Methods and systems for predicting the efficacy of radiotherapy for osteogenic bone metastases

Abstract

This disclosure proposes a method and system for predicting the grading of radiotherapy efficacy for osteoblastic bone metastases, relating to the technical field of radiotherapy efficacy grading prediction for osteoblastic bone metastases. The method includes: acquiring CT scan images of the vertebral body corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases; based on relevant variables in the CT scan images of the vertebral body used to distinguish between osteoblastic bone metastases and bone islands, and using a preset radiotherapy efficacy grading prediction model for osteoblastic bone metastases, predicting the radiotherapy efficacy grading of osteoblastic bone metastases at a set time after at least one radiotherapy stage following radiotherapy. The embodiments of this disclosure can achieve the prediction of radiotherapy efficacy grading for osteoblastic bone metastases.

Description

Technical Field

[0001] This disclosure relates to the field of radiotherapy efficacy grading prediction for osteogenic bone metastases, and particularly to a method and system for radiotherapy efficacy grading prediction for osteogenic bone metastases. Background Technology

[0002] Osteoblastic bone metastases commonly occur in tumors such as prostate cancer, breast cancer, and kidney cancer, accounting for 10% of all bone metastases. Bone metastases are classified into osteolytic metastases, osteoblastic metastases, and mixed metastases. Among these, diagnosing osteoblastic bone metastases is a challenge in current imaging diagnostics, as it is easily confused with diseases such as bone islands. Symptoms of bone metastases include persistent and severe bone pain, pathological fractures and limited mobility, severe hypercalcemia, and compression of adjacent spinal cords and nerves. Failure to make a clear diagnosis and treat bone metastases in a timely manner can pose a significant risk to the patient's life and reduce their quality of life. Currently, bone metastases are often diagnosed through a clear history of the primary tumor, commonly used imaging diagnostic techniques (DR, CT, MR), emission CT (ECT), or positron emission tomography (PET).

[0003] The shortcomings of existing technologies are summarized as follows. Osteolytic and mixed bone metastases both appear as bone destruction on both digital radiography (DR) and computed tomography (CT), showing low-density or mixed low-density lesions, often accompanied by soft tissue masses. On magnetic resonance imaging (MR), osteolytic and mixed bone metastases often show long T1 and long T2 signals, and diffusion-restricted areas can be seen on diffusion-limited sequences. Enhanced MR scans can reveal varying degrees of enhancement of bone destruction and surrounding soft tissue masses. However, osteoblastic bone metastases and bone islands are very similar in imaging. These lesions are high-density on both DR and CT (CT has higher image resolution), while they are generally low-signal on various MR sequences. Therefore, it is often difficult to accurately diagnose and differentiate osteoblastic bone metastases and bone islands using MR.

[0004] Currently, the main treatment for bone metastases is symptomatic drug therapy. If a clear therapeutic effect is not achieved, radiotherapy, chemotherapy, and immunotherapy are required. Chemotherapy for bone metastases often fails to completely cure the tumor and can only be used to relieve symptoms and improve the patient's quality of life. For significant therapeutic effects on local bone metastases, local radiotherapy to the metastatic area is often necessary. When assessing the effectiveness of local radiotherapy, osteolytic and mixed metastatic bone metastases can be evaluated by changes in the size of the corresponding tumor lesion; however, this assessment method is not applicable to osteoblastic bone metastases. After radiotherapy, although the internal tumor tissue of osteoblastic bone metastases has undergone necrosis, changes in the size and density of the corresponding tumor lesion are often not significant due to ossification or bone hyperplasia. Therefore, current technology cannot accurately assess the radiotherapy effect on osteoblastic bone metastases.

[0005] Furthermore, while radionuclide imaging (CT) can diagnose all bone metastases, including osteoblastic bone metastases, it requires intravenous injection of radiolabeled phosphate compounds. PET scans also require the injection of radioactive drugs and are very expensive. Therefore, these techniques are not suitable for patients with bone metastases who are allergic to radiopharmaceuticals, pregnant or breastfeeding women, have severe cardiopulmonary or hepatic / renal insufficiency, or are financially disadvantaged. In addition, radioactive drugs pose a risk not only to the patient's own healthy tissues but also to the surrounding environment and the general population.

[0006] Therefore, finding a technical solution that poses no significant harm to patients with bone metastases and the surrounding environment, and that can accurately predict the efficacy grading of radiotherapy for osteoblastic bone metastases, is of great clinical significance. Summary of the Invention

[0007] This disclosure presents a technical solution for a method and system for predicting the efficacy of radiotherapy for osteogenic bone metastases.

[0008] According to one aspect of this disclosure, a method for predicting the grading of radiotherapy efficacy for osteoblastic bone metastases is provided, comprising: acquiring vertebral images of the target spectral CT scan corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases; based on the relevant variables of the target spectral CT scan vertebral images for differentiating osteoblastic bone metastases from bone islands, using a preset osteoblastic bone metastases radiotherapy efficacy grading prediction model, predicting the grading of osteoblastic bone metastases radiotherapy efficacy at a set radiotherapy time after at least one radiotherapy stage following radiotherapy.

[0009] Preferably, determining the relevant variables for differentiating osteogenic bone metastasis from bone islands includes: extracting multiple imaging parameters corresponding to lesions in multiple first-spectrum CT scan vertebral images of the osteogenic bone metastasis study group and multiple imaging parameters corresponding to lesions in multiple second-spectrum CT scan vertebral images of the bone island control group; determining corresponding quantitative and count features based on the multiple imaging parameters of the study group and the multiple imaging parameters of the control group; and using a feature screening algorithm to screen the risk factors constructed from the quantitative and count features to obtain the relevant variables for differentiating osteogenic bone metastasis from bone islands.

[0010] Preferably, the extraction of multiple imaging parameters corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group includes: extracting features from the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group to obtain one or more of the following multiple imaging parameters: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the long diameter to the short diameter of the lesion with the largest cross-sectional area, whether the boundary of the lesion in the research group is clear, whether the location of the lesion in the research group is close to the bone cortex, and whether there is a spinous process sign at the edge of the lesion in the research group.

[0011] Preferably, the step of extracting features from multiple first-spectrum CT scan vertebral images corresponding to the osteoblastic bone metastasis research group to obtain one or more of the following imaging parameters: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the long axis to the short axis of the research group of the lesion with the largest cross-sectional area, whether the boundary of the research group lesion is clear, whether the location of the research group lesion is close to the bone cortex, and whether there is a spinous process sign at the edge of the research group lesion, includes: using a first preset energy spectrum CT scan vertebral image lesion segmentation model to segment the osteoblastic bone metastasis lesion. The lesions in multiple first-spectrum CT scan vertebral body images corresponding to the transfer research group are segmented to obtain multiple first lesion mask images. Based on the multiple first-spectrum CT scan vertebral body images and the multiple first lesion mask images, one or more of the following imaging parameters are extracted: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the long axis to the short axis of the lesion with the largest cross-sectional area, whether the boundary of the research group lesion is clear, whether the location of the research group lesion is close to the bone cortex, and whether there is a spinous process sign at the edge of the research group lesion.

[0012] Preferably, the average effective atomic number value of the study group corresponding to the lesion is extracted based on the plurality of first spectral CT scan vertebral body images and the plurality of first lesion mask images, including: using a spectral CT atomic number analysis module configured with spectral CT images to extract the average effective atomic number value of the study group corresponding to the lesion based on the plurality of first spectral CT scan vertebral body images and the plurality of first lesion mask images.

[0013] Preferably, based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first-lesion mask images, the average CT value of the study group corresponding to the lesion is extracted, including: performing pixel-level multiplication operations on each first-spectrum CT scan vertebral body image and its corresponding first-lesion mask image in the plurality of first-spectrum CT scan vertebral body images to obtain a plurality of first-spectrum CT scan lesion images corresponding to the plurality of first-spectrum CT scan vertebral body images; calculating the CT value of each lesion corresponding to the plurality of first-spectrum CT scan lesion images; and averaging the CT values ​​of each lesion to determine the average CT value of the study group corresponding to the lesion.

[0014] Preferably, based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images, the study group ratio of the major and minor axes of the lesion with the largest cross-sectional area is extracted, including: calculating the lesion mask cross-sectional area of ​​each lesion in the first lesion mask image corresponding to each of the plurality of first-spectrum CT scan vertebral body images; calculating the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion; and determining the study group ratio corresponding to each first-spectrum CT scan vertebral body image based on the major and minor axes of the lesion with the largest cross-sectional area.

[0015] Preferably, calculating the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion includes: performing edge detection on the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion to obtain the edge of the lesion with the largest cross-sectional area; selecting any edge pixel point in the edge of the lesion with the largest cross-sectional area; calculating multiple edge pixel point distances between the arbitrary edge pixel point and other edge pixel points in the edge of the lesion with the largest cross-sectional area; selecting the maximum edge pixel point distance and the minimum edge pixel point distance from the multiple edge pixel point distances; and configuring the maximum edge pixel point distance and the minimum edge pixel point distance as the major and minor axes of the lesion with the largest cross-sectional area, respectively.

[0016] Preferably, determining whether the lesion boundaries in the study group are clear based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images includes: performing an erosion operation on the lesion mask in the first lesion mask image corresponding to each of the plurality of first-spectrum CT scan vertebral body images to obtain a plurality of first lesion mask erosion images; subtracting the corresponding plurality of first lesion mask erosion images from the plurality of first lesion mask images to obtain a plurality of first lesion mask processed images; mapping the coordinates of multiple lesions in the plurality of first lesion mask processed images to the corresponding first-spectrum CT scan vertebral body images; calculating a plurality of first gradients corresponding to the coordinates of multiple lesions in the first-spectrum CT scan vertebral body images; if the first gradient magnitude value corresponding to the plurality of first gradients is greater than or equal to a first preset gradient magnitude value, then the lesion boundary where the lesion location is greater than or equal to the first preset gradient magnitude value is determined to be clear; otherwise, the lesion boundary is determined to be blurred.

[0017] Preferably, determining whether the lesion location in the study group is adjacent to the bone cortex includes: detecting the bone cortex edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images; and determining whether the lesion location in the study group is adjacent to the bone cortex based on the bone cortex edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images.

[0018] Preferably, the method for extracting multiple imaging parameters of the control group corresponding to the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group is the same as the method for extracting multiple imaging parameters of the study group corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group, including: performing feature extraction on the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group to obtain one or more of the following multiple imaging parameters of the control group: the average effective atomic number value of the control group corresponding to the lesion, the average CT value of the control group corresponding to the lesion, the ratio of the long diameter to the short diameter of the lesion with the largest cross-sectional area to the control group, whether the boundary of the lesion in the control group is clear, whether the location of the lesion in the control group is adjacent to the bone cortex, and whether there is a spinous process sign at the edge of the lesion in the control group.

[0019] Preferably, determining the corresponding quantitative features and count features based on the multiple imaging parameters of the study group and the multiple imaging parameters of the control group includes: grouping the multiple imaging parameters of the study group and the multiple imaging parameters of the control group according to the principles of measurement and counting to obtain the quantitative features and count features to be processed; performing an independent samples t-test on the quantitative features to be processed corresponding to the study group and the control group, and selecting quantitative features with a significance level less than or equal to a first preset significance level from the quantitative features to be processed; performing a chi-square test on the count features to be processed corresponding to the study group and the control group, and selecting count features with a significance level less than or equal to a second preset significance level from the count features to be processed; and / or, the feature selection algorithm is configured as a multi-factor logistic regression analysis algorithm.

