structure, and λ represents the hyperparameter. The hyperparameter λ is related to the minimum diameter of the tubular structure, and in the second embodiment, λ is, for example, 0.8.

[0097] According to equation (1), as Figure 8As shown, when the diameter R of the tubular structure's lumen is small, the threshold value is determined to be negative; when the diameter R of the tubular structure's lumen is large, the threshold value is determined to be positive. For example, in the second embodiment, when the diameter R of the tubular structure's lumen is less than or equal to 0.54, the threshold value is less than 0; when the diameter R of the tubular structure's lumen is 1, the threshold value is approximately 0.24. Therefore, the tag acquisition unit 113A is based on... Figure 8 The functional relationship shown determines the threshold of the interpolation method used based on the diameter R of the tubular structure's lumen.

[0098] Furthermore, the label acquisition unit 113A processes the resampled medical image data by applying a threshold to each pixel in the medical image data using the interpolation method, and obtains a label as a binary label. Here, "binary label" refers to a label obtained by simplifying complex anatomical information into 0 / 1 encoding.

[0099] In addition, the tag acquisition unit 113A compares the value of each pixel in the medical image data with a threshold. When the value of a pixel exceeds the threshold, it is determined that the pixel belongs to the tubular structure region, and when the value of a pixel is less than the threshold, it is determined that the pixel belongs to the background region.

[0100] This enables the accurate differentiation of tubular structures from background areas in medical image data, resulting in highly accurate labels.

[0101] Therefore, according to the second embodiment of the present invention, the threshold of the interpolation method can be applied to each pixel in the medical image data based on the functional relationship between the size of the tubular structure's lumen and the threshold of the interpolation method used. Thus, the labeling accuracy in the medical image data used as training data can be improved, and the tubular structure region and the background region can be accurately distinguished. Furthermore, the segmentation accuracy of the tubular structure can be further improved in the tubular structure segmentation task of medical image data.

[0102] (Flowchart of the second implementation method)

[0103] The following is for reference Figure 9 The process of the second embodiment will be described. Figure 9 This is a flowchart of the second implementation method.

[0104] In step S10, the receiving unit 110 receives medical image data. Then, the process proceeds to step S11.

[0105] In step S11, the analysis unit 111 extracts the centerline of the tubular structure from the medical image data received by the receiving unit 110. Then, the process proceeds to step S12.

[0106] In step S12, the analysis unit 111 performs a distance transformation on the centerline of the extracted tubular structure to calculate the diameter or radius of the lumen of the tubular structure. The analysis unit 111 sends the medical image data and the calculated size of the lumen of the tubular structure to the differentiation unit 112. Then, the process proceeds to step S13.

[0107] In step S13, the differentiation unit 112 assigns multi-level labels to the tubular structure based on the size of the lumen of the tubular structure calculated by the analysis unit 111, thus differentiating the tubular structure into multiple regions. The differentiation unit 112 sends medical image data of the tubular structure, which is divided into multiple regions, to the label acquisition unit 113A. Then, the process proceeds to step S14.

[0108] In step S14, the tag acquisition unit 113A resamples each region of the multiple regions of the tubular structure distinguished by the distinguishing unit 112 using a different interpolation method. Then, the process proceeds to step S25.

[0109] In step S25, the tag acquisition unit 113A establishes a functional relationship between the size of the tubular structure's lumen calculated by the analysis unit 111 and the threshold of the interpolation method used. Furthermore, based on the established functional relationship, the tag acquisition unit 113A determines the threshold of the interpolation method used according to the diameter R of the tubular structure's lumen. Then, the process proceeds to step S26.

[0110] In step S26, the tag acquisition unit 113A processes the resampled medical image data by applying the threshold of the interpolation method to each pixel in the medical image data, and acquires a tag as a binary tag. The tag acquisition unit 113A sends the medical image data and the acquired tag to the preprocessing unit 114. Then, the process proceeds to step S15.

[0111] In step S15, the preprocessing unit 114 preprocesses the resampled medical image data. The preprocessing unit 114 then sends the preprocessed medical image data to the training unit 115. Next, the process proceeds to step S16.

[0112] In step S16, the training unit 115 uses the medical image data preprocessed by the preprocessing unit 114 and the labels obtained by the label acquisition unit 113 as a training dataset to train the inference model. The training unit 115 outputs the trained inference model to the inference unit 116. Then, the process proceeds to step S17.

[0113] In step S17, the inference unit 116 uses the inference model trained by the training unit 115 to perform a segmentation task on the medical image data. This segments the tubular structures within the medical image data. The process then ends.

[0114] (Comparative Example)

[0115] Reference Figure 10 The differences between the second embodiment and the comparative example will be explained. Figure 10 This is a schematic diagram illustrating the difference between the segmentation results of the second implementation method and the comparative example based on the test data. Figure 10 The figure above shows the segmentation results of the tubular structure based on the comparative example (prior art). Figure 10 The figure below shows the segmentation result of the tubular structure based on the second embodiment.

