Methods and systems for determining and intelligently analyzing pleural mask images

Abstract

This disclosure proposes a method and system for determining and intelligently analyzing pleural mask images, relating to the field of intelligent medical image analysis technology. The method for determining the pleural mask image includes: expanding the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain an expanded lung field mask image; and performing pixel-level subtraction operations on the expanded lung field mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura. Embodiments of this disclosure can achieve pleural mask image determination and intelligent medical image analysis.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent medical image analysis technology, and in particular to a method and system for determining and intelligently analyzing pleural mask images. Background Technology

[0002] The pleura is a thin, smooth serous membrane covering the surface of the lungs and the inner surface of the thoracic cavity. It consists of two layers: the visceral pleura and the parietal pleura. The visceral pleura adheres closely to the lung surface and is tightly integrated with the lungs; the parietal pleura covers the inner surface of the chest wall, the upper surface of the diaphragm, and the sides of the mediastinum. The two pleuras transition into each other at the root of the lung, forming a closed potential cavity, the pleural cavity, which contains a small amount of serous fluid to reduce friction during respiration.

[0003] The main functions of the pleura include: lubrication, maintaining negative pressure in the pleural cavity, and protection. Specifically, lubrication involves secreting serous fluid to allow the lungs to slide smoothly against the chest wall during respiration, reducing friction. Maintaining negative pressure in the pleural cavity helps with lung expansion and the return of venous blood and lymph. Protection involves acting as a physical barrier to protect the lungs and other organs within the pleural cavity from external damage and infection.

[0004] Pleural-related diseases include pleurisy, pleural effusion, and pleural adhesions, which may be caused by infection, trauma, tumors, etc. However, the pleura is not clearly displayed on chest medical images, making it difficult to delineate the pleura, thus failing to meet the technical requirements of a pre-defined pleural segmentation model built based on a deep learning segmentation network. Summary of the Invention

[0005] This disclosure presents a technical solution for a method and system for determining and intelligently analyzing pleural mask images.

[0006] According to one aspect of this disclosure, a method for determining a pleural mask image is provided, comprising: performing an expansion operation on a lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain a lung field expansion mask image; and performing a pixel-level subtraction operation on the lung field expansion mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura.

[0007] Preferably, determining the lung field mask image corresponding to the chest medical image includes: segmenting the lung field in the chest medical image to obtain the lung field mask image corresponding to the chest medical image.

[0008] Preferably, the step of segmenting the lung fields in the chest medical image to obtain a lung field mask image corresponding to the chest medical image includes: acquiring a preset lung field segmentation model constructed based on a deep learning segmentation network; and using the preset lung field segmentation model to segment the lung fields in the chest medical image to obtain a lung field mask image corresponding to the chest medical image.

[0009] Preferably, before segmenting the lung fields in the chest medical image using the preset lung field segmentation model, the method includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding lung field mask training images to obtain a preset lung field segmentation model constructed based on the deep learning segmentation network.

[0010] According to one aspect of this disclosure, a medical image intelligent analysis method is provided, comprising: determining a pleural mask image corresponding to the pleura in a chest medical image using the pleural mask image determination method described above; dividing the chest medical image into lung field regions using a lung field dynamic region division method to obtain a lung field region division network corresponding to each nodule in the chest medical image; and determining the nearest pleural region image for each nodule based on the lung field region division network corresponding to each nodule and the pleural mask image.

[0011] Preferably, determining each nodule in the chest medical image includes: segmenting the nodules in the chest medical image to determine each nodule in the chest medical image.

[0012] Preferably, the step of segmenting the nodules in the chest medical image and determining each nodule in the chest medical image includes: segmenting the nodules in the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network to obtain a nodule mask image corresponding to the chest medical image; and determining each nodule in the chest medical image based on the nodule mask image.

[0013] Preferably, the intelligent medical image analysis method further includes: using the nearest nodular pleural region image of each nodule to determine whether the chest medical image exhibits nodular pleural traction sign.

[0014] Preferably, determining whether the chest medical image contains a nodular pleural traction sign using the nearest nodular pleural region image for each nodule includes: using a pre-defined feature target detection model based on deep learning to perform feature target detection on the nodular pleural region image, and determining whether there are cord-like or linear features between the adjacent pleura of the nodule at the preset position in the nodular pleural region image; if so, determining that the chest medical image contains a nodular pleural traction sign; otherwise, determining that the chest medical image does not contain a nodular pleural traction sign.

[0015] Preferably, based on the quantitative imaging features of the selected preset nodules corresponding to the presence or absence of nodular pleural traction signs, preset chest diseases related to nodules are predicted.

[0016] Preferably, the step of predicting nodule-related preset chest diseases based on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of nodular pleural traction signs includes: performing feature selection on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of nodular pleural traction signs to obtain preset nodule quantitative imaging features corresponding to preset chest diseases; and predicting nodule-related preset chest diseases based on the preset nodule quantitative imaging features corresponding to preset chest diseases.

[0017] Preferably, the step of predicting the preset chest disease related to the preset nodule based on the preset quantitative imaging features of the preset nodule corresponding to the preset chest disease includes: training the preset predictor and constructing a preset chest disease model by using the preset quantitative imaging features of the preset nodule corresponding to the preset chest disease in the chest medical image at the first time and the preset chest disease results at the second time after the first time.

[0018] Preferably, the selected preset quantitative imaging features of the nodule further include one or more of the following: nodule long diameter, whether the nodule edge is clear, nodule morphology, nodule spinous process sign, nodule lobulation sign, and nodule cavitation sign.

[0019] Preferably, detecting the nodule spinous process sign includes: calculating the ratio of the major axis to the minor axis of each nodule based on the major axis and minor axis of each nodule in the nodule mask image corresponding to the chest medical image; wherein the chest medical image is configured as a two-dimensional medical image; if the ratio of the major axis to the minor axis is greater than or equal to a first preset ratio of the major axis to the minor axis, then it is determined that the corresponding nodule has a spinous process sign.

[0020] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: an expansion unit, configured to expand a lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain an expanded lung field mask image; and a subtraction unit, configured to perform pixel-level subtraction operations on the expanded lung field mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura.

[0021] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: 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 above-described pleural mask image determination method.

[0022] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: a computer-readable storage medium storing a computer program / instructions and a bit stream thereon, wherein the computer program / instructions, when executed by a processor, implement the above-described pleural mask image determination method to generate the bit stream.

[0023] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: a computer program product configured with a computer program / instruction that, when executed by a processor, implements the aforementioned pleural mask image determination method.

[0024] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: an expansion unit, configured to expand the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain an expanded lung field mask image; a subtraction unit, configured to perform pixel-level subtraction operations on the expanded lung field mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura; a segmentation unit, configured to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image; and a first determination unit, configured to determine the nearest pleural region image to each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image.

[0025] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: an expansion unit, configured to expand the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain a lung field expansion mask image; a subtraction unit, configured to perform pixel-level subtraction operations on the lung field expansion mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura; a segmentation unit, configured to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image; a first determination unit, configured to determine the nearest pleural region image of each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image; and a second determination unit, configured to determine whether the chest medical image contains a nodular pleural traction sign using the nearest nodular pleural region image of each nodule.

[0026] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: an expansion unit, configured to expand the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain a lung field expansion mask image; a subtraction unit, configured to perform pixel-level subtraction operations on the lung field expansion mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura; a segmentation unit, configured to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image; a first determination unit, configured to determine the nearest pleural region image of each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image; a second determination unit, configured to determine whether the chest medical image contains a nodular pleural traction sign using the nearest nodular pleural region image of each nodule; and a prediction unit, configured to predict nodule-related preset chest diseases based on quantitative imaging features of preset nodules corresponding to the presence or absence of nodular pleural traction sign.

[0027] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: 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 aforementioned medical image intelligent analysis method.

[0028] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: a computer-readable storage medium storing a computer program / instructions and a bit stream thereon, wherein the computer program / instructions, when executed by a processor, implement the above-described medical image intelligent analysis method to generate the bit stream.

[0029] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: a computer program product configured with a computer program / instruction, which, when executed by a processor, implements the aforementioned medical image intelligent analysis method.

[0030] In this embodiment, a technical solution is proposed for a pleural mask image determination and intelligent analysis method and system to solve the technical problem that the pleura is not clearly displayed on the chest medical image, making it difficult to delineate the pleura, and thus failing to obtain a preset pleural segmentation model based on a deep learning segmentation network that meets the performance requirements.

[0031] 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.

[0032] 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

[0033] 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.

[0034] Figure 1 A flowchart illustrating a method for determining a pleural mask image according to an embodiment of the present disclosure; Figure 2 A flowchart illustrating a medical image intelligent analysis method according to an embodiment of the present disclosure is shown; Figure 3 This is a block diagram illustrating an electronic device 800 according to an exemplary embodiment; Figure 4 This is a block diagram illustrating an electronic device 1900 according to an exemplary embodiment. Detailed Implementation

[0035] 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.

[0036] 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.

[0037] 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.

[0038] 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.

[0039] 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.

[0040] In addition, this disclosure also provides a pleural mask image determination device or system, a medical image intelligent analysis device or system, an electronic device, a computer-readable storage medium, and a program. All of the above can be used to implement any of the pleural mask image determination and medical image intelligent analysis methods provided in this disclosure. The corresponding technical solutions and descriptions are described in the relevant section on pleural mask image determination and medical image intelligent analysis methods, and will not be repeated here.

[0041] Figure 1 A flowchart illustrating a method for determining a pleural mask image according to an embodiment of the present disclosure is shown, such as... Figure 1 As shown, the pleural mask image determination method includes: Step S1: Dilatating the lung field in the lung field mask image corresponding to the chest medical image using a second preset edge width to obtain a lung field dilatation mask image; Step S2: Performing pixel-level subtraction operations on the lung field dilatation mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura. This addresses the technical problem that the pleura is not clearly displayed on the chest medical image, making it difficult to delineate the pleura and thus preventing the obtaining of a preset pleural segmentation model based on a deep learning segmentation network that meets performance requirements.

[0042] Based on the aforementioned technical problems, and unlike existing technologies, this disclosure fully utilizes the anatomical positional relationship between the lung field and the pleura, and uses the lung field mask image to determine the pleural mask image corresponding to the pleura.

[0043] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can determine the second preset edge width according to actual needs. For example, the first preset edge width can be configured to 1-5 pixels or other pixel values.

[0044] In the embodiments disclosed herein and other possible embodiments, the two-dimensional chest medical image can be configured as a chest X-ray image or a chest ultrasound image; the three-dimensional chest medical image can be configured as one or more of a chest CT image, a chest MRI image, a chest PET image, a chest CT-PET image, and a chest MRI-PET image. Furthermore, the two-dimensional chest medical image can also be configured as one or more corresponding chest projection images of a chest CT image, a chest MRI image, a chest PET image, a chest CT-PET image, and a chest MRI-PET image.

[0045] In this embodiment of the disclosure, determining the lung field mask image corresponding to the chest medical image includes: segmenting the lung field in the chest medical image to obtain the lung field mask image corresponding to the chest medical image; and / or, segmenting the lung field in the chest medical image to obtain the lung field mask image corresponding to the chest medical image includes: acquiring a preset lung field segmentation model constructed based on a deep learning segmentation network; and using the preset lung field segmentation model to segment the lung field in the chest medical image to obtain the lung field mask image corresponding to the chest medical image.