[0020] Preferably, determining the preset osteogenic bone metastasis radiotherapy efficacy grading prediction model includes: obtaining a preset number of relevant variables for distinguishing osteogenic bone metastasis from bone islands at the first time point and osteogenic bone metastasis radiotherapy efficacy grading labels at the second time point after the first time point; training the preset prediction network or preset predictor to obtain the preset osteogenic bone metastasis radiotherapy efficacy grading prediction model.

[0021] According to one aspect of this disclosure, a radiotherapy efficacy grading prediction system for osteogenic bone metastases is provided, comprising: an acquisition unit for acquiring vertebral body images of the target spectral CT scan corresponding to at least one radiotherapy stage before and after radiotherapy for osteogenic bone metastases; and a prediction unit for predicting the osteogenic bone metastases radiotherapy efficacy grading at a set radiotherapy time after at least one radiotherapy stage following radiotherapy, based on relevant variables for distinguishing between osteogenic bone metastases and bone islands from the target spectral CT scan vertebral body images, using a preset osteogenic bone metastases radiotherapy efficacy grading prediction model.

[0022] According to one aspect of this disclosure, a system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the above-described method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0023] According to one aspect of this disclosure, a system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases is provided, comprising: a computer-readable storage medium having a computer program / instructions and a bit stream stored thereon, wherein the computer program / instructions, when executed by a processor, generate the bit stream by implementing the above-described method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0024] According to one aspect of this disclosure, a system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases is provided, comprising: a computer program product configured with a computer program / instruction, which, when executed by a processor, implements the aforementioned method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0025] In the embodiments of this disclosure, a technical solution is proposed for a method and system for predicting the efficacy grading of radiotherapy for osteogenic bone metastases, in order to solve the technical problem that it is difficult to effectively predict the efficacy grading of radiotherapy for osteogenic bone metastases in the prior art.

[0026] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0027] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0028] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0029] Figure 1 A flowchart illustrating a method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to an embodiment of the present disclosure is shown. Figure 2 A flowchart is shown showing a method for feature extraction of lesions from multiple first-spectrum CT scan vertebral images corresponding to the osteoblastic bone metastasis research group, according to an embodiment of the present disclosure. Figure 3 A flowchart is shown showing a method for extracting the ratio of the major and minor axes of the lesion with the largest cross-sectional area based on the plurality of first energy spectrum CT scan vertebral body images and the plurality of first lesion mask images according to an embodiment of the present disclosure; Figure 4 This is a block diagram illustrating a radiotherapy efficacy grading prediction system for osteogenic bone metastases according to an exemplary embodiment. Detailed Implementation

[0030] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0031] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0032] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0033] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0034] It is understood that the various method embodiments mentioned above in this disclosure can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this disclosure will not elaborate further.

[0035] In addition, this disclosure also provides a system for predicting the efficacy grading of radiotherapy for osteogenic bone metastases. All of the above can be used to implement any of the methods for predicting the efficacy grading of radiotherapy for osteogenic bone metastases provided in this disclosure. The corresponding technical solutions and descriptions are described in the relevant section on methods for predicting the efficacy grading of radiotherapy for osteogenic bone metastases, and will not be repeated here.

[0036] Figure 1 A flowchart illustrating a method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to embodiments of the present disclosure is shown, such as... Figure 1 As shown, the method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases includes: Step S101: acquiring vertebral body images of the target spectral CT scan corresponding to at least one radiotherapy stage before and after radiotherapy for osteogenic bone metastases; Step S102: based on the relevant variables of the target spectral CT scan vertebral body images used to distinguish between osteogenic bone metastases and bone islands, using a preset osteogenic bone metastases radiotherapy efficacy grading prediction model, predicting the grading of osteogenic bone metastases radiotherapy efficacy at a set radiotherapy time after at least one radiotherapy stage. This solves the technical problem in the prior art of effectively predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0037] This invention proposes to utilize the atomic number analysis module of spectral CT, which can accurately diagnose osteoblastic bone metastases and differentiate them from common bone island lesions through techniques such as atomic number values ​​and atomic number pseudo-color images. Furthermore, by performing multiple analyses using the atomic number analysis module before, during, or after radiotherapy for osteoblastic bone metastases, the efficacy grading of radiotherapy for bone metastases can be accurately predicted. Most importantly, unlike radionuclide imaging and PET scans, the atomic number analysis module of spectral CT does not require additional intravenous injection of radioactive drugs, thus posing no radiation hazard to patients or the surrounding healthy population. Moreover, spectral CT has similar scanning doses and examination costs to conventional CT scans, minimizing the health and financial burden on patients with bone metastases.

[0038] Step S101: Obtain vertebral images of the target spectral CT scan corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases.

[0039] In this disclosure and other possible embodiments, before obtaining the predicted spectral CT vertebral body images corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases, the method includes: performing spectral CT scans on the vertebral body of the patient with osteoblastic bone metastases before and after at least one radiotherapy stage, to obtain the predicted spectral CT vertebral body images corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases. The setting of the radiotherapy time after at least one radiotherapy stage can be configured to be set after 1 month, 2 months, 3 months, or any other time after radiotherapy.

[0040] Unlike conventional CT, spectral CT is essentially a type of rapid tube voltage-switching dual-source CT. Rapid tube voltage-switching dual-source CT uses two single-energy X-rays for imaging, including: first, obtaining basic attenuation measurements using the first single-energy X-ray spectrum; then, obtaining additional attenuation measurements using the second single-energy X-ray spectrum; and finally, distinguishing and identifying different material components based on the basic and additional attenuation measurements. Furthermore, compared to spectral CT, conventional dual-source CT, which uses two X-ray tubes, is susceptible to cross-scattering during imaging, causing spatial positioning difficulties and affecting CT image quality to some extent. In contrast, spectral CT (i.e., rapid tube voltage-switching dual-source CT) employs instantaneous kVp switching technology to complete the switching between high and low single-energy X-rays in an extremely short time (ms), achieving simultaneous, co-directional, and co-source emission of dual-energy X-rays. This overcomes the aforementioned shortcomings of dual-X-ray tube dual-source CT, avoids the cross-scattering effects caused by subtle angular differences between the two X-ray tube scanning planes, and thus improves imaging accuracy, providing higher-quality spectral CT scan images.

[0041] The high-speed flash volume platform (GSI Xtream) used in energy dispersive spectroscopy (EDS) can generate single-energy spectral images, material density images, virtual plain scan images, metal artifact-free images, and atomic number maps corresponding to EDS scan images. The atomic number analysis module of EDS can measure the effective atomic number values ​​of different substances based on the atomic number map and display these values ​​as quantitative atomic number bar charts, generating corresponding pseudo-color atomic number images for easier observation. Specifically, the effective atomic number value, or Zeff value, refers to the Zeff value of a compound that can be represented by the Zeff value of the element when the X-ray attenuation coefficients of a certain element and a certain compound are the same. The Zeff value can be calculated using a formula based on the elemental composition and material density, taking into account the different absorption of photons by the material at different energy levels.

[0042] Step S102: Based on the relevant variables of the vertebral body image of the CT scan to be predicted for identifying osteogenic bone metastases and bone islands, use a preset osteogenic bone metastasis radiotherapy efficacy grading prediction model to predict the osteogenic bone metastasis radiotherapy efficacy grading at a set time after at least one radiotherapy stage.

[0043] CT values ​​obtained from conventional non-dual-source CT images cannot accurately differentiate between osteogenic bone metastases and bone islands. This is because conventional non-dual-source CT, relying on a single X-ray absorption coefficient, cannot quantitatively assess the degree of tissue density. For tissue structures with inherently high density (such as bone), uncontrollable changes during X-ray absorption attenuation (i.e., the influence of sclerosis artifacts) often lead to significant deviations in CT values ​​from conventional non-dual-source CT images. Furthermore, the CT values ​​of osteogenic bone metastases and bone islands are very similar, thus conventional non-dual-source CT cannot accurately differentiate between them. Compared to CT values ​​from conventional non-dual-source CT, the effective atomic number value, or Zeff value, exhibits significant tissue specificity and is particularly suitable for bone tissues containing a high proportion of inorganic components (such as calcium and phosphorus). Furthermore, spectral CT uses two different energies of X-rays for scanning. The first single-energy X-ray spectrum provides the basic attenuation measurement, and the second single-energy X-ray spectrum provides additional attenuation measurements. This can significantly reduce hardening artifacts, resulting in more accurate CT scan images.

[0044] In the embodiments disclosed herein and other possible embodiments, unlike the prior art, the relevant variables used to identify osteogenic bone metastases and bone islands are used as a key technical solution for predicting the efficacy grading of radiotherapy for osteogenic bone metastases, which directly affects the accuracy of the efficacy grading prediction of radiotherapy for osteogenic bone metastases.

[0045] In this embodiment of the disclosure, determining the relevant variables for differentiating osteogenic bone metastasis from bone islands includes: extracting multiple imaging parameters corresponding to lesions in multiple first-spectrum CT scan vertebral images of the osteogenic bone metastasis study group and multiple imaging parameters corresponding to lesions in multiple second-spectrum CT scan vertebral images of the bone island control group; determining corresponding quantitative and count features based on the multiple imaging parameters of the study group and the multiple imaging parameters of the control group; and using a feature screening algorithm to screen the risk factors constructed by the quantitative and count features to obtain the relevant variables for differentiating osteogenic bone metastasis from bone islands.

[0046] Multiple first-spectrum CT images of the vertebral body corresponding to osteoblastic bone metastases were selected as the study group, while multiple second-spectrum CT images of the vertebral body corresponding to bone island cases were selected as the control group. Inclusion criteria for the study group: ① The primary malignant tumors corresponding to the osteoblastic bone metastases in all patients were confirmed by pathological examination; ② The patients had not received anti-tumor treatment before the spectral CT examination; ③ The spectral CT showed high-density changes in the tumor corresponding to the intraosseous lesions; ④ Radionuclide examination confirmed osteoblastic bone metastases before the spectral CT examination. Inclusion criteria for the control group: ① No history of primary malignant tumors; ② The spectral CT showed high-density changes in the intraosseous lesions; ③ Radionuclide examination confirmed bone islands before the spectral CT examination.

[0047] In the embodiments of this disclosure and other possible embodiments, before extracting multiple imaging parameters of the study group corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group, the method includes: performing vertebral scans on each patient with osteogenic bone metastasis in a first number of osteogenic bone metastasis study groups using spectral CT, thereby obtaining multiple first-spectrum CT scan vertebral images corresponding to the first number of osteogenic bone metastasis study groups. In the embodiments of this disclosure and other possible embodiments, each of the multiple first-spectrum CT scan vertebral images includes: multiple spectral CT scan vertebral images; the multiple spectral CT scan vertebral images corresponding to each study subject in the study group constitute a three-dimensional first-spectrum CT scan vertebral image.