[0116] like Figure 10 As shown, in the segmentation results of the tubular structure based on the second embodiment, compared with the segmentation results of the tubular structure based on the comparative example, as indicated by the arrow, a better segmentation result was obtained in the tubular structure region of the thin lumen, and the connectivity of the tubular structure region of the thin lumen was improved.

[0117] Furthermore, according to the second embodiment of the present invention, in a test set of medical image data containing 22 cases, compared with the comparative example, the Dice coefficient of the tubular structure segmentation result based on the second embodiment is improved, indicating that the segmentation result of the tubular structure has not become thicker, and the number of connected regions in the segmentation result of the tubular structure is reduced, indicating that the connectivity of the segmentation result of the tubular structure is significantly improved. In addition, the tree length detected rate (TD) and branch detected rate (BD) of the tubular structure are significantly improved, indicating that the under-segmentation of the tubular structure with thin lumens in the segmentation result is significantly reduced.

[0118] (Third Implementation)

[0119] The third embodiment of the present invention will now be described. In the following description, only the differences between the third embodiment and the first and second embodiments will be described, while the similarities will be omitted.

[0120] The difference between the third implementation method and the first and second implementation methods is that the third implementation method is used for post-processing of the trained inference model.

[0121] Reference Figure 11 An example of the configuration of the image processing apparatus 1B according to the third embodiment will be described. Figure 11 This is a block diagram illustrating a configuration example of the image processing apparatus 1B according to the third embodiment.

[0122] In the third embodiment, the processing circuit 11B includes a receiving unit 110B, an analysis unit 111B, a differentiation unit 112, a tag acquisition unit 113A, and an output unit 117.

[0123] The receiving unit 110B receives the predicted probability map output by the trained inference model. The receiving unit 110B then sends the predicted probability map output by the inference model to the analysis unit 111B.

[0124] The analysis unit 111B analyzes the size of the lumen of the tubular structure in the received prediction probability map. Specifically, the analysis unit 111B extracts the center line of the tubular structure in the prediction probability map, performs a distance transformation on the extracted center line of the tubular structure, and calculates the diameter or radius of the lumen of the tubular structure as the size of the lumen of the tubular structure.

[0125] The differentiation section 112 assigns multi-level labels to the tubular structure based on the size of the lumen of the extracted tubular structure, and distinguishes multiple regions of the tubular structure.

[0126] The tag acquisition unit 113A establishes a functional relationship between the size of the tubular structure cavity calculated by the analysis unit 111 and the threshold of the interpolation method used. By applying the threshold of the interpolation method to each pixel in the prediction probability map, the resampled prediction probability map is processed, and a tag is obtained as a binary tag.

[0127] The output unit 117 outputs the predicted probability map after the above processing as the segmentation result of the tubular structure based on the inference model.

[0128] Therefore, according to the third embodiment of the present invention, the image processing method of the present invention can be applied to the post-processing of the inference model. Consequently, the connectivity of the tubular structures in the segmentation results of the inference model can be improved.

[0129] (Fourth Implementation)

[0130] The fourth embodiment of the present invention will now be described. In the following description, only the differences between the fourth embodiment and the first to third embodiments will be described, and the similarities will be omitted.

[0131] The fourth embodiment differs from the first to third embodiments in that the fourth embodiment is used to measure the tubular structure of a thin lumen in medical image data.

[0132] Specifically, the tag obtained by the tag acquisition unit 113A of the second embodiment can be used to assist in the measurement of the tubular structure in the thin lumen.

[0133] For example, when the spacing of the original images in medical image data is large, the measurement may not be accurate enough at the original resolution. In this case, by resampling the labels using the second embodiment, an image with smaller spacing can be obtained, thereby enabling accurate measurement of the diameter of the tubular structure of the thin lumen on the image with smaller spacing.

[0134] In the above description, a tubular structure, specifically a blood vessel, was used as an example. However, the invention is not limited to this; the tubular structure can also be other tubular structures such as the trachea.

[0135] Furthermore, the example given is a multi-level tag containing two levels. However, this is not a limitation; a multi-level tag can also have two or more other numbers of levels.

[0136] Furthermore, the example given is "preprocessing," which involves normalizing resampled medical image data. However, this is not the only approach; other preprocessing techniques can also be applied to resampled medical image data.

[0137] Furthermore, in the above embodiments, the processing circuit is not limited to being implemented by a single processor; it can also be constructed by combining multiple independent processors, with each processor executing a program to implement each processing function. Additionally, the processing functions of the processing circuit can be appropriately distributed or integrated into one or more processing circuits. Furthermore, the processing functions of the processing circuit can also be implemented through a combination of hardware (such as circuits) and software. While an example has been described where the program corresponding to each processing function is stored in a single storage circuit, the implementation is not limited to this. For example, it is also possible to configure the program corresponding to each processing function to be distributed and stored in multiple storage circuits, with the processing circuit reading and executing each program from each storage circuit.

[0138] Furthermore, while the above embodiments illustrate examples where each functional part of this specification is implemented by processing circuitry, the embodiments are not limited to these examples. For instance, each functional part of this specification may be implemented not only by the processing circuitry described in the embodiments, but also by hardware alone, software alone, or a combination of hardware and software to achieve the same function.