[0046] In this embodiment of the disclosure, before segmenting the lung fields in the chest medical image using the preset lung field segmentation model, the method includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding lung field mask training images to obtain a preset lung field segmentation model constructed based on the deep learning segmentation network.

[0047] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate the lung fields of a preset number of chest medical images to obtain corresponding lung field mask training images.

[0048] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding lung field mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate the lung fields of the preset number of chest medical images to obtain corresponding lung field mask training images. Those skilled in the art can configure the preset number according to actual needs.

[0049] Figure 2 A flowchart illustrating a medical image intelligent analysis method according to an embodiment of the present disclosure is shown, such as... Figure 2 As shown, the intelligent medical image analysis method includes: Step S3: using the pleural mask image determination method described above, determining the pleural mask image corresponding to the pleura in the chest medical image; Step S4: using the lung field dynamic region division method, dividing the lung field region of the chest medical image to obtain the lung field region division network corresponding to each nodule in the chest medical image; Step S5: based on the lung field region division network corresponding to each nodule and the pleural mask image, determining the nearest pleural region image of each nodule.

[0050] In this embodiment of the disclosure, determining each nodule in the chest medical image includes: segmenting the nodules in the chest medical image to determine each nodule in the chest medical image.

[0051] In this embodiment of the disclosure, segmenting the nodules in the chest medical image and determining each nodule in the chest medical image includes: segmenting the nodules in the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network to obtain a nodule mask image corresponding to the chest medical image; and determining each nodule in the chest medical image based on the nodule mask image.

[0052] In the embodiments of this disclosure and other possible embodiments, before performing nodule segmentation on the chest medical image using the preset nodule segmentation model constructed based on a deep learning segmentation network, the method includes: training the preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain the preset nodule segmentation model constructed based on the deep learning segmentation network.

[0053] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate nodules on a preset number of chest medical images to obtain corresponding nodule mask training images.

[0054] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules on the preset number of chest medical images to obtain corresponding nodule mask training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500.

[0055] In the embodiments disclosed herein and other possible embodiments, the 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).

[0056] In this embodiment of the disclosure, the intelligent medical image analysis method further includes: using the image of the nearest nodular pleural region to each nodule to determine whether the chest medical image exhibits nodular pleural traction. In this embodiment of the disclosure, determining whether the chest medical image contains a nodular pleural traction sign using the nearest nodular pleural region image for each nodule includes: using a pre-defined feature target detection model based on deep learning to perform feature target detection on the nodular pleural region image, and determining whether there are cord-like or linear features between the adjacent pleura of the nodule at the preset position in the nodular pleural region image; if so, determining that the chest medical image contains a nodular pleural traction sign; otherwise, determining that the chest medical image does not contain a nodular pleural traction sign.

[0057] In this embodiment of the disclosure, based on the quantitative imaging features of the selected preset nodules corresponding to the presence or absence of nodular pleural traction signs, preset chest diseases related to nodules are predicted.

[0058] In this embodiment of the disclosure, the step of predicting nodule-related preset chest diseases based on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of nodular pleural traction signs includes: performing feature selection on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of nodular pleural traction signs to obtain preset nodule quantitative imaging features corresponding to preset chest diseases; and predicting nodule-related preset chest diseases based on the preset nodule quantitative imaging features corresponding to the preset chest diseases.

[0059] In this embodiment of the disclosure, the step of predicting the preset chest disease related to the preset nodule based on the preset quantitative imaging features of the preset nodule corresponding to the preset chest disease includes: training the preset predictor and constructing a preset chest disease model by using the preset quantitative imaging features of the preset nodule corresponding to the preset chest disease in the chest medical image at the first time and the preset chest disease results at the second time after the first time.

[0060] In this embodiment of the disclosure, the preset quantitative imaging features to be selected for nodules further include one or more of the following: nodule long diameter, whether the nodule edge is clear, nodule morphology, nodule spinous process sign, nodule lobulation sign, and nodule cavitation sign.

[0061] In this embodiment, detecting the nodule spinous process sign includes: calculating the ratio of the major axis to the minor axis of each nodule based on the major axis and minor axis of each nodule in the nodule mask image corresponding to the chest medical image; wherein the chest medical image is configured as a two-dimensional medical image; if the ratio of the major axis to the minor axis is greater than or equal to a first preset ratio of the major axis to the minor axis, then the corresponding nodule is determined to have a spinous process sign. This addresses at least one of the technical problems currently present, such as the excessive subjectivity of manually determining whether a nodule edge has a spinous process sign, the lack of intelligent determination of whether a nodule edge has a spinous process sign leading to a huge workload for doctors, and the inability to effectively evaluate the subsequent radiotherapy efficacy prediction.

[0062] This disclosure also proposes a method for detecting a spinous process sign, comprising: calculating multiple ratios of the major and minor axes of each three-dimensional nodule in a nodule mask image corresponding to the chest medical image; if any one of the major-minor axis ratios is greater than or equal to a first preset major-minor axis ratio, then the corresponding nodule is determined to have a spinous process sign; wherein the chest medical image is configured as a three-dimensional medical image. This addresses at least one of the technical problems currently present, such as the excessive subjectivity of manually determining whether a spinous process sign exists at the nodule edge, the lack of intelligent determination of whether a spinous process sign exists at the nodule edge leading to a huge workload for doctors, and the inability to effectively evaluate the subsequent prediction of radiotherapy efficacy.

[0063] This disclosure also proposes a method for detecting a spinous process sign, comprising: determining whether a chest medical image is a two-dimensional or three-dimensional medical image; if the chest medical image is a two-dimensional medical image, calculating the ratio of the major axis to the minor axis of each nodule in the nodule mask image corresponding to the chest medical image; if the ratio of the major axis to the minor axis is greater than or equal to a first preset ratio of the major axis to the minor axis, determining that the corresponding nodule has a spinous process sign; otherwise, determining that the corresponding nodule does not have a spinous process sign; if the chest medical image is a three-dimensional medical image, calculating multiple ratios of the major axis to the minor axis of each three-dimensional nodule in the nodule mask image corresponding to the chest medical image; if any one of the ratios of the major axis to the minor axis is greater than or equal to the first preset ratio of the major axis to the minor axis, determining that the corresponding nodule has a spinous process sign.

[0064] In this embodiment of the disclosure, determining the major axis corresponding to each three-dimensional nodule includes: performing three-dimensional reconstruction on each nodule mask in the first nodule mask image to obtain a three-dimensional mask nodule corresponding to each nodule mask; performing edge detection on the three-dimensional mask nodule of each nodule mask to obtain three-dimensional nodule mask edges; selecting any three-dimensional edge pixel point from the three-dimensional nodule mask edges; calculating multiple three-dimensional edge pixel point distances between the arbitrary three-dimensional edge pixel point and other three-dimensional edge pixel points; selecting the maximum three-dimensional edge pixel point distance from the multiple three-dimensional edge pixel point distances; and configuring the maximum three-dimensional edge pixel point distance as the major axis corresponding to each three-dimensional nodule.

[0065] In this embodiment of the disclosure, determining the short path corresponding to each nodule includes: performing three-dimensional reconstruction on each nodule mask in the first nodule mask image to obtain a three-dimensional mask nodule corresponding to each nodule mask; performing edge detection on the three-dimensional mask nodule of each nodule mask to obtain the three-dimensional nodule mask edge; selecting any three-dimensional edge pixel in the three-dimensional nodule 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; configuring the minimum three-dimensional edge pixel distance as the long path corresponding to each three-dimensional nodule; if the ratio of the long path to the short path is less than the first preset ratio of the long path to the short path, then it is determined that the corresponding nodule does not have a spur sign.

[0066] In this embodiment of the disclosure, determining the nodule mask image corresponding to the chest medical image includes: training a first preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule mask training images to obtain a preset nodule segmentation model; segmenting the nodules in the chest medical image using the preset nodule segmentation model to obtain the corresponding nodule mask image; segmenting the arteries using the preset artery segmentation model to obtain the corresponding artery mask image; and / or, the nodules in the chest medical image are configured as nodules to be treated with chemotherapy or nodules to be operated on.

[0067] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each nodule region in the nodule mask image and the nodule mask training image is configured as 1, and the mask value corresponding to non-nodule regions is configured as 0.

[0068] In the embodiments of this disclosure and other possible embodiments, before training the first preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules in the preset number of preset site medical training images to obtain corresponding nodule mask training images.

[0069] In this embodiment of the disclosure, if the nodule has a spinous process, it is determined whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask image and artery mask image corresponding to the chest medical image; if the spinous process corresponding to the nodule is adjacent to an artery, the spinous process is determined to be a dangerous spinous process.

[0070] In this embodiment of the disclosure, if the nodule has a spinous process, it is determined whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask and artery mask image corresponding to the chest medical image; if the spinous process corresponding to the nodule is adjacent to an artery, the spinous process is determined to be a dangerous spinous process.

[0071] In this embodiment of the disclosure, determining whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask image and the artery mask image corresponding to the chest medical image includes: extracting multiple spinous process position coordinates corresponding to multiple spinous processes in the nodule mask image and artery mask edge position coordinates corresponding to the artery mask edge in the artery mask image; calculating multiple distances between the multiple spinous process position coordinates and the artery mask edge position coordinates; if the minimum distance among the multiple distances is less than a preset distance, then it is determined that the spinous process corresponding to the nodule is adjacent to an artery.

[0072] In this embodiment of the disclosure, determining whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask and artery mask image corresponding to the chest medical image includes: extracting multiple spinous process position coordinates corresponding to multiple spinous processes and artery mask edge position coordinates corresponding to the artery mask edge in the nodule mask and artery mask image; calculating multiple distances between the multiple spinous process position coordinates and the artery mask edge position coordinates respectively; if the minimum distance among the multiple distances is less than a preset distance, then it is determined that the spinous process corresponding to the nodule is adjacent to an artery.

[0073] In the embodiments of this disclosure and other possible embodiments, determining the spike includes: detecting the edges of the nodule mask in the nodule mask image to obtain nodule mask edge lines; calculating a first pixel distance and a second pixel distance between any nodule mask edge pixel on the nodule mask edge line and its adjacent nodule mask edge pixels; if the first pixel distance and the second pixel distance are greater than a set pixel distance, then configuring the arbitrary nodule mask edge pixel as a spike.

[0074] In this disclosure and other possible embodiments, determining whether the location of the nodule corresponding to the nodule to be treated or the nodule to be operated on is adjacent to an artery based on the nodule 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 nodule edge mask position coordinates corresponding to the nodule mask image; if the minimum positional distance among the multiple positional distances is less than or equal to a preset positional distance, then the nodule location is determined to be adjacent to an artery; otherwise, the nodule location is determined to be far from an artery. The artery edge mask position coordinates and the nodule edge mask position coordinates corresponding to the nodule are respectively configured as the position coordinates of the artery edge in the artery edge mask image corresponding to the artery mask image and the position coordinates of the nodule edge mask in the nodule edge mask image corresponding to the nodule mask image.