[0048] Similarly, in the embodiments of this disclosure and other possible embodiments, before extracting multiple imaging parameters of the control group corresponding to the lesions in the multiple second-spectrum CT scan vertebral body images corresponding to the bone island control group, the method includes: performing vertebral body scans on each bone island patient in the second number of bone island control groups using spectral CT, thereby obtaining multiple second-spectrum CT scan vertebral body images corresponding to the second number of bone island control groups. In the embodiments of this disclosure and other possible embodiments, each of the multiple second-spectrum CT scan vertebral body images includes: multiple spectral CT scan vertebral body images; the multiple spectral CT scan vertebral body images corresponding to each control subject in the control group constitute a three-dimensional second-spectrum CT scan vertebral body image.

[0049] In embodiments of this disclosure and other possible embodiments, the configuration of the first number and the second number satisfies statistical significance. For example, the configuration of the first number and the second number satisfies the number requirements for independent samples t-tests and chi-square tests.

[0050] In this embodiment of the disclosure, the extraction of multiple imaging parameters corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group includes: performing feature extraction on the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group to obtain one or more of the following multiple imaging parameters: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the long diameter to the short diameter of the lesion with the largest cross-sectional area, whether the boundary of the research group lesion is clear, whether the location of the research group lesion is adjacent to the bone cortex, and whether there is a spinous process sign at the edge of the research group lesion.

[0051] For each lesion in both the study and control groups, spectral CT scans of the vertebral body were used to measure the effective atomic number of each component within each lesion, the average effective atomic number of all components in all lesions, and the average CT value of all lesions. Specifically, spectral CT scans of the vertebral body were used to generate quantitative bar charts of the atomic number of each component within each lesion. The effective atomic number of each component within each lesion was determined based on the multiple peaks corresponding to the quantitative bar charts. The average effective atomic number of all components within all lesions was then calculated by averaging the effective atomic number values.

[0052] In addition, it is necessary to measure the ratio of the long axis to the short axis of the largest cross-sectional area lesion at the bone window level among all lesions, the clarity or indistinctness of the lesion boundary (whether the lesion boundary is clear), the location of the lesion (whether it is adjacent to the cortical bone), and whether there is a spiky sign at the edge of the lesion. The spiky sign is that the edge of the lesion is accompanied by radial bone stripes, and the bone stripes fuse with the adjacent trabeculae of the cancellous bone to form a "spiky" change.

[0053] In the embodiments disclosed herein and other possible embodiments, the extraction of multiple imaging parameters corresponding to lesions in multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group includes at least imaging parameters related to whether the lesion boundary is clear and / or whether the lesion location is adjacent to the cortical bone. The extraction of multiple imaging parameters corresponding to lesions in multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group further includes: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the major axis to the minor axis of the lesion with the largest cross-sectional area, and one or more of the multiple imaging parameters of the research group, including whether there is a spinous process sign at the edge of the lesion.

[0054] In the embodiments disclosed herein and other possible embodiments, the number of one or more imaging parameters of the study group corresponding to the lesion, the average effective atomic number value of the study group corresponding to the lesion, the average CT value of the study group corresponding to the lesion, the ratio of the long diameter to the short diameter of the lesion with the largest cross-sectional area, whether the boundary of the lesion in the study group is clear, whether the location of the lesion in the study group is close to the cortical bone, and whether there is a spinous process sign at the edge of the lesion in the study group, is consistent with the first number corresponding to the multiple first energy spectrum CT scan vertebral body images in the study group.

[0055] Figure 2 A flowchart illustrating a method for feature extraction from multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group, according to an embodiment of this disclosure, is provided. Figure 2 As shown in this embodiment, the step of extracting features from multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group to obtain one or more of the following imaging parameters: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the long axis to the short axis of the research group of the lesion with the largest cross-sectional area, whether the boundary of the research group lesion is clear, whether the location of the research group lesion is close to the bone cortex, and whether there is a spinous process sign at the edge of the research group lesion. This includes: Step 1101: Using the first preset energy spectrum CT scan vertebral image lesion segmentation model, the feature extraction is performed on the lesion of the research group. In the osteogenic bone metastasis study group, lesions in multiple first-spectrum CT scan vertebral body images are segmented to obtain multiple first lesion mask images; Step 1102: Based on the multiple first-spectrum CT scan vertebral body images and the multiple first lesion mask images, one or more of the following imaging parameters are extracted: the average effective atomic number value of the study group corresponding to the lesion, the average CT value of the study group corresponding to the lesion, the ratio of the long axis to the short axis of the lesion with the largest cross-sectional area, whether the boundary of the study group lesion is clear, whether the location of the study group lesion is close to the bone cortex, and whether there is a spinous process sign at the edge of the study group lesion.

[0056] In the embodiments of this disclosure and other possible embodiments, before segmenting lesions in multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group using a first preset energy-spectrum CT scan lesion segmentation model to obtain multiple corresponding first lesion mask images, the method includes: training a first preset deep learning segmentation network using a preset number of energy-spectrum CT scan vertebral training images and their corresponding energy-spectrum CT scan lesion mask training images to obtain a first preset energy-spectrum CT scan vertebral image lesion segmentation model. Each first lesion mask image and each energy-spectrum CT scan lesion mask training image may contain multiple lesion masks.

[0057] In the embodiments disclosed herein and other possible embodiments, the lesions in the plurality of first spectral CT scan vertebral images are configured as osteoblastic bone metastases; the mask value corresponding to each lesion region in the plurality of first lesion mask images is configured as 1, and the mask value corresponding to the non-lesion region is configured as 0.

[0058] In the embodiments of this disclosure and other possible embodiments, before training the first preset deep learning segmentation network using a preset number of spectral CT scan vertebral body training images and their corresponding spectral CT scan lesion mask training images to obtain the first preset spectral CT scan vertebral body image lesion segmentation model, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate lesions in the preset number of spectral CT scan vertebral body training images to obtain corresponding spectral CT scan lesion mask training images.

[0059] In the embodiments disclosed herein and other possible embodiments, the first preset deep learning segmentation network is configured as one or more deep learning segmentation networks such as Fully Convolutional Network (FCN), U-Net, DeepLab, and Pyramid Spatial Pooling Network (PSPNet).

[0060] In this embodiment of the disclosure, the average effective atomic number value of the study group corresponding to the lesion is extracted based on the plurality of first spectral CT scan vertebral body images and the plurality of first lesion mask images. This includes: using a spectral CT atomic number analysis module configured with spectral CT images, the average effective atomic number value of the study group corresponding to the lesion is extracted based on the plurality of first spectral CT scan vertebral body images and the plurality of first lesion mask images.

[0061] In this embodiment of the disclosure, the spectral CT atomic number analysis module configured using spectral CT images extracts the average effective atomic number value of the study group corresponding to the lesion based on the plurality of first spectral CT scan vertebral body images and the plurality of first lesion mask images. This includes: using the spectral CT atomic number analysis module configured using spectral CT images, extracting a quantitative bar chart of the atomic number corresponding to each component within each lesion generated from each first spectral CT scan vertebral body image and its corresponding first lesion mask image; determining the effective atomic number value of each component within each lesion based on multiple peak values ​​corresponding to the quantitative bar chart of the atomic number corresponding to each component within each lesion; and averaging the effective atomic number values ​​of all components within all lesions to determine the average effective atomic number value of the study group for all components within the lesions corresponding to the plurality of first spectral CT scan vertebral body images.

[0062] In the embodiments of this disclosure and other possible embodiments, the step of extracting a quantitative bar chart of the atomic number corresponding to each component within each lesion generated from each of the plurality of first-spectrum CT scan vertebral body images and their corresponding first lesion mask images includes: performing pixel-level multiplication operations on each of the plurality of first-spectrum CT scan vertebral body images and their corresponding first lesion mask images to obtain a plurality of first-spectrum CT scan lesion images corresponding to each of the plurality of first-spectrum CT scan vertebral body images; and obtaining a quantitative bar chart of the atomic number corresponding to each component within each lesion generated from each of the plurality of first-spectrum CT scan lesion images based on the plurality of first-spectrum CT scan lesion images.

[0063] In this embodiment of the disclosure, the average CT value of the study group corresponding to the lesion is extracted based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images. This includes: performing pixel-level multiplication operations on each first-spectrum CT scan vertebral body image and its corresponding first lesion mask image in the plurality of first-spectrum CT scan vertebral body images to obtain a plurality of first-spectrum CT scan lesion images corresponding to the plurality of first-spectrum CT scan vertebral body images; calculating the CT value of each lesion corresponding to the plurality of first-spectrum CT scan lesion images; and averaging the CT values ​​of each lesion to determine the average CT value of the study group corresponding to the lesion.

[0064] Figure 3 A flowchart illustrating a method for extracting the major and minor axes of a lesion with the largest cross-sectional area based on a plurality of first-energy-spectrum CT scan vertebral body images and a plurality of first-lesion mask images, according to an embodiment of this disclosure, is provided. Figure 3As shown, based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images, the study group ratio of the major and minor axes of the lesion with the largest cross-sectional area is extracted, including: Step 1103: Calculate the lesion mask cross-sectional area of ​​each lesion in the first lesion mask image corresponding to each of the plurality of first-spectrum CT scan vertebral body images; Step 1104: Calculate the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion; Based on the major and minor axes of the lesion with the largest cross-sectional area, determine the study group ratio corresponding to each first-spectrum CT scan vertebral body image.

[0065] In the embodiments of this disclosure and other possible embodiments, the step of calculating the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion includes: performing edge detection on the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion to obtain the edge of the lesion with the largest cross-sectional area; selecting any edge pixel point in the edge of the lesion with the largest cross-sectional area; calculating multiple edge pixel point distances between the arbitrary edge pixel point and other edge pixel points in the edge of the lesion with the largest cross-sectional area; selecting the maximum edge pixel point distance and the minimum edge pixel point distance from the multiple edge pixel point distances; and configuring the maximum edge pixel point distance and the minimum edge pixel point distance as the major and minor axes of the lesion with the largest cross-sectional area, respectively.

[0066] In the embodiments disclosed herein and other possible embodiments, the clarity of lesion boundaries is of significant reference value in clinical diagnosis and treatment, including: determining benignity or malignancy, assessing inflammation and infection, and formulating treatment strategies. Firstly, clear boundaries usually indicate that the lesion may be benign, such as cysts, lipomas, or some benign tumors. These lesions often have a complete capsule or localized growth, clear boundaries with surrounding tissues, slow growth, and low invasiveness. Unclear boundaries are often associated with malignant lesions, such as lung cancer, breast cancer, and liver cancer. These malignant tumors have invasive cells that infiltrate surrounding tissues, leading to blurred boundaries, which may be accompanied by spiculated, lobulated, or spiky morphologies. Secondly, some inflammatory lesions may have relatively clear boundaries when the inflammation is localized, indicating that the inflammation is in a relatively stable stage. If the inflammation is in an active phase, inflammatory cells infiltrate the surrounding tissues, and the boundaries become blurred, often accompanied by symptoms such as fever, cough, and pain, requiring further anti-infective or anti-inflammatory treatment. Finally, if the boundaries are clear, for benign lesions that are small and asymptomatic, an observation and follow-up strategy is usually adopted, with regular follow-up imaging examinations. If treatment is required, local treatments such as surgical resection or minimally invasive ablation are effective and have a low recurrence rate. If the boundaries are unclear, the possibility of malignancy should be highly suspected, and further examinations, such as puncture biopsy and enhanced CT / MRI, are usually recommended to clarify the pathological diagnosis. If malignancy is diagnosed, a comprehensive treatment plan should be developed according to the stage, such as surgical resection, chemotherapy, radiotherapy, and targeted therapy. Based on the above important reference significance for clinical diagnosis and treatment, and differentiating itself from existing technologies, a technical solution for determining whether the lesion boundary is clear is proposed to solve at least one of the technical problems of existing technologies that require manual determination of lesion edge clarity, which is too subjective and lacks intelligent determination of lesion edge clarity, resulting in a huge workload for doctors.