[0139] Furthermore, the term "processor" used in the above description of the embodiments can refer to circuits such as CPU (Central Processing Unit), GPU (Graphics Processing Unit), Application Specific Integrated Circuit (ASIC), or programmable logic device (e.g., Simple Programmable Logic Device (SPLD), Complex Programmable Logic Device (CPLD), and Field Programmable Gate Array (FPGA)). Alternatively, instead of storing the computer program in a memory circuit, the computer program can be directly loaded into the processor circuit. In this case, the processor performs its function by reading and executing the program loaded into the circuit. Furthermore, the processors in this embodiment are not limited to being configured as a single circuit; multiple independent circuits can be combined to form a single processor and perform its functions.

[0140] Furthermore, the "computer program product" used in the above description of the embodiments includes a computer program executed by a processor, which is provided pre-loaded into ROM (Read Only Memory) or storage circuitry, etc. Additionally, the computer program can also be provided as a file that can be installed on or executed on computer-readable non-volatile storage media such as CD (Compact Disk)-ROM, FD (Flexible Disk), CD-R (Recordable), DVD (Digital Versatile Disk). Furthermore, the computer program can also be stored on a computer connected to a network such as the Internet and provided or distributed via network download. For example, the computer program is composed of modules including the aforementioned processing functions. As actual hardware, the CPU reads and executes the computer program from a storage medium such as ROM, loads each module onto the main storage device, and generates it on the main storage device.

[0141] Furthermore, in the above embodiments, the constituent elements of each device illustrated are functional conceptual elements and do not necessarily need to be physically configured as shown in the figures. That is, the specific manner in which the devices are distributed or concentrated is not limited to that illustrated, and can be configured in any unit, functionally or physically, in accordance with various loads, usage conditions, etc. Furthermore, all or any part of the processing functions performed by each device can also be implemented by a CPU and the program parsed and executed by the CPU, or as hardware implementation based on wiring logic.

[0142] Furthermore, among the processes described in the above embodiments, all or part of the processes described as automatically performed can be performed manually, or all or part of the processes described as manually performed can be performed automatically using known methods. In addition, the information regarding the processing procedures, control procedures, specific names, various data, or parameters shown in the above documents or drawings can be arbitrarily changed, except where specifically stated.

[0143] Furthermore, the various types of data processed in this specification are typically digital data.

[0144] Several embodiments have been described above, but these embodiments are given as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other ways, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope or spirit of the invention, and are included in the scope of the invention as described in the claims and its equivalents.

Claims

1. An image processing method, characterized in that, include: Steps for analyzing the size of the lumen of tubular structures in medical image data; The step of distinguishing multiple regions of the tubular structure according to the size of the lumen; as well as The steps include resampling the medical image data in multiple regions of the tubular structure and obtaining labels.

2. The image processing method according to claim 1, characterized in that, In the step of analyzing the size of the lumen It includes the steps of extracting the centerline of the tubular structure, performing distance transformation, and calculating the diameter or radius of the lumen.

3. The image processing method according to claim 1, characterized in that, In the step of obtaining the label For each of the multiple regions of the tubular structure, resampling is performed using different interpolation methods.

4. The image processing method according to claim 1, characterized in that, In the step of distinguishing the multiple regions Multiple levels of labels are assigned based on the size of the lumen.

5. The image processing method according to claim 4, characterized in that, Also includes: The step of preprocessing the resampled medical image data; The steps of training the inference model using the medical image data and the labels; as well as The steps of using the inference model to perform segmentation tasks on the medical image data.

6. The image processing method according to claim 4, characterized in that, In the step of obtaining the label Establish a functional relationship between the size and the threshold of the interpolation method. The threshold is applied to each pixel to process the resampled medical image data, and the label is obtained as a binary label.

7. An image processing method, characterized in that, The learned inference model is used to perform segmentation tasks on medical image data. The inference model is learned through the following method: Based on the size of the lumen of the tubular structure in the medical image data, the tubular structure is divided into multiple regions. In each of the multiple regions of the tubular structure, the medical image data is resampled and labeled.

8. An image processing apparatus, characterized in that, include: The analysis department analyzes the size of the lumen of tubular structures in medical image data; The segmentation section divides the tubular structure into multiple regions based on the size of the lumen; as well as The acquisition unit resamples the medical image data in multiple regions of the tubular structure and acquires labels.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed, it includes the following steps: Steps for analyzing the size of the lumen of tubular structures in medical image data; The step of distinguishing multiple regions of the tubular structure according to the size of the lumen; as well as The steps include resampling the medical image data in multiple regions of the tubular structure and obtaining labels.

10. An image processing method, characterized in that, Using the image processing method according to any one of claims 1 to 7, The reasoning results of the inference model are post-processed.

11. An image processing method, characterized in that, Using the image processing method according to any one of claims 1 to 7, The diameter of the tubular structure is measured in the medical image data containing the label.

Images