[0075] In this embodiment and other possible embodiments, determining whether the location of the nodule corresponding to the nodule to be treated or the nodule to be operated on is adjacent to an artery based on the nodule mask and artery mask images includes: calculating multiple positional distances between the artery edge mask position coordinates corresponding to the nodule mask and the artery mask image and the nodule edge mask position coordinates; if the minimum positional distance among the multiple positional distances is less than or equal to a preset positional distance, then the nodule location is determined to be adjacent to an artery; otherwise, the nodule location is determined to be far from an artery. The artery edge mask position coordinates and the nodule edge mask position coordinates corresponding to the nodule are respectively configured as the position coordinates of the artery edge in the artery edge mask and the position coordinates of the nodule edge mask in the nodule mask and artery mask image.

[0076] 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.

[0077] In this embodiment, the preset chest disease is configured as lymph node metastasis in partially solid lung adenocarcinoma. Specifically, the prediction of the preset chest disease based on the quantitative imaging features of the preset nodules corresponding to the preset chest disease includes: training a preset predictor using the quantitative imaging features of the preset nodules corresponding to partially solid lung adenocarcinoma in the chest medical image at a first time point and the pathological results of the lymph nodes in partially solid lung adenocarcinoma corresponding to metastasis or non-metastasis at a second time point after the first time point to construct a lymph node metastasis prediction model; and using the lymph node metastasis prediction model and the quantitative imaging features of the preset nodules corresponding to partially solid lung adenocarcinoma in the chest medical image to be predicted to predict lymph node metastasis in partially solid lung adenocarcinoma. This addresses the technical problem of the lack of accurate, effective, and non-invasive means to assess lymph node metastasis in partially solid lung adenocarcinoma nodules preoperatively.

[0078] In embodiments of this disclosure and other possible embodiments, the preset predictor may be configured as a prediction network based on deep learning (e.g., CNN, RNN, GCN, etc.) or a predictor based on machine learning (e.g., SVM, Random Forest, XGBoost, etc.).

[0079] In this embodiment of the disclosure, before training the preset predictor, the method includes: extracting quantitative imaging features of candidate nodules corresponding to partial solid lung adenocarcinoma in chest medical images; and selecting preset quantitative imaging features of nodules related to the pathological results of lymph nodes in the partial solid lung adenocarcinoma from the candidate nodule quantitative imaging features.

[0080] In this embodiment of the disclosure, determining the preset quantitative imaging features of nodules related to the pathological results of lymph nodes in the partial solid lung adenocarcinoma includes: using a univariate logistic regression analysis algorithm, the quantitative imaging features of the nodules to be selected and their corresponding pathological results of lymph nodes in the partial solid lung adenocarcinoma, performing univariate logistic regression on each of the quantitative imaging features of the nodules to be selected, and determining the preset quantitative imaging features of nodules related to the pathological results of lymph nodes in the partial solid lung adenocarcinoma.

[0081] In this embodiment of the disclosure, the quantitative imaging features of the nodule to be selected include one or more of the following: the long diameter of the nodule, the long diameter of the solid component of the nodule, the proportion of the solid component of the nodule to the nodule itself, the location of the nodule, whether the edge of the nodule is clear, the morphology of the nodule, the nodule spinous process sign, the nodule lobulation sign, the nodule pleural traction sign, the nodule cavitation sign, and the nodule air bronchus sign.

[0082] In this embodiment of the present disclosure, the step of extracting quantitative imaging features of candidate nodules corresponding to partial solid lung adenocarcinoma in chest medical images includes: segmenting the nodules corresponding to partial solid lung adenocarcinoma in the chest medical images to obtain a nodule mask image; and based on the nodule mask image, determining one or more of the following quantitative imaging features of the candidate nodules: nodule major diameter, whether the nodule edge is clear, nodule morphology, nodule spinous process sign, nodule lobulation sign, and nodule vacuolation sign.

[0083] In this embodiment of the disclosure, before training the preset predictor, the numerical quantitative imaging features of the candidate nodules are standardized or normalized, and the categorical quantitative imaging features of the candidate nodules are encoded.

[0084] In the embodiments of this disclosure and other possible embodiments, for example, before training the preset predictor, the quantitative imaging features of the nodule corresponding to the nodule major diameter, the nodule solid component major diameter, and the proportion of nodule solidity to the nodule are standardized or normalized to eliminate the influence of different dimensions on the training of the preset predictor.

[0085] In this disclosed embodiment and other possible embodiments, after determining one or more of the following quantitative imaging features of the nodule to be selected based on the nodule mask image: nodule location, clarity of nodule edges, nodule shape, presence of spinous process, presence of lobulation, presence of pleural traction, presence of cavitation, and presence of air bronchogram, respectively, the following features are encoded to obtain... The system uses one or more of the coding features corresponding to the quantitative imaging features of the nodule to be selected, such as nodule location, clarity of nodule edge, nodule shape, presence of spinous process, presence of lobulation, presence of pleural traction, presence of cavitation, and presence of air bronchus, to train a preset predictor and construct a lymph node metastasis prediction model.

[0086] In this embodiment and other possible embodiments, the nodule locations corresponding to the quantitative imaging features of the selected nodules are encoded to obtain different nodule location encoding features corresponding to a first numerical queue (0, 1, 2, ... or 0, 2, 4, ...) for different nodule locations; wherein, the different nodule location encoding features corresponding to the first numerical queue (0, 1, 2, ... or 0, 2, 4, ...) for different nodule locations are configured according to a predefined lung field anatomical region. Similarly, the nodule morphology corresponding to the quantitative imaging features of the selected nodules is encoded to obtain different nodule morphology encoding features corresponding to a second numerical queue (0, 1, 2, ... or 1, 3, 5, ...); wherein, the second numerical queue (0, 1, 2, ... or 1, 3, 5, ...) corresponding to the different nodule morphology encoding features are configured according to different types corresponding to the preset closed curve equation or the preset closed surface equation.

[0087] In this disclosure and other possible embodiments, one or more of the following features corresponding to the quantitative imaging features of the nodule to be selected are encoded: whether the nodule edge is clear, whether the nodule has a spiky sign, whether the nodule has a lobulated sign, whether the nodule has a pleural traction sign, whether the nodule has a cavitation sign, and whether the nodule has an air bronchus sign. This includes: if the features are present, the corresponding quantitative imaging feature of the nodule to be selected is configured with a first value (e.g., 1); otherwise, the corresponding quantitative imaging feature of the nodule to be selected is configured with a second value different from the first value (e.g., 0). For example, if the nodule edge is clear, then when encoding the clear nodule edge, the clear nodule edge is configured with a first value; if the nodule has a spiky sign, then when encoding the spiky nodule sign, the spiky nodule sign is configured with a first value.

[0088] In embodiments of this disclosure and other possible embodiments, determining whether the nodule edge corresponding to the quantitative imaging features of the selected nodule is clear based on the nodule mask image includes: performing an erosion operation on the nodule mask image to obtain a nodule mask eroded image; subtracting the nodule mask eroded image from the nodule mask image to obtain a nodule mask processed image; mapping the nodule position coordinates in the nodule mask processed image to the chest medical image; calculating multiple gradients corresponding to the nodule position coordinates in the chest medical image; if the gradient magnitude values ​​corresponding to the multiple nodules are greater than or equal to a preset gradient magnitude value, then the nodule boundary where the nodule position corresponding to the preset gradient magnitude value is located is determined to be clear; otherwise, the nodule boundary is determined to be unclear.

[0089] In embodiments of this disclosure and other possible embodiments, the step of calculating multiple gradients corresponding to the nodule location coordinates in the chest medical image includes: determining multiple first position coordinates and multiple second position coordinates corresponding to both sides of the nodule location coordinates in the chest medical image in the x-direction, and multiple third position coordinates and multiple fourth position coordinates corresponding to the y-direction, based on the nodule location coordinates; calculating the differences between the first pixel values ​​corresponding to the multiple first position coordinates and the second pixel values ​​corresponding to the multiple second position coordinates to obtain multiple x-direction gradients corresponding to the nodule location coordinates; calculating the differences between the third pixel values ​​corresponding to the multiple third position coordinates and the fourth pixel values ​​corresponding to the multiple fourth position coordinates to obtain multiple y-direction gradients corresponding to the nodule location coordinates; and configuring the multiple x-direction gradients and their corresponding multiple y-direction gradients as multiple gradients corresponding to the nodule location coordinates.

[0090] In the embodiments disclosed herein and other possible embodiments, determining multiple first position coordinates and multiple second position coordinates on both sides of the nodule position coordinates in the x-direction in the chest medical image based on the nodule position coordinates includes: using the nodule position coordinates, determining multiple first position coordinates and multiple second position coordinates on both sides in the x-direction according to a preset distance.

[0091] In the embodiments of this disclosure and other possible embodiments, determining multiple third position coordinates and multiple fourth position coordinates on both sides of the nodule position coordinates in the chest medical image in the y direction based on the nodule position coordinates includes: using the nodule position coordinates, determining multiple third position coordinates and multiple fourth position coordinates on both sides in the x direction according to a preset distance.

[0092] In the embodiments disclosed herein and other possible embodiments, determining the nodule morphology corresponding to the quantitative imaging features of the selected nodule based on the nodule mask image includes: determining a nodule mask processing image corresponding to each nodule in the nodule mask image of a chest medical image; extracting multiple nodule mask position coordinates corresponding to the nodule mask lines in the nodule mask processing image corresponding to each nodule; and fitting or reconstructing the multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology corresponding to each nodule.

[0093] In the embodiments of this disclosure and other possible embodiments, the step of fitting or reconstructing the multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology includes: under the optimal criterion, approximating multiple preset closed curve equations or multiple preset closed surface equations using the multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology corresponding to each nodule.

[0094] In the embodiments disclosed herein and other possible embodiments, the step of approximating a preset closed curve equation or a preset closed surface equation using the multiple nodule mask position coordinates corresponding to each nodule under the optimal criterion to determine the nodule morphology corresponding to each nodule includes: under the optimal criterion, minimizing the error or absolute residual of the nodule mask position coordinates between the approximation equations and the multiple near-preset closed curve equations or multiple preset closed surface equations when the multiple nodule mask position coordinates corresponding to each nodule approximate the multiple preset closed curve equations or multiple preset closed surface equations, thereby determining the nodule morphology corresponding to each nodule.

[0095] In the embodiments disclosed herein and other possible embodiments, the step of approximating a preset closed curve equation or a preset closed surface equation using multiple nodule mask position coordinates corresponding to each nodule under the optimal criterion to determine the nodule morphology corresponding to each nodule includes: if the chest medical image is configured as a two-dimensional medical image, then under the optimal criterion, approximating multiple preset closed curve equations using multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology corresponding to each nodule; if the chest medical image is configured as a three-dimensional medical image, then under the optimal criterion, approximating multiple preset closed surface equations using multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology corresponding to each nodule.

[0096] In the embodiments disclosed herein and other possible embodiments, the optimal criterion is configured as one or more of the least squares criterion, the minimum absolute residual criterion, etc.

[0097] In the embodiments of this disclosure and other possible embodiments, the optimal criterion is configured as the least squares method criterion. Then, after minimizing the error between the approximation equation and the multiple near-preset closed curve equation or multiple preset closed surface equation when the multiple nodule mask position coordinates corresponding to each nodule are approximated to the multiple preset closed curve equation or multiple preset closed surface equation, multiple nodule shapes corresponding to each nodule are obtained. The nodule mask position coordinate errors corresponding to the multiple nodule shapes are extracted respectively. The nodule shape with the smallest error among the multiple nodule shape's corresponding nodule position coordinate errors is configured as the final nodule shape corresponding to each nodule.