[0067] In this embodiment of the disclosure, determining whether the lesion boundaries of the research group are clear based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images includes: performing an erosion operation on the lesion mask in the first lesion mask image corresponding to each of the plurality of first-spectrum CT scan vertebral body images to obtain a plurality of first lesion mask erosion images; subtracting the corresponding plurality of first lesion mask erosion images from the plurality of first lesion mask images to obtain a plurality of first lesion mask processed images; mapping the coordinates of multiple lesions in the plurality of first lesion mask processed images to the corresponding first-spectrum CT scan vertebral body images; calculating a plurality of first gradients corresponding to the coordinates of multiple lesions in the first-spectrum CT scan vertebral body images; if the first gradient magnitude value corresponding to the plurality of first gradients is greater than or equal to a first preset gradient magnitude value, then the lesion boundary where the lesion location is greater than or equal to the first preset gradient magnitude value is determined to be clear; otherwise, the lesion boundary is determined to be blurred.

[0068] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the first preset gradient magnitude value according to actual needs. For example, the first preset gradient magnitude value can be configured to 1-80 or other possible values.

[0069] In the embodiments of this disclosure and other possible embodiments, after subtracting the corresponding first lesion mask erosion images from the plurality of first lesion mask images to obtain a plurality of first lesion mask processing images, the plurality of first lesion mask processing images are configured as a plurality of first lesion edge mask images.

[0070] In the embodiments of this disclosure and other possible embodiments, the step of calculating multiple first gradients corresponding to multiple lesion location coordinates in the first energy-spectrum CT scan vertebral body image includes: determining multiple first position coordinates and multiple second position coordinates on both sides of the multiple lesion location coordinates in the first energy-spectrum CT scan vertebral body image in the x-direction, and multiple third position coordinates and multiple fourth position coordinates corresponding to the y-direction, based on the multiple lesion location coordinates; calculating the difference between the first pixel value corresponding to the multiple first position coordinates and the second pixel value corresponding to the multiple second position coordinates to obtain multiple first x-direction gradients corresponding to the multiple lesion location coordinates; and calculating the difference between the third pixel value corresponding to the multiple third position coordinates and the fourth pixel value corresponding to the multiple fourth position coordinates to obtain multiple first y-direction gradients corresponding to the multiple lesion location coordinates.

[0071] In the embodiments of this disclosure and other possible embodiments, determining the first gradient magnitude corresponding to the plurality of first gradients includes: squaring the plurality of first x-direction gradients corresponding to the plurality of lesion location coordinates to obtain a plurality of first x-direction squared gradients; squaring the plurality of first y-direction gradients corresponding to the plurality of lesion location coordinates to obtain a plurality of first y-direction squared gradients; summing the plurality of first x-direction squared gradients with their corresponding first y-direction squared gradients to obtain the sum of the first gradients corresponding to the plurality of lesion location coordinates; and taking the square root of the sum of the first gradients corresponding to the plurality of lesion location coordinates to obtain the first gradient magnitude corresponding to the plurality of first gradients.

[0072] In the embodiments disclosed herein and other possible embodiments, multiple lesion location coordinates (x1, y1), (x2, y2), ..., (x n ,y n ), determine the multiple first position coordinates (x1+1, y1), (x2+1, y2), ..., (x2+1, y2) on both sides of the multiple lesion location coordinates in the first energy spectrum CT scan vertebral body image in the x direction. n +1,y n) and multiple second position coordinates (x1-1, y1), (x2-1, y2), ..., (x n -1,y n ) and the corresponding third position coordinates in the y direction (x1, y1+1), (x2, y2+1), ..., (x n ,y n +1) and multiple fourth position coordinates (x1, y1-1), (x2, y2-1), ..., (x n ,y n -1); calculate the differences between the first pixel values ​​p1 corresponding to multiple first position coordinates and the second pixel values ​​p2 corresponding to multiple second position coordinates to obtain multiple first x-direction gradients corresponding to the multiple lesion position coordinates; calculate the differences between the third pixel values ​​p3 corresponding to multiple third position coordinates and the fourth pixel values ​​p4 corresponding to multiple fourth position coordinates to obtain the multiple lesion position coordinates (x1, y1), (x2, y2), ..., (x n ,y n The corresponding gradients in the first y-direction.

[0073] In the embodiments of this disclosure and other possible embodiments, the step of performing an erosion operation on the lesion mask in the plurality of first lesion mask images to obtain a plurality of first lesion mask eroded images includes: using a set erosion pixel to perform an erosion operation on the lesion mask in the plurality of first lesion mask images to obtain a plurality of first lesion mask eroded images.

[0074] In this embodiment of the disclosure, determining whether the lesion location in the study group is adjacent to the bone cortex includes: detecting the bone cortex edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images; and determining whether the lesion location in the study group is adjacent to the bone cortex based on the bone cortex edge mask position coordinates corresponding to each of the first-spectrum CT scan vertebral body images and the plurality of first lesion edge mask position coordinates corresponding to each of the first-spectrum CT scan vertebral body images.

[0075] In this embodiment of the disclosure, the method for extracting multiple imaging parameters of the control group corresponding to the lesions in multiple second-spectrum CT scan vertebral images corresponding to the bone island control group is the same as the method for extracting multiple imaging parameters of the study group corresponding to the lesions in multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group. The method includes: extracting features from the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group to obtain one or more of the following multiple imaging parameters of the control group: the average effective atomic number value of the control group corresponding to the lesion, the average CT value of the control group corresponding to the lesion, the ratio of the long diameter to the short diameter of the lesion with the largest cross-sectional area, whether the boundary of the lesion in the control group is clear, whether the location of the lesion in the control group is adjacent to the bone cortex, and whether there is a spinous process sign at the edge of the lesion in the control group.

[0076] Similarly, in the embodiments of this disclosure and other possible embodiments, the multiple imaging parameters of the control group corresponding to the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group and the multiple imaging parameters of the study group corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group are, at least whether the boundary of the lesion in the control group is clear and / or whether the location of the lesion in the control group is adjacent to the bone cortex. The multiple imaging parameters of the control group corresponding to the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group and the multiple imaging parameters of the study group corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group also include: the average effective atomic number value of the control group corresponding to the lesion, the average CT value of the control group corresponding to the lesion, the ratio of the long axis to the short axis of the lesion with the largest cross-sectional area to the control group, and one or more of the multiple imaging parameters of the control group corresponding to the lesion edge.

[0077] In the embodiments disclosed herein and other possible embodiments, determining whether a spinous process sign exists at the edge of the lesion in the study group includes: calculating the ratio of the long axis to the short axis of each three-dimensional lesion in the first lesion mask image corresponding to each first lesion in the plurality of first spectral CT scan vertebral body images; if the ratio of the long axis to the short axis is greater than or equal to a first preset ratio of the long axis to the short axis, then it is determined that the corresponding lesion has a spinous process sign; otherwise, it is determined that the corresponding lesion does not have a spinous process sign.

[0078] In the embodiments of this disclosure and other possible embodiments, determining the major axis corresponding to each three-dimensional lesion includes: performing three-dimensional reconstruction on each lesion mask in the first lesion mask image to obtain the three-dimensional mask lesion corresponding to each lesion mask; performing edge detection on the three-dimensional mask lesion of each lesion mask to obtain the three-dimensional lesion mask edge; selecting any three-dimensional edge pixel in the three-dimensional lesion mask edge; calculating multiple three-dimensional edge pixel distances between the arbitrary three-dimensional edge pixel and other three-dimensional edge pixel; selecting the maximum three-dimensional edge pixel distance from the multiple three-dimensional edge pixel distances; and configuring the maximum three-dimensional edge pixel distance as the major axis corresponding to each three-dimensional lesion.

[0079] In the embodiments of this disclosure and other possible embodiments, determining the short path corresponding to each lesion includes: performing three-dimensional reconstruction on each lesion mask in the first lesion mask image to obtain a three-dimensional mask lesion corresponding to each lesion mask; performing edge detection on the three-dimensional mask lesions of each lesion mask to obtain the three-dimensional lesion mask edges; selecting any three-dimensional edge pixel in the three-dimensional lesion mask edges; calculating multiple three-dimensional edge pixel distances between the arbitrary three-dimensional edge pixel and other three-dimensional edge pixel points; selecting the minimum three-dimensional edge pixel distance from the multiple three-dimensional edge pixel distances; and configuring the minimum three-dimensional edge pixel distance as the long path corresponding to each three-dimensional lesion.

[0080] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the first preset major-to-minor diameter ratio according to actual needs. For example, the first preset major-to-minor diameter ratio can be configured as 1-50 or other possible values.

[0081] In the embodiments of this disclosure and other possible embodiments, since each first spectral CT scan vertebral body image is a three-dimensional first spectral CT scan vertebral body image, the first lesion mask image corresponding to each first spectral CT scan vertebral body image is also a three-dimensional first lesion mask image, and there may be multiple lesions on each dimension of the lesion mask image in the three-dimensional first lesion mask image.

[0082] In the embodiments disclosed herein and other possible embodiments, determining whether the lesion location in the study group is adjacent to the cortical bone based on the cortical bone edge mask position coordinates corresponding to each first-spectrum CT scan vertebral body image and the multiple first lesion edge mask position coordinates corresponding to each first-spectrum CT scan vertebral body image includes: calculating multiple position distances between the cortical bone edge mask position coordinates corresponding to each first-spectrum CT scan vertebral body image and the multiple first lesion edge mask position coordinates; if the minimum position distance among the multiple position distances is less than or equal to a preset position distance, then the lesion location in the first-spectrum CT scan vertebral body image corresponding to the study group is determined to be adjacent to the cortical bone; otherwise, the lesion location in the first-spectrum CT scan vertebral body image corresponding to the study group is determined to be far from the cortical bone. Wherein, the multiple first lesion edge mask position coordinates are configured as the position coordinates of the lesion edge mask in the multiple first lesion edge mask images.

[0083] In the embodiments disclosed herein and other possible embodiments, the step of detecting the cortical bone edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images includes: using a preset cortical bone segmentation model for cortical bone scan vertebral body images, performing cortical bone segmentation on the cortical bone in the plurality of first-spectrum CT scan vertebral body images corresponding to the osteogenic bone metastasis research group to obtain a plurality of corresponding first cortical bone mask images; performing edge detection on the plurality of first cortical bone mask images to obtain a plurality of corresponding first cortical bone edge mask images; and extracting the cortical bone edge coordinates in the plurality of first cortical bone edge mask images to obtain the cortical bone edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images.

[0084] In the embodiments disclosed herein and other possible embodiments, before the bone cortex in the multiple first spectral CT scan vertebral images corresponding to the osteogenic bone metastasis research group is segmented using a preset energy spectrum CT scan vertebral cortex segmentation model to obtain multiple corresponding first bone cortex mask images, the method includes: training a second preset deep learning segmentation network using a preset number of energy spectrum CT scan vertebral training images and their corresponding energy spectrum CT scan bone cortex mask training images to obtain a preset energy spectrum CT scan vertebral image bone cortex segmentation model.