[0098] In the embodiments of this disclosure and other possible embodiments, the optimal criterion is configured as the least squares criterion. Then, after minimizing the absolute residuals of the nodule mask position coordinates of the multiple nodule mask position coordinates that approximate multiple preset closed curve equations or multiple preset closed surface equations when approximating multiple nodule mask position coordinates of each nodule, multiple nodule shapes corresponding to each nodule are obtained. The absolute residuals of the nodule mask position coordinates corresponding to the multiple nodule shapes are extracted respectively. The nodule shape corresponding to the minimum error of the absolute residuals of the nodule mask position coordinates corresponding to the multiple nodule shapes is configured as the final nodule shape corresponding to each nodule.

[0099] In the embodiments disclosed herein and other possible embodiments, the preset closed curve equation is configured as one or more of the two-dimensional closed curves corresponding to a first-order polynomial, a second-order polynomial, or a higher-order polynomial, which are used in polynomial fitting. For example, the preset closed curve equation is configured as a closed curve corresponding to a circle, ellipse, or conic section, a closed curve enclosed by line segments corresponding to a triangle, quadrilateral, pentagon, or hexagon, a spline curve, or a polyline closed curve, etc.

[0100] In the embodiments of this disclosure and other possible embodiments, the preset closed surface equation is configured as a bivariate polynomial or a multivariate polynomial. For example, the preset closed surface equation is configured as a closed surface formed by connecting the beginning and end of line segments, such as a circular surface, elliptical surface, conical surface, triangular surface, quadrilateral surface, pentagonal surface, hexagonal surface, etc., corresponding to a closed surface formed by basic geometric closed surfaces, or a closed surface composed of multiple surfaces.

[0101] In embodiments of this disclosure and other possible embodiments, determining the nodule mask processing image includes: performing an erosion operation on the nodule mask image to obtain a nodule mask erosion image; and subtracting the nodule mask erosion image from the nodule mask image to obtain the nodule mask processing image.

[0102] In the embodiments of this disclosure and other possible embodiments, before determining the nodule mask processing image corresponding to the nodule in the nodule mask image of the chest medical image, the method includes: using a preset nodule segmentation model constructed based on a deep learning segmentation network to segment the chest medical image into nodules, and determining the nodule mask image of the chest medical image.

[0103] In the embodiments of this disclosure and other possible embodiments, before segmenting chest medical images into nodules using a preset segmentation model constructed based on a deep learning segmentation network, the method includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset nodule segmentation model constructed based on the deep learning segmentation network.

[0104] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset nodule segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software to delineate nodule lesions on a preset number of chest medical images to obtain corresponding nodule mask training images.

[0105] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodule lesions on the preset number of chest medical images to obtain corresponding nodule mask training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500.

[0106] In the embodiments disclosed herein and other possible embodiments, the 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).

[0107] In this embodiment of the disclosure, determining the major diameter of the nodule corresponding to the quantitative imaging features of the nodule to be selected includes: performing edge detection on each nodule in the nodule mask image corresponding to the chest medical image to obtain the edge line of each nodule mask; setting a reference mask point on each nodule mask edge line respectively; calculating multiple distances from the reference mask point to multiple mask points on each nodule mask edge line excluding the reference mask point; and configuring the maximum distance among the multiple distances as the major diameter of each nodule.

[0108] In this embodiment of the disclosure, determining the nodular pleural traction sign includes: detecting an image of a nodular pleural region between a nodule at a predetermined location in a lung field and the pleura adjacent to the nodule at the predetermined location in a chest medical image; and determining whether the nodular pleural traction sign exists in the chest medical image based on the nodular pleural region image. This addresses at least one of the technical problems of existing methods that require manual determination of the nodular pleural traction sign, such as excessive subjectivity, lack of intelligent determination of the nodular pleural traction sign leading to a huge workload for doctors, and inability to further predict lymph node metastasis in some solid lung adenocarcinomas.

[0109] In the embodiments of this disclosure and other possible embodiments, determining the nodule at a preset location in the lung field includes: segmenting the nodule in the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network to obtain a nodule mask image corresponding to the chest medical image; and determining the nodule at the preset location in the lung field based on the nodule position coordinates in the nodule mask image.

[0110] In the embodiments of this disclosure and other possible embodiments, determining the nodule at a preset position in the lung field based on the nodule position coordinates in the nodule mask image includes: dividing the chest medical image into lung field regions using a lung field dynamic region division method to obtain a lung field region division network corresponding to each nodule; locating the lung field anatomical region corresponding to each nodule based on its position and the corresponding lung field region division network; determining the nodule at a preset position in the lung field based on the lung field anatomical region corresponding to each nodule; and / or, determining the nodule at a preset position in the lung field based on the lung field anatomical region corresponding to each nodule includes: if the lung field anatomical region corresponding to the nodule is close to the lung field boundary, then the nodule close to the lung field boundary is determined as the nodule at the preset position in the lung field.

[0111] In the embodiments of this disclosure and other possible embodiments, determining whether the lung field anatomical region corresponding to the nodule is close to the lung field boundary includes: calculating multiple nodule lung field boundary distances for each nodule from the lung field boundary in the lung field mask image corresponding to the chest medical image; and determining whether the lung field anatomical region corresponding to the nodule is close to the lung field boundary based on the multiple nodule lung field boundary distances and a preset distance.

[0112] In the embodiments of this disclosure and other possible embodiments, determining whether the anatomical region of the lung field corresponding to the nodule is close to the lung field boundary based on the distances between the multiple nodule lung field boundaries and a preset distance includes: if the distances between the multiple nodule lung field boundaries are less than or equal to the preset distance, then the anatomical region of the lung field corresponding to the nodule is determined to be close to the lung field boundary; otherwise, the anatomical region of the lung field corresponding to the nodule is determined to be far from the lung field boundary. Those skilled in the art can determine the preset distance according to actual needs. For example, the preset distance can be configured as 1-10 pixels or other pixel values.

[0113] In the embodiments of this disclosure and other possible embodiments, before performing nodule segmentation on the chest medical image using the preset nodule segmentation model constructed based on a deep learning segmentation network, the method includes: training the preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain the preset nodule segmentation model constructed based on the deep learning segmentation network.

[0114] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate nodules on a preset number of chest medical images to obtain corresponding nodule mask training images.

[0115] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules on the preset number of chest medical images to obtain corresponding nodule mask training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500.

[0116] In the embodiments disclosed herein and other possible embodiments, the 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).

[0117] In the embodiments of this disclosure and other possible embodiments, determining the nodules at preset locations in the lung field includes: segmenting the lung field in the chest medical image using a preset lung field segmentation model constructed based on a deep learning segmentation network to obtain a lung field mask image corresponding to the chest medical image; extracting the lung field from the lung field mask image by a first preset edge width to obtain a lung field image with a first preset edge width; segmenting the nodules in the first preset edge width lung field image using a preset nodule segmentation model constructed based on a deep learning segmentation network to obtain a nodule mask image with a first preset edge width; and determining the nodules in the first preset edge width nodule mask image as nodules at preset locations in the lung field. This addresses at least one of the technical problems, such as excessive subjectivity in nodule location positioning, the heavy workload for doctors due to the lack of intelligent nodule location positioning, and the inability to further predict lymph node metastasis in partially solid lung adenocarcinoma.

[0118] In the embodiments of this disclosure and other possible embodiments, the step of extracting the lung field from the lung field mask image by a first preset edge width to obtain a lung field image with a first preset edge width includes: performing an erosion operation on the lung field in the lung field mask image using the first preset edge width to obtain a lung field erosion mask image; performing a pixel-level subtraction operation on the lung field mask image and the lung field erosion mask image to obtain a lung field mask image with a first preset edge width; and performing a pixel-level multiplication operation on the lung field mask image with the chest medical image to obtain a lung field image with a first preset edge width.

[0119] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to the lung field in the lung field mask image can be configured to 1. Those skilled in the art can determine the first preset edge width according to actual needs. For example, the first preset edge width can be configured to 1-10 pixels or other pixel values.

[0120] In the embodiments of this disclosure and other possible embodiments, before segmenting the chest medical image using a preset lung field segmentation model constructed based on a deep learning segmentation network, the method includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding lung field mask training images to obtain a preset lung field segmentation model constructed based on the deep learning segmentation network.

[0121] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate the lung fields of a preset number of chest medical images to obtain corresponding lung field mask training images.

[0122] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding lung field mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate the lung fields of the preset number of chest medical images to obtain corresponding lung field mask training images. Those skilled in the art can configure the preset number according to actual needs.

[0123] In the embodiments of this disclosure and other possible embodiments, determining the pleural mask image corresponding to the pleura includes: expanding the lung field in the lung field mask image using a second preset edge width to obtain an expanded lung field mask image; performing pixel-level subtraction operations on the expanded lung field mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura; and / or, to solve the technical problem that the pleura is not clearly displayed on the chest medical image, making it difficult to delineate the pleura, thus failing to obtain a preset pleural segmentation model based on a deep learning segmentation network with satisfactory performance.

[0124] In the embodiments disclosed herein and other possible embodiments, the detection of the nodular pleural region image between a nodule at a predetermined location in the lung field and the pleura adjacent to the nodule at the predetermined location in a chest medical image includes: drawing multiple rays corresponding to multiple predetermined directions from the nodule boundary line or nodule surface of the nodule at the predetermined location to the pleural mask boundary line or pleural mask surface adjacent to the pleura at the predetermined location; and determining the image area enclosed by the rays intersecting the multiple rays and the pleural mask boundary line or pleural mask surface, the nodule boundary line or nodule surface of the nodule at the predetermined location, and the pleural mask boundary line or pleural mask surface adjacent to the pleura at the predetermined location as the nodular pleural region image.

[0125] In the embodiments disclosed herein and other possible embodiments, determining the image area enclosed by the rays intersecting the multiple rays with the pleural mask boundary line or pleural mask surface, the nodule boundary line or nodule surface of the preset position nodule, and the pleural mask boundary line or pleural mask surface of the pleura adjacent to the preset position nodule as a nodular pleural region image includes: determining the image area enclosed by the rays intersecting the multiple rays with the pleural mask boundary line or pleural mask surface, the nodule boundary line or nodule surface of the preset position nodule, and the pleural mask boundary line or pleural mask surface of the pleura adjacent to the preset position nodule as a nodular pleural region mask image; and performing a pixel-level multiplication operation on the nodular pleural region mask image and the chest medical image to determine the nodular pleural region image.

[0126] In the embodiments disclosed herein and other possible embodiments, if the chest medical image is configured as a two-dimensional medical image, then detecting the nodular pleural region image between a nodule at a preset position in the lung field of the chest medical image and the pleura adjacent to the nodule at the preset position includes: drawing multiple rays corresponding to multiple preset directions from the nodule boundary line of the nodule at the preset position to the pleural mask boundary line adjacent to the pleura at the preset position; and determining the area enclosed by the rays intersecting the multiple rays and the pleural mask boundary line, the nodule boundary line of the nodule at the preset position, and the pleural mask boundary line adjacent to the pleura at the preset position as the nodular pleural region image.