[0085] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each cortical region in the plurality of first cortical bone mask images is configured as 1, and the mask value corresponding to the non-cortical bone regions is configured as 0.

[0086] In the embodiments of this disclosure and other possible embodiments, before training the second preset deep learning segmentation network using a preset number of spectral CT scan vertebral body training images and their corresponding spectral CT scan cortical bone mask training images to obtain a preset spectral CT scan vertebral body image cortical bone segmentation model, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate the cortical bone of the preset number of spectral CT scan vertebral body training images to obtain corresponding spectral CT scan cortical bone mask training images.

[0087] In the embodiments disclosed herein and other possible embodiments, the second preset deep learning segmentation network is configured as one or more deep learning segmentation networks such as Fully Convolutional Network (FCN), U-Net, DeepLab, and Pyramid Spatial Pooling Network (PSPNet).

[0088] Specifically, in the embodiments of this disclosure and other possible embodiments, the step of extracting features from the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group to obtain one or more of the following imaging parameters for the control group: the average effective atomic number value of the control group corresponding to the lesion, the average CT value of the control group corresponding to the lesion, the ratio of the long axis to the short axis of the lesion with the largest cross-sectional area to the control group, whether the boundary of the lesion in the control group is clear, whether the location of the lesion in the control group is adjacent to the bone cortex, and whether there is a spinous process sign at the edge of the lesion in the control group, includes: using a second preset spectral CT scan vertebral image lesion segmentation model to segment the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group to obtain multiple corresponding second lesion mask images; and extracting one or more of the following imaging parameters for the control group based on the multiple second-spectrum CT scan vertebral images and the multiple second lesion mask images: the average effective atomic number value of the control group corresponding to the lesion, the average CT value of the control group corresponding to the lesion, the ratio of the long axis to the short axis of the lesion with the largest cross-sectional area to the control group, whether the boundary of the lesion in the control group is clear, whether the location of the lesion in the control group is adjacent to the bone cortex, and whether there is a spinous process sign at the edge of the lesion in the control group.

[0089] In the embodiments of this disclosure and other possible embodiments, before segmenting lesions in multiple second-spectrum CT scan vertebral images corresponding to the bone island control group using the second preset energy spectrum CT scan lesion segmentation model, the method includes: training a third preset deep learning segmentation network using a preset number of energy spectrum CT scan vertebral training images and their corresponding energy spectrum CT scan lesion mask training images to obtain the second preset energy spectrum CT scan vertebral image lesion segmentation model. Each second lesion mask image and each energy spectrum CT scan lesion mask training image may contain multiple lesion masks.

[0090] In the embodiments disclosed herein and other possible embodiments, the lesions in the plurality of second-spectrum CT scan vertebral images are configured as bone dysplasia or hamartomas (non-true tumors); the mask value corresponding to each lesion region in the plurality of second lesion mask images is configured as 1, and the mask value corresponding to the non-lesion region is configured as 0.

[0091] In the embodiments of this disclosure and other possible embodiments, before training the first preset deep learning segmentation network using a preset number of spectral CT scan vertebral body training images and their corresponding spectral CT scan lesion mask training images to obtain the second preset spectral CT scan vertebral body image lesion segmentation model, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate lesions in the preset number of spectral CT scan vertebral body training images to obtain the corresponding spectral CT scan lesion mask training images.

[0092] In the embodiments disclosed herein and other possible embodiments, the third preset deep learning segmentation network is configured as one or more deep learning segmentation networks such as Fully Convolutional Network (FCN), U-Net, DeepLab, and Pyramid Spatial Pooling Network (PSPNet).

[0093] In the embodiments of this disclosure and other possible embodiments, the average effective atomic number value of the control group corresponding to the lesion is extracted based on the plurality of second spectral CT scan vertebral body images and the plurality of second lesion mask images. This includes: using a spectral CT atomic number analysis module configured with spectral CT images, the average effective atomic number value of the control group corresponding to the lesion is extracted based on the plurality of second spectral CT scan vertebral body images and the plurality of second lesion mask images.

[0094] In the embodiments of this disclosure and other possible embodiments, the spectral CT atomic number analysis module configured using spectral CT images extracts the average effective atomic number value of the control group corresponding to the lesion based on the plurality of second spectral CT scan vertebral images and the plurality of second lesion mask images. This includes: using the spectral CT atomic number analysis module configured using spectral CT images, extracting a quantitative bar chart of the atomic number corresponding to each component in each lesion generated by each second spectral CT scan vertebral image and its corresponding second lesion mask image; determining the effective atomic number value of each component in each lesion based on multiple peak values ​​corresponding to the quantitative bar chart of the atomic number corresponding to each component in each lesion; and averaging the effective atomic number values ​​of all components in all lesions to determine the average effective atomic number value of the control group for all components in the lesions corresponding to the plurality of second spectral CT scan vertebral images.

[0095] In the embodiments of this disclosure and other possible embodiments, the step of extracting a quantitative bar chart of the atomic number corresponding to each component within each lesion generated from each of the plurality of second-spectrum CT scan vertebral body images and their corresponding second lesion mask images includes: performing pixel-level multiplication operations on each of the plurality of second-spectrum CT scan vertebral body images and their corresponding second lesion mask images to obtain a plurality of second-spectrum CT scan lesion images corresponding to each of the plurality of second-spectrum CT scan vertebral body images; and obtaining a quantitative bar chart of the atomic number corresponding to each component within each lesion generated from each of the plurality of second-spectrum CT scan lesion images based on the plurality of second-spectrum CT scan lesion images.

[0096] In the embodiments of this disclosure and other possible embodiments, based on the plurality of second-spectrum CT scan vertebral body images and the plurality of second lesion mask images, the average CT value of the control group corresponding to the lesion is extracted, including: performing pixel-level multiplication operations on each second-spectrum CT scan vertebral body image and its corresponding second lesion mask image in the plurality of second-spectrum CT scan vertebral body images to obtain a plurality of second-spectrum CT scan lesion images corresponding to the plurality of second-spectrum CT scan vertebral body images; calculating the CT value of each lesion corresponding to the plurality of second-spectrum CT scan lesion images; and averaging the CT values ​​of each lesion to determine the average CT value of the control group corresponding to the lesion.

[0097] In the embodiments of this disclosure and other possible embodiments, based on the plurality of second-spectrum CT scan vertebral body images and the plurality of second lesion mask images, the control group ratio of the major and minor axes of the lesion with the largest cross-sectional area is extracted, including: calculating the lesion mask cross-sectional area of ​​each lesion in the second lesion mask image corresponding to each of the plurality of second-spectrum CT scan vertebral body images; calculating the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion; and determining the control group ratio corresponding to each second-spectrum CT scan vertebral body image based on the major and minor axes of the lesion with the largest cross-sectional area.

[0098] In the embodiments of this disclosure and other possible embodiments, the step of calculating the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion includes: performing edge detection on the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion to obtain the edge of the lesion with the largest cross-sectional area; selecting any edge pixel point in the edge of the lesion with the largest cross-sectional area; calculating multiple edge pixel point distances between the arbitrary edge pixel point and other edge pixel points in the edge of the lesion with the largest cross-sectional area; selecting the maximum edge pixel point distance and the minimum edge pixel point distance from the multiple edge pixel point distances; and configuring the maximum edge pixel point distance and the minimum edge pixel point distance as the major and minor axes of the lesion with the largest cross-sectional area, respectively.

[0099] In the embodiments of this disclosure and other possible embodiments, determining whether the lesion boundary of the control group is clear based on the plurality of second-spectrum CT scan vertebral body images and the plurality of second lesion mask images includes: performing an erosion operation on the lesion mask in the second lesion mask image corresponding to each of the plurality of second-spectrum CT scan vertebral body images to obtain a plurality of second lesion mask erosion images; subtracting the corresponding plurality of second lesion mask erosion images from the plurality of second lesion mask images to obtain a plurality of second lesion mask processed images; mapping the coordinates of multiple lesions in the plurality of second lesion mask processed images to the corresponding second-spectrum CT scan vertebral body images; calculating a plurality of second gradients corresponding to the coordinates of multiple lesions in the second-spectrum CT scan vertebral body images; if the second gradient magnitude value corresponding to the plurality of second gradients is greater than or equal to a second preset gradient magnitude value, then the lesion boundary where the lesion location is greater than or equal to the second preset gradient magnitude value is determined to be clear; otherwise, the lesion boundary is determined to be blurred.

[0100] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the second preset gradient magnitude value according to actual needs. For example, the second preset gradient magnitude value can be configured to 1-80 or other possible values.

[0101] In the embodiments of this disclosure and other possible embodiments, after subtracting the corresponding eroded images of the plurality of second lesion masks from the plurality of second lesion mask images to obtain a plurality of processed images of second lesion masks, the plurality of processed images of second lesion masks are configured as a plurality of edge mask images of second lesions.

[0102] In the embodiments of this disclosure and other possible embodiments, the step of calculating multiple gradients corresponding to multiple lesion location coordinates in the second spectral CT scan vertebral body image includes: determining multiple first position coordinates and multiple second position coordinates on both sides of the multiple lesion location coordinates in the second spectral CT scan vertebral body image in the x-direction, and multiple third position coordinates and multiple fourth position coordinates in the y-direction, based on the multiple lesion location coordinates; calculating the difference between the fifth pixel value corresponding to the multiple first position coordinates and the sixth pixel value corresponding to the multiple second position coordinates to obtain multiple second x-direction gradients corresponding to the multiple lesion location coordinates; and calculating the difference between the seventh pixel value corresponding to the multiple third position coordinates and the eighth pixel value corresponding to the multiple fourth position coordinates to obtain multiple second y-direction gradients corresponding to the multiple lesion location coordinates.

[0103] In the embodiments of this disclosure and other possible embodiments, determining the gradient magnitude corresponding to the plurality of gradients includes: squaring the plurality of second x-direction gradients corresponding to the plurality of lesion location coordinates to obtain a plurality of second x-direction squared gradients; squaring the plurality of second y-direction gradients corresponding to the plurality of lesion location coordinates to obtain a plurality of second y-direction squared gradients; summing the plurality of second x-direction squared gradients with their corresponding second y-direction squared gradients to obtain a sum of second gradients corresponding to the plurality of lesion location coordinates; and taking the square root of the sum of second gradients corresponding to the plurality of lesion location coordinates to obtain the second gradient magnitude corresponding to the plurality of gradients.

[0104] In the embodiments disclosed herein and other possible embodiments, multiple lesion location coordinates (x1, y1), (x2, y2), ..., (x n ,y n ), determine the multiple fifth position coordinates (x1+1, y1), (x2+1, y2), ..., (x) on both sides of the multiple lesion location coordinates in the second energy spectrum CT scan vertebral body image in the x direction. n +1,y n ) and multiple sixth position coordinates (x1-1, y1), (x2-1, y2), ..., (x n -1,y n ) and the corresponding seventh position coordinates in the y direction (x1, y1+1), (x2, y2+1), ..., (x n ,y n +1) and multiple eighth position coordinates (x1, y1-1), (x2, y2-1), ..., (x n ,y n-1); calculate the differences between the fifth pixel values ​​p5 corresponding to multiple fifth position coordinates and the sixth pixel values ​​p6 corresponding to multiple sixth position coordinates to obtain multiple second x-direction gradients corresponding to the multiple lesion position coordinates; calculate the differences between the seventh pixel values ​​p3 corresponding to multiple seventh position coordinates and the eighth pixel values ​​p4 corresponding to multiple eighth position coordinates to obtain the multiple lesion position coordinates (x1, y1), (x2, y2), ..., (x n ,y n The corresponding gradients in the second y-direction.