[0127] In the embodiments disclosed herein and other possible embodiments, the step of drawing multiple rays corresponding to multiple predetermined directions from the nodule boundary line of the nodule at the preset location to the pleural mask boundary line adjacent to the nodule at the preset location includes: setting an initial radiation point on the nodule boundary line of the nodule at the preset location; determining multiple radiation points on the nodule boundary line based on the initial radiation point and a preset radiation point step distance; and drawing multiple rays corresponding to multiple predetermined two-dimensional directions starting from the multiple radiation points.

[0128] In this disclosure and other possible embodiments, if the chest medical image is configured as a three-dimensional medical image, then detecting the nodular pleural region image between a nodule at a predetermined location in the lung field and the pleura adjacent to the nodule at the predetermined location in the chest medical image includes: drawing multiple rays corresponding to multiple predetermined directions from the nodule surface of the nodule at the predetermined location to the pleural mask surface adjacent to the pleura at the predetermined location; determining the area enclosed by the rays intersecting the multiple rays and the pleural mask surface, the nodule surface of the nodule at the predetermined location, and the pleura or pleural mask surface adjacent to the nodule at the predetermined location as the nodular pleural region image; and / or, The step of casting multiple rays corresponding to multiple predetermined directions from the surface of the nodule at the predetermined location to the pleural mask surface adjacent to the pleura at the predetermined location includes: determining a side pleural mask surface corresponding to the lung field anatomical region adjacent to the nodule at the predetermined location; setting an initial radiation point on the side pleural mask surface; determining multiple radiation points on the side pleural mask surface based on the initial radiation point and the predetermined radiation point step distance; and casting multiple rays corresponding to multiple predetermined three-dimensional directions starting from the multiple radiation points.

[0129] In this disclosure and other possible embodiments, the corresponding pleural mask boundary line or pleural mask surface is determined based on the pleural mask image corresponding to the pleura. Specifically, determining the corresponding pleural mask boundary line or pleural mask surface based on the pleural mask image corresponding to the pleura includes: if the chest medical image is configured as a two-dimensional medical image, edge detection is performed on the pleural mask image corresponding to the pleura to determine the corresponding pleural mask boundary line; if the chest medical image is configured as a three-dimensional medical image, edge detection is performed on each layer of the pleural mask image corresponding to the pleura to obtain the edge image of each layer of the pleural mask; and three-dimensional reconstruction is performed on each layer of the pleural mask edge image to determine the corresponding pleural mask surface.

[0130] In this disclosure and other possible embodiments, determining the pleural mask surface corresponding to the lung field anatomical region adjacent to the preset position nodule includes: dividing the pleura into left lung pleura and right lung pleura according to the lung field anatomical region of the preset position nodule; wherein the left lung pleura and right lung pleura correspond to the left lung pleural mask surface and the right lung pleural mask surface, respectively. Therefore, before determining the pleural mask surface corresponding to the lung field anatomical region adjacent to the preset position nodule, the method includes: dividing the pleura into multiple lung field anatomical pleural mask surfaces according to the lung field anatomical region of the preset position nodule; determining the pleural mask surface corresponding to the lung field anatomical region adjacent to the preset position nodule according to the multiple lung field anatomical pleural mask surfaces; setting initial radiation points on the pleural mask surface; and determining multiple radiation points on the pleural mask surface using the initial radiation points and the preset radiation point step distance.

[0131] For example, based on the lung field anatomical region of the pre-defined nodule, the pleura is divided into left lung pleura and right lung pleura; a left lung pleura mask surface or a right lung pleura mask surface corresponding to the lung field anatomical region adjacent to the pre-defined nodule is determined; an initial radiation point is set on the left lung pleura mask surface or the right lung pleura mask surface; multiple radiation points on the pleura mask surface are determined based on the initial radiation point and the pre-defined radiation point step distance.

[0132] In this disclosure and other possible embodiments, if the lung field anatomical region of the pre-positioned nodule is close to the anterior side of the left lung pleura, then the left lung pleural mask surface corresponding to the anterior side of the left lung pleura is retained; if the lung field anatomical region of the pre-positioned nodule is close to the anterior side of the right lung pleura, then the right lung pleural mask surface corresponding to the anterior side of the left lung pleura is retained. Based on the above, those skilled in the art can determine the specific pleural mask surface according to the actual anatomical structure of the lung field. Therefore, the pleural mask surface can also be configured as a pleural mask surface corresponding to different lung lobes. For example, the pleural mask surface of any one side of the right lung lobe.

[0133] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can determine the second preset edge width according to actual needs. For example, the second preset edge width can be configured to 1-5 pixels or other pixel values.

[0134] In the embodiments of this disclosure and other possible embodiments, determining whether there is a nodular pleural traction sign in the chest medical image based on the nodular pleural region image includes: detecting whether there are cord-like or linear features between the pleura adjacent to the nodule at the preset position in the nodular pleural region image; if so, determining that there is a nodular pleural traction sign in the chest medical image; otherwise, determining that there is no nodular pleural traction sign in the chest medical image.

[0135] In the embodiments of this disclosure and other possible embodiments, the step of detecting whether there are cord-like or linear features between the pleura adjacent to the nodule at the preset position in the nodular pleural region image includes: using a preset feature target detection model based on deep learning to perform feature target detection on the nodular pleural region image, and determining whether there are cord-like or linear features between the pleura adjacent to the nodule at the preset position in the nodular pleural region image.

[0136] In the embodiments of this disclosure and other possible embodiments, before performing feature target detection on the nodular pleural region image using a preset feature target detection model based on deep learning, the method includes: training feature point training images using a preset number of chest medical training images and their corresponding cord-like features and / or line-like features, training a preset deep learning target detection network, and obtaining a preset feature target detection model constructed based on the deep learning target detection network.

[0137] In the embodiments disclosed herein and other possible embodiments, before training the preset deep learning object detection network, the method includes: using segmentation and annotation software to delineate a preset number of chest medical images with linear and / or linear feature bounding boxes to obtain corresponding linear and / or linear feature training images. For example, annotation software such as CasiaLabeler or Labelme can be used to delineate linear and / or linear feature bounding boxes on the preset number of chest medical training images to obtain corresponding linear and / or linear feature training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500 or other possible values.

[0138] In the embodiments disclosed herein and other possible embodiments, the preset deep learning object detection network may be configured as a YOLO series or an object detection network based on YOLO.

[0139] In this embodiment of the disclosure, determining the proportion of solid nodules includes: segmenting the chest medical image using a deep learning-based solid nodule segmentation model to obtain a nodule segmentation mask image and a solid nodule segmentation mask image; and determining the proportion of solid nodules within the nodules based on the nodule segmentation mask image and the solid nodule segmentation mask image.

[0140] In the embodiments of this disclosure and other possible embodiments, the step of segmenting the chest medical image using a deep learning-based solid nodule segmentation model to obtain a nodule segmentation mask image and a solid nodule segmentation mask image includes: constructing training images of nodules in each set of chest medical training images in a chest medical training image set, and training images of binary mixed location masks corresponding to solid nodules within the nodules; training the preset deep learning segmentation network using the binary mixed location mask training images, the solid nodule edge mask training images corresponding to the solid nodules, and the corresponding chest medical training images to obtain a solid nodule segmentation model; and segmenting the chest medical image to be segmented based on the solid nodule segmentation model to obtain a nodule segmentation mask image and a solid nodule segmentation mask image. This addresses at least one of the technical problems of existing methods that require manual segmentation of solid nodules, resulting in a huge workload for doctors and hindering further surgical procedures and prediction of lymph node metastasis in some solid lung adenocarcinomas.

[0141] In the embodiments disclosed herein and other possible embodiments, unlike the prior art, a binary mixed location mask training image is constructed for each set of chest medical training images in the chest medical training image set, corresponding to nodules and solid nodules within the nodules. During the training of the preset deep learning segmentation network, the preset deep learning segmentation network pays more attention to the anatomical relationship between the nodules and solid nodules within the nodules, forming a collaborative learning and collaborative response to the nodules and solid nodules within the nodules. This enables effective anatomical segmentation of solid nodules within the nodules, creating anatomical constraints on the segmentation of solid nodules within the nodules.

[0142] In the embodiments disclosed herein and other possible embodiments, when each set of chest medical training images and the chest medical images to be segmented in the chest medical training image set can be configured as two-dimensional chest medical images, each set of chest medical training images may be one chest medical training image corresponding to a subject at a specific moment, or multiple chest medical training images corresponding to subjects at multiple specific moments. For example, when the two-dimensional chest medical images are configured as chest X-ray images or chest ultrasound images, each set of chest medical training images may be one chest medical training image corresponding to a subject at a specific moment corresponding to breath-holding, deep breathing, or deep inspiration, or multiple chest medical training images corresponding to subjects at multiple specific moments corresponding to the breathing process or breath-holding process.

[0143] In the embodiments disclosed herein and other possible embodiments, when each set of chest medical training images and the chest medical images to be segmented in the chest medical training image set can be configured as three-dimensional chest medical images, each set of chest medical training images may be multiple three-dimensional chest medical training images corresponding to the examinee at a specific time or multiple four-dimensional chest medical training images corresponding to the examinee at multiple specific times. For example, when the three-dimensional chest medical images can be configured as one or more of chest CT images, chest MRI images, chest PET images, chest CT-PET images, and chest MRI-PET images, each set of chest medical training images may be multiple three-dimensional chest medical training images corresponding to the examinee at a specific time corresponding to breath-holding or deep breathing or deep inspiration, or multiple four-dimensional chest medical training images corresponding to the examinee at multiple specific times corresponding to the breathing process or breath-holding process.

[0144] In the embodiments disclosed herein and other possible embodiments, the 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).

[0145] In the embodiments disclosed herein and other possible embodiments, each set of chest medical training images may include multiple nodules, and each nodule may include a solid nodule. That is, the multiple nodules in each set of chest medical training images may or may not contain solid nodules.

[0146] In the embodiments of this disclosure and other possible embodiments, the construction of nodules in each set of chest medical training images in the chest medical training image set and binary mixed position mask training images corresponding to solid nodules within the nodules includes: extracting nodule mask training images with first mask values ​​corresponding to nodules in each set of chest medical training images in the chest medical training image set and solid nodule mask training images with second mask values ​​corresponding to solid nodules within the nodules; mapping the second mask values ​​to the nodule positions in the solid nodule mask training images according to the positions of solid nodules in the solid nodule mask training images to construct binary mixed position mask training images.

[0147] In the embodiments of this disclosure and other possible embodiments, determining the solid nodule edge mask training image corresponding to the solid nodule includes: performing an erosion operation on the solid nodule mask training image corresponding to each set of chest medical training images in the chest medical training image set to obtain a solid nodule mask erosion training image; performing a pixel-level subtraction operation on the solid nodule mask training image and the solid nodule mask erosion training image to determine the solid nodule edge mask training image corresponding to the solid nodule.

[0148] In the embodiments disclosed herein and other possible embodiments, the second mask value is mapped to the nodule position in the nodule mask training image according to the position of the real nodule in the nodule mask training image, and the first mask value corresponding to the nodule is changed to the second mask value corresponding to the real nodule in the nodule.

[0149] In the embodiments of this disclosure and other possible embodiments, those skilled in the art can configure the first mask value and the second mask value according to actual needs. For example, the first mask value and the second mask value can be configured as 1 and 2, respectively. However, it is worth noting that, unlike the prior art where the preset mask value only represents a region corresponding to a lesion or area to be segmented, the second mask value in the embodiments of this disclosure represents both the first mask value corresponding to the nodule and the second mask value corresponding to the solid nodule within the nodule.