[0105] In the embodiments disclosed herein and other possible embodiments, the step of performing an erosion operation on the lesion masks in the plurality of second lesion mask images to obtain a plurality of second lesion mask eroded images includes: using a set erosion pixel to perform an erosion operation on the lesion masks in the plurality of second lesion mask images to obtain a plurality of second lesion mask eroded images.

[0106] In the embodiments disclosed herein and other possible embodiments, determining whether the lesion location in the control group is adjacent to the bone cortex includes: detecting the bone cortex edge mask position coordinates corresponding to each of the plurality of second-spectrum CT scan vertebral body images; and determining whether the lesion location in the control group is adjacent to the bone cortex based on the bone cortex edge mask position coordinates corresponding to each of the plurality of second-spectrum CT scan vertebral body images and the plurality of second lesion edge mask position coordinates corresponding to each of the plurality of second-spectrum CT scan vertebral body images.

[0107] The presence of a spiky sign at the lesion's margin is of significant reference value in suggesting the possibility of malignancy, assisting in differential diagnosis, and guiding further examination and treatment in clinical diagnosis. First, the spiky sign is a common imaging feature of malignant lung tumors (especially peripheral lung cancer). Studies show that the incidence of the spiky sign is approximately 66%-72% in peripheral lung cancer, but only about 20% in benign pulmonary nodules. Its formation is related to factors such as invasive tumor growth, uneven vascular distribution, and obstruction by surrounding tissues, suggesting that tumor cells may have breached local tissue boundaries and have a high malignant potential. Second, in the differential diagnosis of solitary pulmonary nodules, the spiky sign helps distinguish between benign and malignant lesions. If a spiky sign is present at the nodule's margin, combined with other signs (such as lobulation, pleural indentation, and vascular convergence), it can increase the suspicion of malignancy; conversely, if the nodule margin is smooth, without a spiky sign, and without other malignant signs, the possibility of a benign lesion is relatively high. Finally, upon discovering the spinous process sign, clinicians usually recommend further examinations, such as enhanced CT, PET-CT, or biopsy, to determine the nature of the lesion. If a malignant tumor is diagnosed, the presence of the spinous process sign may indicate that the tumor has progressed, requiring early development of a comprehensive treatment plan including surgery, radiotherapy, chemotherapy, or targeted therapy. This paper proposes a technique to determine the presence of a spinous process sign at the lesion's edge, distinct from existing technologies.

[0108] In the embodiments of this disclosure and other possible embodiments, determining whether the control group lesion edge has a spinous process sign includes: calculating multiple long-to-short diameter ratios corresponding to each three-dimensional lesion in the second lesion mask image corresponding to each second spectral CT scan vertebral body image in the multiple second spectral CT scan vertebral body images; if any one of the multiple long-to-short diameter ratios is greater than or equal to a second preset long-to-short diameter ratio, then it is determined that the corresponding lesion has a spinous process sign; otherwise, it is determined that the corresponding lesion does not have a spinous process sign.

[0109] In the embodiments of this disclosure and other possible embodiments, determining the major axis corresponding to each three-dimensional lesion includes: performing three-dimensional reconstruction on each lesion mask in the second lesion mask image to obtain the three-dimensional mask lesion corresponding to each lesion mask; performing edge detection on the three-dimensional mask lesion of each lesion mask to obtain the three-dimensional lesion mask edge; selecting any three-dimensional edge pixel in the three-dimensional lesion mask edge; calculating multiple three-dimensional edge pixel distances between the arbitrary three-dimensional edge pixel and other three-dimensional edge pixel; selecting the maximum three-dimensional edge pixel distance from the multiple three-dimensional edge pixel distances; and configuring the maximum three-dimensional edge pixel distance as the major axis corresponding to each three-dimensional lesion.

[0110] In the embodiments of this disclosure and other possible embodiments, determining the short path corresponding to each lesion includes: performing three-dimensional reconstruction on each lesion mask in the second lesion mask image to obtain a three-dimensional mask lesion corresponding to each lesion mask; performing edge detection on the three-dimensional mask lesion of each lesion mask to obtain the three-dimensional lesion mask edge; selecting any three-dimensional edge pixel in the three-dimensional lesion mask edge; calculating multiple three-dimensional edge pixel distances between the arbitrary three-dimensional edge pixel and other three-dimensional edge pixel; selecting the minimum three-dimensional edge pixel distance from the multiple three-dimensional edge pixel distances; and configuring the minimum three-dimensional edge pixel distance as the long path corresponding to each three-dimensional lesion.

[0111] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the second preset major-to-minor diameter ratio according to actual needs. For example, the second preset major-to-minor diameter ratio can be configured to 1-50 or other possible values.

[0112] In the embodiments disclosed herein and other possible embodiments, since each second-spectrum CT scan vertebral body image is a three-dimensional second-spectrum CT scan vertebral body image, the second lesion mask image corresponding to each second-spectrum CT scan vertebral body image is also a three-dimensional second lesion mask image, and there may be multiple lesions on each dimension of the lesion mask image in the three-dimensional second lesion mask image.

[0113] In the embodiments of this disclosure and other possible embodiments, determining whether the lesion location of the control group is adjacent to the cortical bone based on the cortical bone edge mask position coordinates corresponding to each second-spectrum CT scan vertebral body image and the multiple second lesion edge mask position coordinates corresponding to each second-spectrum CT scan vertebral body image includes: calculating the cortical bone edge mask position coordinates corresponding to each second-spectrum CT scan vertebral body image and multiple position distances corresponding to each of the multiple second lesion edge mask position coordinates; if the minimum position distance among the multiple position distances is less than or equal to a preset position distance, then the lesion location in the second-spectrum CT scan vertebral body image corresponding to the control group is determined to be adjacent to the cortical bone; otherwise, the lesion location in the second-spectrum CT scan vertebral body image corresponding to the control group is determined to be far from the cortical bone. The multiple second lesion edge mask position coordinates are configured as the position coordinates of the lesion edge mask in the multiple second lesion edge mask images.

[0114] In the embodiments disclosed herein and other possible embodiments, the step of detecting the cortical bone edge mask position coordinates corresponding to each of the plurality of second-spectrum CT scan vertebral body images includes: using a preset cortical bone segmentation model for spectral CT scan vertebral body images, performing cortical bone segmentation on the cortical bone in the plurality of second-spectrum CT scan vertebral body images corresponding to the bone island control group to obtain a plurality of corresponding second cortical bone mask images; performing edge detection on the plurality of second cortical bone mask images to obtain a plurality of corresponding second cortical bone edge mask images; and extracting the cortical bone edge coordinates in the plurality of second cortical bone edge mask images to obtain the cortical bone edge mask position coordinates corresponding to each of the plurality of second-spectrum CT scan vertebral body images.

[0115] In the embodiments of this disclosure and other possible embodiments, before the bone cortex in the multiple second spectral CT scan vertebral images corresponding to the bone island control group is segmented using a preset energy spectrum CT scan vertebral cortex segmentation model to obtain the corresponding multiple second bone cortex mask images, the method includes: training a fourth preset deep learning segmentation network using a preset number of energy spectrum CT scan vertebral training images and their corresponding energy spectrum CT scan bone cortex mask training images to obtain a preset energy spectrum CT scan vertebral image bone cortex segmentation model.

[0116] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each cortical region in the plurality of second cortical bone mask images is configured as 1, and the mask value corresponding to the non-cortical bone regions is configured as 0.

[0117] In the embodiments of this disclosure and other possible embodiments, before training the fourth preset deep learning segmentation network using a preset number of spectral CT scan vertebral body training images and their corresponding spectral CT scan cortical bone mask training images to obtain a preset spectral CT scan vertebral body image cortical bone segmentation model, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate the cortical bone of the preset number of spectral CT scan vertebral body training images to obtain corresponding spectral CT scan cortical bone mask training images.

[0118] In the embodiments disclosed herein and other possible embodiments, the fourth preset deep learning segmentation network is configured as one or more deep learning segmentation networks such as Fully Convolutional Network (FCN), U-Net, DeepLab, and Pyramid Spatial Pooling Network (PSPNet).

[0119] In this embodiment of the disclosure, determining the corresponding quantitative features and count features based on multiple imaging parameters of the study group and the control group includes: grouping the multiple imaging parameters of the study group and the control group according to quantitative and count principles to obtain the quantitative features and count features to be processed; performing an independent samples t-test on the quantitative features to be processed corresponding to the study group and the control group, and selecting quantitative features with a significance level less than or equal to a first preset significance level from the quantitative features to be processed; and performing a chi-square test on the count features to be processed corresponding to the study group and the control group, and selecting count features with a significance level less than or equal to a second preset significance level from the count features to be processed. The feature selection algorithm is configured as a multi-factor logistic regression analysis algorithm.

[0120] Specifically, SPSS 22.0 software was used for statistical analysis. Independent samples t-tests were used to compare the mean effective atomic number, mean CT value, and the ratio of the long to short diameter of the largest cross-sectional area lesion between the study group and the control group. Quantitative features with a significance level less than or equal to the first pre-specified significance (e.g., 0.05) were selected from these metrics. Similarly, chi-square tests were used to evaluate lesion boundary clarity / blurred boundary, lesion location, and presence of a spinous process in both the study group and the control group. Count features with a significance level less than or equal to the second pre-specified significance (e.g., 0.05) were selected from these metrics. Multivariate logistic regression analysis was used to screen risk factors constructed from the quantitative and count features to obtain relevant variables for differentiating osteoblastic bone metastases from bone islands. Then, receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC) were used for evaluation.

[0121] More specifically, the area under the first ROC curve for osteoblastic bone metastases and the area under the second ROC curve for bone islands can demonstrate the diagnostic efficacy in differentiating osteoblastic bone metastases from bone islands. Furthermore, quantitative atomic number bar charts and pseudo-color atomic number plots are helpful in observing the difference in effective atomic number values ​​between osteoblastic bone metastases and bone islands. Combined with histopathological analysis, bone islands are small pieces of mature cortical bone within cancellous bone, composed of irregular lamellar bone with no obvious interstitial components; their effective atomic number values ​​are close to those of the bone cortex. In contrast, osteoblastic metastases are characterized by extensive pathological osteoblasts; these newly formed bones are woven in appearance, with incomplete local calcium salt deposition, lacking the normal cortical bone structure; their effective atomic number values ​​are lower than those of bone islands.

[0122] In this embodiment of the disclosure, determining the preset osteogenic bone metastasis radiotherapy efficacy grading prediction model includes: obtaining a preset number of relevant variables for distinguishing osteogenic bone metastasis from bone islands at a first time point and osteogenic bone metastasis radiotherapy efficacy grading labels at a second time point after the first time point; training a preset prediction network or preset predictor to obtain the preset osteogenic bone metastasis radiotherapy efficacy grading prediction model.

[0123] In the embodiments disclosed herein and other possible embodiments, the preset prediction network or preset predictor may be configured as one or more of the following: decision tree classifier, random forest classifier, fully connected neural network (FCN), recurrent neural network (RNN), support vector machine (SVM), K-nearest neighbor algorithm (KNN).

[0124] In the embodiments disclosed herein and other possible embodiments, if the first time point is configured as before radiotherapy, the second time point can be configured as 1 month, 2 months, 3 months after radiotherapy, or any other time point after radiotherapy. If the first time point is configured as 1 month after radiotherapy, the second time point can be configured as 2 months, 3 months, or any other time point after 1 month after radiotherapy.

[0125] In the embodiments disclosed herein and other possible embodiments, when quantifying the tumor activity status corresponding to bone metastases, it is necessary to determine whether the bone metastases are in a radiotherapy efficacy grading system for osteoblastic bone metastases, categorized as complete remission (CR), partial remission (PR), stable disease (SD), or progressive disease (PD). This directly determines the subsequent treatment plan for patients with bone metastases and has a profound impact on their prognosis. The radiotherapy effect on osteolytic and mixed metastatic bone metastases can be assessed by evaluating the tumor activity status through changes in the size of the tumor corresponding to the bone metastases; however, this method is not applicable to osteoblastic bone metastases. This is because, after radiotherapy, although the tumor tissue corresponding to the lesion has undergone necrosis, changes in the size and density of the osteoblastic bone metastases are often not significant due to ossification or bone hyperplasia. Therefore, multiple analyses before and after radiotherapy are required using the atomic number analysis module of spectral CT to accurately predict the radiotherapy efficacy grading of osteoblastic bone metastases.

[0126] For example, cases of osteoblastic bone metastasis are selected as research subjects, and spectral CT imaging of the vertebral body is performed before radiotherapy to obtain spectral CT scan images of the vertebral body to be predicted before radiotherapy. Based on the relevant variables of the spectral CT scan images of the vertebral body to be predicted for distinguishing between osteoblastic bone metastasis and bone islands, a pre-set osteoblastic bone metastasis radiotherapy efficacy grading prediction model is used to predict the osteoblastic bone metastasis radiotherapy efficacy grading at a set time (1 month or 3 months after radiotherapy) after at least one radiotherapy stage.

[0127] In the embodiments disclosed herein and other possible embodiments, the radiotherapy efficacy grading for osteoblastic bone metastases is configured as one or more of the following: complete remission (CR), partial remission (PR), stable disease (SD), and progressive disease (PD). Complete remission (CR) is characterized by X-ray, CT, or MRI showing complete sclerosis of the osteoblastic lesion, normalized bone density, normalized bone signal intensity, and normalized bone scintigraphy tracer uptake; clinically, it is characterized by complete disappearance of pain, no need for analgesics, and basic functional recovery. Partial remission (PR) is characterized by X-ray, CT, or MRI showing sclerosis or partial sclerosis filling at the margins of the osteoblastic lesion, or a measurable lesion shrinkage of ≥50%, or a subjective shrinkage of ≥50% in unmeasurable lesions; a subjective reduction of ≥50% in bone scintigraphy tracer uptake; clinically, it is characterized by significant pain reduction, reduced analgesic use, and partial functional recovery. Stable disease (SD) is characterized by an increase of <25% or decrease of <50% in measurable lesion size on imaging, an increase of ≤25% and a decrease of ≤50% in unmeasurable lesion size, and no new bone metastases. Clinically, pain is not significantly changed or slightly improved, and function is basically stable. Progressive disease (PD) is characterized by an increase of ≥25% in measurable lesion size on X-ray, CT, or MRI, a subjective increase of ≥25% in unmeasurable lesion size, or an increase of ≥25% in bone scintigraphy tracer uptake. New bone metastases may also be present. Clinically, pain may worsen, function may deteriorate, or new bone-related events (such as pathological fractures or spinal cord compression) may occur.

[0128] In the embodiments disclosed herein and other possible embodiments, the method further includes: determining whether the location of the lesion corresponding to the lesion to be chemo or the lesion to be operated on is adjacent to an artery; if the location of the lesion corresponding to the lesion to be chemo or the lesion to be operated on is adjacent to an artery, then determining whether the artery passes through the lesion to be chemo or the lesion to be operated on based on the lesion mask area corresponding to the lesion to be chemo or the lesion to be operated on and the artery mask area corresponding to the artery.

[0129] In the embodiments disclosed herein and other possible embodiments, if the artery passes through the lesion to be treated with chemotherapy or the lesion to be operated on, radiotherapy is performed on the artery passing through the lesion under ultrasound image guidance at a first radiotherapy dose, and radiotherapy is performed on the lesion outside the artery under ultrasound image guidance at a second radiotherapy dose lower than the first radiotherapy dose. This addresses the issue that tumor cells corresponding to the lesion may be affected by blood flow in the blood vessels, and the concentration of chemotherapy drugs in the local tumor area may be easily diluted, leading to reduced efficacy and increased risk of residual cancer cells and recurrence.

[0130] In the embodiments disclosed herein and other possible embodiments, a vertebral body image scanned with a CT scan to be predicted is used to determine a lesion mask image and an arterial vessel mask image corresponding to the lesion to be treated or the lesion to be operated on; based on the lesion mask image and the arterial vessel mask image, it is determined whether the location of the lesion to be treated or the lesion to be operated on is adjacent to an artery; or, a vertebral body image scanned with a CT scan to be predicted is used to determine a lesion mask and an arterial vessel mask image corresponding to the lesion to be treated or the lesion to be operated on; based on the lesion mask and the arterial vessel mask image, it is determined whether the location of the lesion to be treated or the lesion to be operated on is adjacent to an arterial vessel.

[0131] In this disclosure and other possible embodiments, the step of determining the lesion mask image and arterial vessel mask image corresponding to the lesion to be treated or the lesion to be operated on using the vertebral body image of the CT scan to be predicted includes: training a first preset deep learning segmentation network using a preset number of vertebral body training images of the CT scan to be predicted and their corresponding lesion mask training images to obtain a preset lesion segmentation model; training a second preset deep learning segmentation network using a preset number of vertebral body training images of the CT scan to be predicted and their corresponding arterial vessel mask training images to obtain a preset arterial vessel segmentation model; and using the preset lesion segmentation model to segment the lesion to be treated or the lesion to be operated on in the vertebral body image of the CT scan to be predicted. The process involves segmenting the lesion to obtain a corresponding lesion mask image; segmenting the arteries using the preset arterial segmentation model to obtain a corresponding arterial mask image; or, determining the lesion mask and arterial mask images corresponding to the lesion to be treated or the lesion to be operated on using the vertebral body image to be predicted by CT scan, which includes: training a preset deep learning segmentation network using a preset number of vertebral body training images to be predicted by CT scan and their corresponding lesion mask and arterial mask training images to obtain a preset lesion and arterial segmentation model; segmenting the lesion to be treated or the lesion to be operated on and the arteries using the preset lesion and arterial segmentation model to obtain the corresponding lesion mask and arterial mask images.

[0132] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each lesion region in the lesion mask image and the lesion mask training image is configured as 1, and the mask value corresponding to the non-lesion region is configured as 0; similarly, the mask value corresponding to the artery in the artery mask image and the artery mask training image is configured as 1, and the mask value corresponding to the non-artery region is configured as 0.

[0133] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each lesion region in the lesion mask and artery mask image and the lesion mask and artery mask training image is configured as 1, the mask value corresponding to the artery is configured as 2, and the mask value corresponding to the non-lesion region and non-artery region is configured as 0.

[0134] In the embodiments of this disclosure and other possible embodiments, before training the first preset deep learning segmentation network using a preset number of CT scan vertebral body training images of the energy spectrum to be predicted and their corresponding lesion mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate lesions in the preset number of CT scan vertebral body training images of the energy spectrum to be predicted, thereby obtaining corresponding lesion mask training images.

[0135] In the embodiments of this disclosure and other possible embodiments, before training the second preset deep learning segmentation network using a preset number of CT scan vertebral body training images of the energy spectrum to be predicted and their corresponding arterial blood vessel mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate the arterial blood vessels in the preset number of CT scan vertebral body training images of the energy spectrum to be predicted, thereby obtaining the corresponding arterial blood vessel mask training images.

[0136] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network using a preset number of CT scan vertebral body training images to be predicted and their corresponding lesion masks and arterial vessel mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate lesions and arterial vessels in the preset number of CT scan vertebral body training images to be predicted, thereby obtaining corresponding lesion masks and arterial vessel mask training images.

[0137] In the embodiments disclosed herein and other possible embodiments, the preset deep learning segmentation network, the first preset deep learning segmentation network and the second preset deep learning segmentation network are configured as one or more deep learning segmentation networks such as Fully Convolutional Network (FCN), U-Net, DeepLab, Pyramid Spatial Pooling Network (PSPNet).

[0138] In the embodiments disclosed herein and other possible embodiments, determining whether the location of the lesion to be treated or operated on is adjacent to an artery based on the lesion mask image and the artery mask image includes: calculating multiple positional distances between the artery edge mask position coordinates corresponding to the artery mask image and the lesion edge mask position coordinates corresponding to the lesion mask image; if the minimum positional distance among the multiple positional distances is less than or equal to a preset positional distance, then the lesion location is determined to be adjacent to an artery; otherwise, the lesion location is determined to be far from an artery. Wherein, the position coordinates of the arterial vessel edge mask and the position coordinates of the lesion edge mask corresponding to the lesion are respectively configured as the position coordinates of the arterial vessel edge in the arterial vessel edge mask image corresponding to the arterial vessel mask image and the position coordinates of the lesion edge mask in the lesion edge mask image corresponding to the lesion mask image; or, the step of determining whether the lesion location corresponding to the lesion to be treated or the lesion to be operated on is close to an arterial vessel based on the lesion mask and the arterial vessel mask image includes: calculating multiple position distances corresponding to the position coordinates of the arterial vessel edge mask and the position coordinates of the lesion edge mask, respectively; if the minimum position distance among the multiple position distances is less than or equal to a preset position distance, then the lesion location is determined to be close to an arterial vessel; otherwise, the lesion location is determined to be far from an arterial vessel. Wherein, the position coordinates of the arterial vessel edge mask and the position coordinates of the lesion edge mask corresponding to the lesion are respectively configured as the position coordinates of the arterial vessel edge in the arterial vessel edge mask image and the position coordinates of the lesion edge mask in the lesion edge mask image.

[0139] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the preset position distance according to actual needs. For example, the preset position distance can be any value between 0 and 10 pixels or other values.

[0140] In the embodiments of this disclosure and other possible embodiments, before determining the lesion mask image and arterial mask image corresponding to the lesion to be treated or the lesion to be operated on using the vertebral body image of the CT scan to be predicted, the method includes: detecting whether the lesion boundary in the lesion mask image corresponding to the vertebral body image of the CT scan to be predicted is blurred and whether there is a spinous process sign at the edge of the lesion; if the lesion boundary in the lesion mask image is blurred and there is a spinous process sign at the edge of the lesion, then the lesion is configured as a lesion to be treated or a lesion to be operated on; or, before determining the lesion mask and arterial mask image corresponding to the lesion to be treated or the lesion to be operated on using the vertebral body image of the CT scan to be predicted, the method includes: detecting whether the lesion boundary in the lesion mask image corresponding to the vertebral body image of the CT scan to be predicted is blurred and whether there is a spinous process sign at the edge of the lesion; if the lesion boundary in the lesion mask image is blurred and there is a spinous process sign at the edge of the lesion, then the lesion is configured as a lesion to be treated or a lesion to be operated on.