[0150] In the embodiments disclosed herein and other possible embodiments, if there is a solid nodule within the nodule, the binary mixed location mask training image includes a first numerical mask corresponding to the nodule and a second numerical mask corresponding to the solid nodule; if there is no solid nodule within the nodule, the binary mixed location mask training image includes a first numerical mask corresponding to the nodule.

[0151] In the embodiments of this disclosure and other possible embodiments, extracting the nodule mask training image with the first mask value corresponding to the nodules and the solid nodule mask training image with the second mask value corresponding to the solid nodules within the nodules for each set of chest medical training images in the chest medical training image set includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules and solid nodules within the nodules for each set of chest medical training images in the chest medical training image set of a preset number of chest medical training images, to obtain the nodule mask training image with the first mask value corresponding to the nodules and the solid nodule mask training image with the second mask value corresponding to the solid nodules within the nodules for each set of chest medical training images.

[0152] In the embodiments of this disclosure and other possible embodiments, the step of training the preset deep learning segmentation network using the binary mixed location mask training image, the solid nodule edge mask training image corresponding to the solid nodule, and the corresponding chest medical training image to obtain a solid nodule segmentation model includes: obtaining a binary mixed location mask image during the training process of the preset deep learning segmentation network using the binary mixed location mask training image and its corresponding chest medical training image; using a preset solid nodule edge encoder to encode the solid nodule edge mask image corresponding to the solid nodule segmentation mask image in the binary mixed location mask image and the solid nodule edge mask training image, respectively, to obtain a first edge encoding feature map and a second edge encoding feature map; calculating a first loss of the binary mixed location mask image and its corresponding solid nodule edge mask training image and calculating a second loss of the first edge encoding feature map and the second edge encoding feature map to obtain a solid nodule segmentation model.

[0153] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the loss functions corresponding to the first loss and the second loss according to actual needs. For example, the loss functions corresponding to the first loss and the second loss can be configured as one or more of the loss functions selected from Cross-Entropy Loss, Dice Loss, Tversky Loss, and Focal Loss.

[0154] In embodiments of this disclosure and other possible embodiments, the binary mixed location mask image includes: a nodule and a mask image corresponding to a solid nodule within the nodule; wherein the mask values ​​of the nodule and the mask values ​​corresponding to the solid nodule within the nodule are different. Furthermore, the solid nodule segmentation mask image in the binary mixed location mask image can be determined based on the mask values ​​of the nodule and the mask values ​​corresponding to the solid nodule within the nodule.

[0155] Furthermore, in the embodiments of this disclosure and other possible embodiments, determining the solid nodule edge mask image includes: performing an erosion operation on the solid nodule segmentation mask image to obtain a solid nodule mask erosion image; performing a pixel-level subtraction operation on the solid nodule mask image and the solid nodule mask erosion image to determine the solid nodule edge mask image corresponding to the solid nodule.

[0156] In the embodiments of this disclosure and other possible embodiments, the construction of the preset solid nodule edge encoder includes: constructing the preset solid nodule edge encoder using the solid nodule edge mask training image corresponding to the solid nodule and its corresponding solid nodule edge training image.

[0157] In the embodiments of this disclosure and other possible embodiments, the step of constructing a preset solid nodule edge encoder using the solid nodule edge mask training image corresponding to the solid nodule and its corresponding solid nodule edge training image includes: extracting the solid nodule edge training image corresponding to each set of chest medical training images in the chest medical training image set; and training a preset downsampling convolutional network using the solid nodule edge training image and its corresponding solid nodule edge mask training image to obtain the preset solid nodule edge encoder.

[0158] In the embodiments disclosed herein and other possible embodiments, the step of extracting the solid nodule edge training image corresponding to each set of chest medical training images in the chest medical training image set includes: performing pixel-level multiplication operations on the solid nodule edge mask training image corresponding to each set of chest medical training images in the chest medical training image set and its corresponding chest medical training image to extract the solid nodule edge training image.

[0159] In the embodiments disclosed herein and other possible embodiments, unlike the prior art, the preset solid nodule edge encoder enables the edges corresponding to the solid nodules within the nodules to participate in the training process of the preset deep learning segmentation network using the training images of the two-class mixed position mask and their corresponding chest medical training images. This makes the edges corresponding to the solid nodules within the nodules controllable and more conducive to the edge segmentation of the solid nodules within the nodules.

[0160] In the embodiments disclosed herein and other possible embodiments, during the construction of the preset solid nodule edge encoder, a third loss is calculated between the solid nodule edge mask training image corresponding to the solid nodule and its corresponding solid nodule edge training image. Similarly, those skilled in the art can configure the loss function corresponding to the third loss according to actual needs. For example, the loss function corresponding to the third loss can be configured as one or more of the following loss functions: Cross-Entropy Loss, Dice Loss, Tversky Loss, and Focal Loss.

[0161] In the embodiments disclosed herein and other possible embodiments, the preset downsampling convolutional network can be configured as one or more of the downsampling convolutional networks corresponding to ResNet, EfficientNet, LeNet, AlexNet, YOLOv10, SRFD / DRFD modules, HWD modules, MobileNet, ShuffleNet, etc.; wherein, those skilled in the art can configure the kernel size, stride of the kernel, pooling and other conventional layers of the above-mentioned preset downsampling convolutional network according to actual needs.

[0162] In the embodiments disclosed herein and other possible embodiments, the preset solid nodule edge encoder corresponding to the preset downsampling convolutional network can encode the solid nodule edge mask image corresponding to the solid nodule segmentation mask image in the two-class mixed position mask image and the solid nodule edge mask training image, thereby achieving one or more objectives such as compressing the spatial size of the solid nodule edge mask image and the solid nodule edge mask training image, expanding the receptive field, and enhancing translation invariance. Specifically, the following features are considered: Compressing spatial dimensions: Reducing the height and width of the feature maps corresponding to the solid nodule edge mask image and the solid nodule edge mask training image reduces subsequent computational load, making the preset solid nodule edge encoder more efficient; Expanding the receptive field: As the size of the feature maps corresponding to the solid nodule edge mask image and the solid nodule edge mask training image decreases, neurons in subsequent layers can cover a wider input region, helping to capture the global features corresponding to the solid nodule edge mask image and the solid nodule edge mask training image; Enhancing translation invariance: The encoder is more robust to small displacements or deformations of the solid nodule edge mask image and the solid nodule edge mask training image, improving the generalization ability of the preset solid nodule edge encoder.

[0163] For example, the pre-defined downsampling convolutional network can be configured as a hierarchical feature extraction structure or a gradually expanding receptive field structure. For a hierarchical feature extraction structure, it typically consists of multiple convolutional layers and pooling layers stacked alternately. With each downsampling operation, the resolution of the feature maps corresponding to the solid nodule edge mask image and the training image decreases, but the number of channels may increase to retain richer semantic information. For a gradually expanding receptive field structure, with each downsampling operation on the solid nodule edge mask image and the training image, the input region that the deep neurons of the network can "see" becomes larger and larger, facilitating a gradual transition from local features corresponding to the solid nodule edge mask image and the training image to global semantic understanding.

[0164] For example, by setting the stride of the convolutional kernel to be greater than 1 (e.g., stride=2), the convolutional kernel slides in jumps across the input feature maps corresponding to the solid nodule edge mask image and the training image, directly reducing the size of the output feature maps corresponding to these images. For instance, if the input size is 5×5, the kernel size is 3×3, and the stride is 2, the output size of the solid nodule edge mask image and the training image will become 2×2. Common max pooling and average pooling downsample the feature maps corresponding to the solid nodule edge mask image and the training image by selecting the maximum or average value of the local regions corresponding to these images. For example, 2×2 max pooling will halve the size of the feature map corresponding to the above solid nodule edge mask image and the solid nodule edge mask training image.

[0165] In the embodiments of this disclosure and other possible embodiments, before segmenting the chest medical image using a deep learning-based solid nodule segmentation model, the process includes: constructing nodules for each set of chest medical training images in a chest medical training image set and binary mixed location mask training images corresponding to solid nodules within the nodules; training the preset deep learning segmentation network using the binary mixed location mask training images, the solid nodule edge mask training images corresponding to the solid nodules, and the corresponding chest medical training images to obtain a solid nodule segmentation model; and performing solid nodule segmentation on the chest medical image to be segmented (chest medical image) based on the solid nodule segmentation model.

[0166] In the embodiments of this disclosure and other possible embodiments, based on the nodule mask image corresponding to the chest medical image, the nodule image corresponding to the nodule mask image is extracted from the chest medical image; based on the nodule image, one or more of the following are determined: the long diameter of the solid component of the nodule, the proportion of solid component of the nodule to the nodule, and the nodule air bronchogram sign corresponding to the quantitative imaging features of the nodule to be selected.

[0167] In this disclosure and other possible embodiments, detecting the nodule spinous process sign includes: calculating the ratio of the major axis to the minor axis of each nodule in the nodule mask image corresponding to the chest medical image; wherein the chest medical image is configured as a two-dimensional medical image; if the ratio of the major axis to the minor axis is greater than or equal to a first preset ratio of the major axis to the minor axis, then it is determined that the corresponding nodule has a spinous process sign. This addresses at least one of the technical problems currently present, such as the need for manual determination of whether a nodule edge has a spinous process sign, which is too subjective, lacks intelligent determination of whether a nodule edge has a spinous process sign, resulting in a huge workload for doctors and preventing further effective evaluation of radiotherapy efficacy prediction.

[0168] This disclosure also proposes a method for detecting a spinous process sign, comprising: calculating multiple ratios of the major and minor axes of each three-dimensional nodule in a nodule mask image corresponding to the chest medical image; if any one of the major-minor axis ratios is greater than or equal to a first preset major-minor axis ratio, then the corresponding nodule is determined to have a spinous process sign; wherein the chest medical image is configured as a three-dimensional medical image. This addresses at least one of the technical problems currently present, such as the excessive subjectivity of manually determining whether a spinous process sign exists at the nodule edge, the lack of intelligent determination of whether a spinous process sign exists at the nodule edge leading to a huge workload for doctors, and the inability to effectively evaluate the subsequent prediction of radiotherapy efficacy.

[0169] This disclosure also proposes a method for detecting a spinous process sign, comprising: determining whether a chest medical image is a two-dimensional or three-dimensional medical image; if the chest medical image is a two-dimensional medical image, calculating the ratio of the major axis to the minor axis of each nodule in the nodule mask image corresponding to the chest medical image; if the ratio of the major axis to the minor axis is greater than or equal to a first preset ratio of the major axis to the minor axis, determining that the corresponding nodule has a spinous process sign; otherwise, determining that the corresponding nodule does not have a spinous process sign; if the chest medical image is a three-dimensional medical image, calculating multiple ratios of the major axis to the minor axis of each three-dimensional nodule in the nodule mask image corresponding to the chest medical image; if any one of the ratios of the major axis to the minor axis is greater than or equal to the first preset ratio of the major axis to the minor axis, determining that the corresponding nodule has a spinous process sign.

[0170] In this embodiment of the disclosure, determining the major axis corresponding to each three-dimensional nodule includes: performing three-dimensional reconstruction on each nodule mask in the first nodule mask image to obtain a three-dimensional mask nodule corresponding to each nodule mask; performing edge detection on the three-dimensional mask nodule of each nodule mask to obtain three-dimensional nodule mask edges; selecting any three-dimensional edge pixel point from the three-dimensional nodule mask edges; calculating multiple three-dimensional edge pixel point distances between the arbitrary three-dimensional edge pixel point and other three-dimensional edge pixel points; selecting the maximum three-dimensional edge pixel point distance from the multiple three-dimensional edge pixel point distances; and configuring the maximum three-dimensional edge pixel point distance as the major axis corresponding to each three-dimensional nodule.

[0171] In this embodiment of the disclosure, determining the short path corresponding to each nodule includes: performing three-dimensional reconstruction on each nodule mask in the first nodule mask image to obtain a three-dimensional mask nodule corresponding to each nodule mask; performing edge detection on the three-dimensional mask nodule of each nodule mask to obtain the three-dimensional nodule mask edge; selecting any three-dimensional edge pixel in the three-dimensional nodule 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; configuring the minimum three-dimensional edge pixel distance as the long path corresponding to each three-dimensional nodule; if the ratio of the long path to the short path is less than the first preset ratio of the long path to the short path, then it is determined that the corresponding nodule does not have a spur sign.

[0172] In this embodiment of the disclosure, determining the nodule mask image corresponding to the chest medical image includes: training a first preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule mask training images to obtain a preset nodule segmentation model; segmenting the nodules in the chest medical image using the preset nodule segmentation model to obtain the corresponding nodule mask image; segmenting the arteries using the preset artery segmentation model to obtain the corresponding artery mask image; and / or, the nodules in the chest medical image are configured as nodules to be treated with chemotherapy or nodules to be operated on.

[0173] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each nodule region in the nodule mask image and the nodule mask training image is configured as 1, and the mask value corresponding to non-nodule regions is configured as 0.

[0174] In the embodiments of this disclosure and other possible embodiments, before training the first preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules in the preset number of preset site medical training images to obtain corresponding nodule mask training images.

[0175] In this embodiment of the disclosure, if the nodule has a spinous process, it is determined whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask image and artery mask image corresponding to the chest medical image; if the spinous process corresponding to the nodule is adjacent to an artery, the spinous process is determined to be a dangerous spinous process.

[0176] In this embodiment of the disclosure, if the nodule has a spinous process, it is determined whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask and artery mask image corresponding to the chest medical image; if the spinous process corresponding to the nodule is adjacent to an artery, the spinous process is determined to be a dangerous spinous process.

[0177] In this embodiment of the disclosure, determining whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask image and the artery mask image corresponding to the chest medical image includes: extracting multiple spinous process position coordinates corresponding to multiple spinous processes in the nodule mask image and artery mask edge position coordinates corresponding to the artery mask edge in the artery mask image; calculating multiple distances between the multiple spinous process position coordinates and the artery mask edge position coordinates; if the minimum distance among the multiple distances is less than a preset distance, then it is determined that the spinous process corresponding to the nodule is adjacent to an artery.

[0178] In this embodiment of the disclosure, determining whether the spinous process corresponding to the nodule is adjacent to an artery based on the nodule mask and artery mask image corresponding to the chest medical image includes: extracting multiple spinous process position coordinates corresponding to multiple spinous processes and artery mask edge position coordinates corresponding to the artery mask edge in the nodule mask and artery mask image; calculating multiple distances between the multiple spinous process position coordinates and the artery mask edge position coordinates respectively; if the minimum distance among the multiple distances is less than a preset distance, then it is determined that the spinous process corresponding to the nodule is adjacent to an artery.

[0179] In the embodiments of this disclosure and other possible embodiments, determining the spike includes: detecting the edges of the nodule mask in the nodule mask image to obtain nodule mask edge lines; calculating a first pixel distance and a second pixel distance between any nodule mask edge pixel on the nodule mask edge line and its adjacent nodule mask edge pixels; if the first pixel distance and the second pixel distance are greater than a set pixel distance, then configuring the arbitrary nodule mask edge pixel as a spike.

[0180] In this disclosure and other possible embodiments, determining whether the location of the nodule corresponding to the nodule to be treated or the nodule to be operated on is adjacent to an artery based on the nodule 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 nodule edge mask position coordinates corresponding to the nodule mask image; if the minimum positional distance among the multiple positional distances is less than or equal to a preset positional distance, then the nodule location is determined to be adjacent to an artery; otherwise, the nodule location is determined to be far from an artery. The artery edge mask position coordinates and the nodule edge mask position coordinates corresponding to the nodule are respectively configured as the position coordinates of the artery edge in the artery edge mask image corresponding to the artery mask image and the position coordinates of the nodule edge mask in the nodule edge mask image corresponding to the nodule mask image.

[0181] In this embodiment and other possible embodiments, determining whether the location of the nodule corresponding to the nodule to be treated or the nodule to be operated on is adjacent to an artery based on the nodule mask and artery mask images includes: calculating multiple positional distances between the artery edge mask position coordinates corresponding to the nodule mask and the artery mask image and the nodule edge mask position coordinates; if the minimum positional distance among the multiple positional distances is less than or equal to a preset positional distance, then the nodule location is determined to be adjacent to an artery; otherwise, the nodule location is determined to be far from an artery. The artery edge mask position coordinates and the nodule edge mask position coordinates corresponding to the nodule are respectively configured as the position coordinates of the artery edge in the artery edge mask and the position coordinates of the nodule edge mask in the nodule mask and artery mask image.

[0182] 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.

[0183] In the embodiments disclosed herein and other possible embodiments, the location of nodules and their proximity to arteries are of significant clinical importance for indicating disease type, assisting in differential diagnosis, and facilitating subsequent chemotherapy and surgery. It is necessary to propose a technical solution for nodule location determination and intelligent medical image processing system to address at least one of the following technical problems: the current reliance on manual determination of nodule location and proximity to arteries leads to excessive subjectivity; the lack of intelligent determination results in a heavy workload for doctors; and it prevents effective evaluation of subsequent radiotherapy efficacy prediction.

[0184] In the embodiments disclosed herein and other possible embodiments, the step of determining the nodule mask image and arterial mask image corresponding to the nodule to be treated or the nodule to be operated on using chest medical images includes: training a first preset deep learning segmentation network using a preset number of preset location medical training images and their corresponding nodule mask training images to obtain a preset nodule segmentation model; training a second preset deep learning segmentation network using a preset number of preset location medical training images and their corresponding arterial mask training images to obtain a preset arterial segmentation model; segmenting the nodule to be treated or the nodule to be operated on in the chest medical image using the preset nodule segmentation model to obtain the corresponding nodule mask image; and segmenting the arterial vessels using the preset arterial segmentation model to obtain the corresponding arterial mask image.

[0185] In the embodiments disclosed herein and other possible embodiments, the step of using chest medical images to determine the nodule mask and arterial vessel mask images corresponding to the nodules to be treated or the nodules to be operated on includes: training a preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule mask and arterial vessel mask training images to obtain a preset nodule and arterial vessel segmentation model; and segmenting the nodules to be treated or the nodules to be operated on and the arterial vessels using the preset nodule and arterial vessel segmentation model to obtain the corresponding nodule mask and arterial vessel mask images.

[0186] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each nodule region in the nodule mask image and the nodule mask training image is configured as 1, and the mask value corresponding to the non-nodule 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.

[0187] In the embodiments disclosed herein and other possible embodiments, the mask value corresponding to each nodule region in the nodule mask and artery mask image and the nodule 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-nodule region and non-artery region is configured as 0.

[0188] In the embodiments of this disclosure and other possible embodiments, before training the first preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules in the preset number of preset site medical training images to obtain corresponding nodule mask training images.

[0189] In the embodiments of this disclosure and other possible embodiments, before training the second preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding arterial blood vessel mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate arteries and blood vessels in the preset number of preset site medical training images to obtain corresponding arterial blood vessel mask training images.

[0190] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network using a preset number of preset site medical training images and their corresponding nodule masks and artery mask training images, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules and arteries in the preset number of preset site medical training images to obtain corresponding nodule masks and artery mask training images.

[0191] 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).

[0192] In the embodiments of this disclosure and other possible embodiments, if the edge of the nodule has a spiky sign, determining whether the nodule in the chest medical image is a nodule to be treated with chemotherapy or a nodule to be operated on includes: detecting whether the nodule boundary in the nodule mask image corresponding to the chest medical image is blurred; if the nodule boundary in the nodule mask image is blurred, then the nodule is configured as a nodule to be treated with chemotherapy or a nodule to be operated on.

[0193] In the embodiments disclosed herein and other possible embodiments, if the nodule edge has a spiky sign, determining whether the nodule in the chest medical image is a nodule to be treated with chemotherapy or a nodule to be operated on includes: detecting whether the nodule boundary in the nodule and arterial mask image corresponding to the chest medical image is blurred and whether the nodule edge has a spiky sign; if the nodule boundary in the nodule and arterial mask image is blurred and the nodule edge has a spiky sign, then the nodule is configured as a nodule to be treated with chemotherapy or a nodule to be operated on.

[0194] In the embodiments of this disclosure and other possible embodiments, determining whether the nodule boundary is blurred includes: mapping the nodule location coordinates in the nodule masking image to a chest medical image; calculating multiple gradients corresponding to the nodule location coordinates in the chest medical image; if the gradient magnitude value corresponding to the multiple gradients is less than a preset gradient magnitude value, then the nodule boundary where the nodule location corresponding to the preset gradient magnitude value is located is determined to be blurred.

[0195] In the embodiments disclosed herein and other possible embodiments, if the gradient magnitude value corresponding to the plurality of gradients is greater than or equal to a preset gradient magnitude value, then the boundary of the nodule corresponding to the location of the nodule that is greater than or equal to the preset gradient magnitude value is determined to be clear.

[0196] In the embodiments of this disclosure and other possible embodiments, before mapping the nodule location coordinates in the nodule masking image to the chest medical image, the method includes: performing an erosion operation on the nodule mask in the nodule masking image corresponding to the chest medical image to obtain a nodule mask erosion image; and subtracting the nodule mask erosion image from the nodule masking image to obtain a nodule masking image.

[0197] In the embodiments disclosed herein and other possible embodiments, determining the gradient magnitude corresponding to the plurality of gradients includes: squaring the plurality of x-direction gradients corresponding to the nodule position coordinates to obtain a plurality of x-direction squared gradients; squaring the plurality of y-direction gradients corresponding to the nodule position coordinates to obtain a plurality of y-direction squared gradients; summing the plurality of x-direction squared gradients with their corresponding y-direction squared gradients to obtain a gradient sum corresponding to the plurality of nodule position coordinates; and taking the square root of the gradient sum corresponding to the plurality of nodule position coordinates to obtain the gradient magnitude corresponding to the plurality of gradients.

[0198] In the embodiments of this disclosure and other possible embodiments, the step of calculating multiple gradients corresponding to the nodule location coordinates in the chest medical image includes: determining multiple first position coordinates and multiple second position coordinates on both sides of the nodule location coordinates in the chest medical image in the x-direction, and multiple third position coordinates and multiple fourth position coordinates in the y-direction, based on the nodule location coordinates; calculating the differences between the first pixel values ​​corresponding to the multiple first position coordinates and the second pixel values ​​corresponding to the multiple second position coordinates to obtain multiple x-direction gradients corresponding to the nodule location coordinates; calculating the differences between the third pixel values ​​corresponding to the multiple third position coordinates and the fourth pixel values ​​corresponding to the multiple fourth position coordinates to obtain multiple y-direction gradients corresponding to the nodule location coordinates; and configuring the multiple x-direction gradients and their corresponding multiple y-direction gradients as multiple gradients corresponding to the nodule location coordinates.