[0141] The execution entity for the method of predicting the grading of radiotherapy efficacy for osteogenic bone metastases can be an image processing device or system. For example, the method can be executed by a terminal device, server, or other processing device. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, vehicle-mounted device, wearable device, etc. In some possible implementations, the method can be implemented by a processor calling computer-readable instructions stored in memory.

[0142] Those skilled in the art will understand that, in the above-described method for predicting the efficacy of radiotherapy for osteogenic bone metastases in specific embodiments, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined based on its function and possible internal logic.

[0143] Figure 4 This is a block diagram illustrating a radiotherapy efficacy grading prediction system for osteogenic bone metastases according to an exemplary embodiment. Figure 4 As shown, the osteogenic bone metastasis radiotherapy efficacy grading prediction system includes: an acquisition unit 10, used to acquire vertebral body images of the target energy spectrum CT scan corresponding to at least one radiotherapy stage before and after radiotherapy for osteogenic bone metastasis; and a prediction unit 11, used to predict the osteogenic bone metastasis radiotherapy efficacy grading at a set radiotherapy time after at least one radiotherapy stage following radiotherapy, based on the relevant variables of the target energy spectrum CT scan vertebral body images used to distinguish between osteogenic bone metastasis and bone islands, using a preset osteogenic bone metastasis radiotherapy efficacy grading prediction model.

[0144] According to one aspect of this disclosure, a system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the above-described method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0145] According to one aspect of this disclosure, a system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases is provided, comprising: a computer-readable storage medium having a computer program / instructions and a bit stream stored thereon, wherein the computer program / instructions, when executed by a processor, generate the bit stream by implementing the above-described method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0146] According to one aspect of this disclosure, a system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases is provided, comprising: a computer program product configured with a computer program / instruction, which, when executed by a processor, implements the aforementioned method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases.

[0147] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for predicting the grading of radiotherapy efficacy in osteoblastic bone metastases, characterized in that, include: Obtain spectral CT images of the vertebral body corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases; Based on the relevant variables used to distinguish between osteogenic bone metastases and bone islands from the vertebral body images of the CT scan to be predicted, the osteogenic bone metastases radiotherapy efficacy grading prediction model is used to predict the osteogenic bone metastases radiotherapy efficacy grading at a set time after at least one radiotherapy stage following radiotherapy.

2. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to claim 1, characterized in that, The determination of the relevant variables used to differentiate osteogenic bone metastases from bone islands includes: extracting multiple imaging parameters corresponding to lesions in multiple first-spectrum CT vertebral images of the osteogenic bone metastasis study group and multiple imaging parameters corresponding to lesions in multiple second-spectrum CT vertebral images of the bone island control group; determining corresponding quantitative and count features based on the multiple imaging parameters of the study group and the multiple imaging parameters of the control group; and using a feature screening algorithm to screen the risk factors constructed from the quantitative and count features to obtain the relevant variables used to differentiate osteogenic bone metastases from bone islands.

3. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to claim 2, characterized in that, The extraction of multiple imaging parameters corresponding to the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group includes: extracting features from the lesions in the multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis research group to obtain one or more of the following multiple imaging parameters: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the long diameter to the short diameter of the research group of the lesion with the largest cross-sectional area, whether the boundary of the research group lesion is clear, whether the location of the research group lesion is close to the bone cortex, and whether there is a spinous process sign at the edge of the research group lesion.

4. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to claim 3, characterized in that, The process involves extracting features from multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group, obtaining one or more of the following imaging parameters: the average effective atomic number value of the study group corresponding to the lesion, the average CT value of the study group corresponding to the lesion, the ratio of the long axis to the short axis of the lesion with the largest cross-sectional area, whether the lesion boundary is clear, whether the lesion location is adjacent to the bone cortex, and whether there is a spinous process sign at the edge of the lesion. This includes using a first-preset first-spectrum CT scan vertebral image lesion segmentation model to segment the osteogenic bone metastasis... Lesions in multiple first-spectrum CT scan vertebral images corresponding to the research group were segmented to obtain multiple first-lesion mask images. Based on the multiple first-spectrum CT scan vertebral images and the multiple first-lesion mask images, one or more of the following imaging parameters were extracted: the average effective atomic number value of the research group corresponding to the lesion, the average CT value of the research group corresponding to the lesion, the ratio of the major axis to the minor axis of the lesion with the largest cross-sectional area, whether the boundary of the research group lesion is clear, whether the location of the research group lesion is close to the bone cortex, and whether there is a spinous process sign at the edge of the research group lesion.

5. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to any one of claims 3 or 4, characterized in that, Based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first-lesion mask images, the average effective atomic number value of the study group corresponding to the lesion is extracted, including: using a spectrum CT atomic number analysis module configured with spectrum CT images, extracting the average effective atomic number value of the study group corresponding to the lesion based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first-lesion mask images; and / or, Based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images, the average CT value of the study group corresponding to the lesion is extracted, including: performing pixel-level multiplication operations on each first-spectrum CT scan vertebral body image and its corresponding first lesion mask image in the plurality of first-spectrum CT scan vertebral body images to obtain a plurality of first-spectrum CT scan lesion images corresponding to the plurality of first-spectrum CT scan vertebral body images; calculating the CT value of each lesion corresponding to the plurality of first-spectrum CT scan lesion images; and averaging the CT values ​​of each lesion to determine the average CT value of the study group corresponding to the lesion.

6. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to any one of claims 3-5, characterized in that, Based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images, the study group ratio of the major and minor axes of the lesion with the largest cross-sectional area is extracted, including: calculating the lesion mask cross-sectional area of ​​each lesion in the first lesion mask image corresponding to each of the plurality of first-spectrum CT scan vertebral body images; calculating the major and minor axes of the lesion with the largest cross-sectional area among the lesion mask cross-sectional areas of each lesion; determining the study group ratio corresponding to each first-spectrum CT scan vertebral body image based on the major and minor axes of the lesion with the largest cross-sectional area; and / or, The calculation of the major and minor axes of the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion includes: performing edge detection on the lesion with the largest cross-sectional area in the lesion mask cross-sectional area of ​​each lesion to obtain the edge of the lesion with the largest cross-sectional area; selecting any edge pixel point in the edge of the lesion with the largest cross-sectional area; calculating multiple edge pixel point distances between the arbitrary edge pixel point and other edge pixel points in the edge of the lesion with the largest cross-sectional area; selecting the maximum edge pixel point distance and the minimum edge pixel point distance from the multiple edge pixel point distances; and configuring the maximum edge pixel point distance and the minimum edge pixel point distance as the major and minor axes of the lesion with the largest cross-sectional area, respectively.

7. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to any one of claims 3-6, characterized in that, Based on the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion mask images, determining whether the lesion boundaries in the study group are clear includes: performing an erosion operation on the lesion mask in the first lesion mask image corresponding to each of the plurality of first-spectrum CT scan vertebral body images to obtain a plurality of first lesion mask erosion images; subtracting the corresponding plurality of first lesion mask erosion images from the plurality of first lesion mask images to obtain a plurality of first lesion mask processed images; mapping the coordinates of multiple lesion locations in the plurality of first lesion mask processed images to the corresponding first-spectrum CT scan vertebral body images; calculating a plurality of first gradients corresponding to the coordinates of multiple lesion locations in the first-spectrum CT scan vertebral body images; if the first gradient magnitude value corresponding to the plurality of first gradients is greater than or equal to a first preset gradient magnitude value, then the lesion boundary where the lesion location is greater than or equal to the first preset gradient magnitude value is determined to be clear; otherwise, the lesion boundary is determined to be blurred.

8. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to any one of claims 3-7, characterized in that, Determining whether the lesion location in the study group is adjacent to the bone cortex includes: detecting the bone cortex edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images; determining whether the lesion location in the study group is adjacent to the bone cortex based on the bone cortex edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images and the plurality of first lesion edge mask position coordinates corresponding to each of the plurality of first-spectrum CT scan vertebral body images; and / or, The method for extracting multiple imaging parameters of the control group corresponding to the lesions in multiple second-spectrum CT scan vertebral images corresponding to the bone island control group is the same as the method for extracting multiple imaging parameters of the study group corresponding to the lesions in multiple first-spectrum CT scan vertebral images corresponding to the osteogenic bone metastasis study group. The method includes: extracting features from the lesions in the multiple second-spectrum CT scan vertebral images corresponding to the bone island control group to obtain one or more of the following multiple imaging parameters of the control group: the average effective atomic number value of the control group corresponding to the lesion, the average CT value of the control group corresponding to the lesion, the ratio of the long diameter to the short diameter of the lesion with the largest cross-sectional area, whether the boundary of the lesion in the control group is clear, whether the location of the lesion in the control group is close to the bone cortex, and whether there is a spinous process sign at the edge of the lesion in the control group.

9. The method for predicting the grading of radiotherapy efficacy for osteogenic bone metastases according to any one of claims 2-8, characterized in that, The step of determining corresponding quantitative and count features based on multiple imaging parameters of the study group and the control group includes: grouping the multiple imaging parameters of the study group and the control group according to quantitative and count principles to obtain quantitative and count features to be processed; performing independent samples t-tests on the quantitative features to be processed corresponding to the study group and the control group, and selecting quantitative features with significance less than or equal to a first preset significance from the quantitative features to be processed; performing chi-square tests on the count features to be processed corresponding to the study group and the control group, and selecting count features with significance less than or equal to a second preset significance from the count features to be processed; and / or, the feature selection algorithm is configured as a multi-factor logistic regression analysis algorithm; and / or Determining the preset osteogenic bone metastasis radiotherapy efficacy grading prediction model includes: obtaining a preset number of relevant variables for distinguishing osteogenic bone metastasis from bone islands at a first time point and osteogenic bone metastasis radiotherapy efficacy grading labels at a second time point after the first time point; training a preset prediction network or preset predictor to obtain the preset osteogenic bone metastasis radiotherapy efficacy grading prediction model.

10. A system for predicting the grading of radiotherapy efficacy for osteogenic bone metastases, characterized in that, include: The acquisition unit is used to acquire vertebral images of the spectral CT scan corresponding to at least one radiotherapy stage before and after radiotherapy for osteoblastic bone metastases. The prediction unit is used to predict the radiotherapy efficacy grade of osteogenic bone metastases at a set time after at least one radiotherapy stage, based on the relevant variables used to distinguish between osteogenic bone metastases and bone islands from the vertebral body image of the CT scan to be predicted, and using a preset radiotherapy efficacy grading prediction model for osteogenic bone metastases; or, The method comprises: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the method for predicting the efficacy grading of radiotherapy for osteogenic bone metastases according to any one of claims 1-9; or, a computer-readable storage medium having a computer program / instructions and a bit stream stored thereon, wherein the computer program / instructions, when executed by a processor, implement the method for predicting the efficacy grading of radiotherapy for osteogenic bone metastases according to any one of claims 1-9 to generate the bit stream; or, a computer program product having a computer program / instructions configured thereon, wherein the computer program / instructions, when executed by a processor, implement the method for predicting the efficacy grading of radiotherapy for osteogenic bone metastases according to any one of claims 1-9.

Images