[0199] In the embodiments disclosed herein and other possible embodiments, determining multiple first position coordinates and multiple second position coordinates corresponding to both sides of the nodule position coordinates in the x-direction in the chest medical image based on the nodule position coordinates includes: using the nodule position coordinates to determine multiple first position coordinates and multiple second position coordinates corresponding to both sides in the x-direction according to a preset distance; and / or, determining multiple third position coordinates and multiple fourth position coordinates corresponding to both sides of the nodule position coordinates in the y-direction in the chest medical image based on the nodule position coordinates includes: using the nodule position coordinates to determine multiple third position coordinates and multiple fourth position coordinates corresponding to both sides in the x-direction according to a preset distance.

[0200] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the preset distance according to actual needs. For example, the nodule location coordinates can be configured as (x1, y1), (x2, y2), ..., (x... n ,y n ), respectively determine the coordinates of multiple first positions (x1+1, y1), (x2+1, y2), ..., (x2+1, y2) in the x-direction on both sides according to the 1 pixel value corresponding to the preset distance. n +1,y n ) and multiple second position coordinates (x1-1, y1), (x2-1, y2), ..., (x n -1,y n For example, the nodule location coordinates are configured as (x1, y1), (x2, y2), ..., (x... n ,y n ), respectively determine the coordinates of multiple first positions (x1, y1+1), (x2, y2+1), ..., (x2, y2+1) in the x-direction on both sides according to the 1 pixel value corresponding to the preset distance. n ,yn +1) and multiple second position coordinates (x1, y1-1), (x2, y2-1), ..., (x n ,y n -1).

[0201] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the preset gradient magnitude value according to actual needs. For example, the preset gradient magnitude value can be configured as any value between 2 and 80 or other values.

[0202] In the embodiments of this disclosure and other possible embodiments, locating the nodule location includes: using a lung field dynamic region segmentation method to segment the chest medical image into lung fields to obtain a lung field region segmentation network corresponding to each nodule; and locating the lung field anatomical region corresponding to each nodule based on the location of each nodule and its corresponding lung field region segmentation network.

[0203] In the embodiments of this disclosure and other possible embodiments, the method of using dynamic lung field region division to divide the lung field region of the chest medical image includes: Step S101: Constructing a set of nodule locations corresponding to all nodule regions using the nodule mask image corresponding to nodules in the lung field of the chest medical image; Step S102: Traversing the area or volume corresponding to each nodule location in the set of nodule locations to determine the minimum nodule area or minimum nodule volume of all nodule locations; Step S103: Dividing the lung field region in the lung field mask image using a preset grid corresponding to the minimum grid area or minimum grid volume of all nodule locations; Step S104: Deleting the nodule location corresponding to the minimum grid area or minimum grid volume in the set of nodule locations to update the set of nodule locations; Step S105: Repeating steps S102-S104 until the lung field region division is completed using all nodule locations in the set of nodule locations. To address at least one of the following technical problems: excessive subjectivity in nodule location, lack of intelligent nodule location, resulting in a huge workload for doctors and preventing further prediction of lymph node metastasis in partially solid lung adenocarcinoma.

[0204] In the embodiments disclosed herein and other possible embodiments, each grid in the preset grid predefines a lung field anatomical region; the location of each nodule is configured as the coordinate point corresponding to each nodule in a chest medical image.

[0205] For example, in the embodiments of this disclosure and other possible embodiments, if the chest medical image is configured as a two-dimensional chest medical image, then the coordinate point of each nodule in the chest medical image is a two-dimensional coordinate (x, y); if the chest medical image is configured as a three-dimensional chest medical image, then the coordinate point of each nodule in the chest medical image is a two-dimensional coordinate (x, y, z).

[0206] In the embodiments disclosed herein and other possible embodiments, the two-dimensional chest medical image can be configured as a chest X-ray image or a chest ultrasound image; the three-dimensional chest medical image can be configured as one or more of a chest CT image, a chest MRI image, a chest PET image, a chest CT-PET image, and a chest MRI-PET image. Furthermore, the two-dimensional chest medical image can also be configured as one or more corresponding chest projection images of a chest CT image, a chest MRI image, a chest PET image, a chest CT-PET image, and a chest MRI-PET image.

[0207] In the embodiments disclosed herein and other possible embodiments, if the chest medical image is configured as a two-dimensional chest medical image, the area corresponding to each nodule location in the nodule location set is traversed to determine the minimum nodule area of ​​all nodule locations; a preset grid corresponding to the minimum grid area of ​​all nodule locations is used to divide the lung field region in the lung field mask image; the nodule location corresponding to the minimum grid area in the nodule location set is deleted to update the nodule location set; and / or, if the chest medical image is configured as a three-dimensional chest medical image, the area or volume corresponding to each nodule location in the nodule location set is traversed to determine the minimum nodule volume of all nodule locations; a preset grid corresponding to the minimum grid volume of all nodule locations is used to divide the lung field region in the lung field mask image; the nodule location corresponding to the minimum grid volume in the nodule location set is deleted to update the nodule location set.

[0208] For example, in embodiments of this disclosure and other possible embodiments, the predefined lung field anatomical regions can be configured to longitudinally divide the lungs into upper, middle, and lower lung fields, with the lower borders of the 2nd and 4th anterior ribs as boundaries, and to transversely divide each lung field vertically into an inner zone (near the mediastinum), a middle zone, and an outer zone (near the chest wall). The lung field anatomical regions can also be configured as the apical region corresponding to the area surrounded by the outer border of the 1st rib, and the subclavian region corresponding to the area between the lower border of the clavicle and the lower border of the 2nd anterior rib. The predefined lung field anatomical regions can also be configured as the upper, middle, and lower lobes of the right lung and the upper and lower lobes of the left lung; wherein the upper lobe of the left lung may include the lingular segment. The predefined lung field anatomical regions can be configured as lung segments corresponding to independent distribution units of bronchi and blood vessels, each lung segment having independent bronchial and vascular bundles; the upper lobe corresponding to the right lung may include the apical segment, posterior segment, and anterior segment; the middle lobe corresponding to the right lung may include the lateral segment and medial segment; the lower lobe corresponding to the right lung may include the dorsal segment, medial basal segment, anterior basal segment, lateral basal segment, and posterior basal segment; the upper lobe corresponding to the left lung may include the apical-posterior segment, anterior segment, superior lingular segment, and inferior lingular segment; the lower lobe corresponding to the left lung may include the dorsal segment, medial anterior basal segment, lateral basal segment, and posterior basal segment.

[0209] In the embodiments disclosed herein and other possible embodiments, the predefined lung field anatomical regions can be combined arbitrarily without violating the anatomical structure of the lung field.

[0210] In the embodiments of this disclosure and other possible embodiments, determining the nodule mask image corresponding to nodules in the lung field of a chest medical image includes: performing nodule segmentation on the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network, and determining the nodule mask image of the chest medical image.

[0211] In the embodiments of this disclosure and other possible embodiments, before segmenting the chest medical image into nodules using the preset segmentation model constructed based on a deep learning segmentation network, the method includes: training the preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain the preset nodule segmentation model constructed based on the deep learning segmentation network.

[0212] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate nodules on a preset number of chest medical images to obtain corresponding nodule mask training images.

[0213] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate nodules on the preset number of chest medical images to obtain corresponding nodule mask training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500.

[0214] In the embodiments disclosed herein and other possible embodiments, the 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).

[0215] In this disclosure and other possible embodiments, the method further includes: determining whether each nodule is adjacent to a pulmonary artery.

[0216] In the embodiments disclosed herein and other possible embodiments, determining whether each nodule is adjacent to a pulmonary artery includes: calculating multiple positional distances between the pulmonary artery edge mask position coordinates corresponding to the pulmonary artery mask image of the chest medical image and the nodule edge mask position coordinates corresponding to the nodule mask image; if the minimum positional distance among the multiple positional distances is less than or equal to a preset positional distance, then the nodule position is determined to be adjacent to a pulmonary artery; otherwise, the nodule position is determined to be far from a pulmonary artery.

[0217] In the embodiments of this disclosure and other possible embodiments, determining the pulmonary artery mask image of the chest medical image includes: segmenting the chest medical image into pulmonary arteries using a preset pulmonary artery segmentation model constructed based on a deep learning segmentation network, and determining the pulmonary artery mask image of the chest medical image.

[0218] In the embodiments of this disclosure and other possible embodiments, before segmenting the chest medical image into pulmonary arteries using a preset segmentation model constructed based on a deep learning segmentation network, the method includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding pulmonary artery mask training images to obtain a preset pulmonary artery segmentation model constructed based on the deep learning segmentation network. In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate the pulmonary artery vessels in a preset number of chest medical images to obtain corresponding pulmonary artery vessel mask training images.

[0219] In the embodiments disclosed herein and other possible embodiments, the pulmonary artery edge mask position coordinates and the nodule edge mask position coordinates corresponding to the nodule are respectively configured as the position coordinates of the pulmonary artery edge in the pulmonary artery edge mask image corresponding to the pulmonary artery mask image and the position coordinates of the nodule edge mask in the nodule edge mask image corresponding to the nodule mask image.

[0220] In the embodiments of this disclosure and other possible embodiments, before training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding pulmonary artery vascular mask training images to obtain a preset segmentation model constructed based on the deep learning segmentation network, the method includes: using segmentation annotation software CasiaLabeler or Labelme to delineate the pulmonary arteries in the preset number of chest medical images to obtain corresponding pulmonary artery vascular mask training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500.

[0221] Similarly, in the embodiments disclosed herein and other possible embodiments, the 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).

[0222] 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.

[0223] In the embodiments disclosed herein and other possible embodiments, the location of nodules and their proximity to pulmonary arteries are of significant clinical importance for indicating disease type, assisting in differential diagnosis, and subsequent chemotherapy and surgery. This addresses at least one of the technical problems currently present, such as the need for manual determination of nodule location and proximity to pulmonary arteries, which is too subjective, lacks intelligent determination of nodule location and proximity to pulmonary arteries, resulting in a huge workload for doctors and preventing further surgery, as well as the prediction of lymph node metastasis in some solid lung adenocarcinomas.

[0224] In this embodiment of the disclosure, determining the nodule lobulation sign includes: performing an erosion operation on the nodule mask image corresponding to each nodule in the chest medical image to obtain a nodule mask erosion image; subtracting the corresponding nodule mask erosion image from the nodule mask image of each nodule to obtain a nodule mask processing image corresponding to each nodule; judging the nodule morphology in the nodule mask processing image corresponding to each nodule; if the nodule morphology is irregular, determining whether the nodule lobulation sign exists based on the nodule mask processing image corresponding to each nodule.

[0225] In the embodiments of this disclosure and other possible embodiments, the method for determining the nodule lobulation sign includes: determining a nodule mask processing image corresponding to each nodule in a nodule mask image of a chest medical image; judging the nodule morphology in the nodule mask processing image corresponding to each nodule; if the nodule morphology is irregular, determining whether the nodule lobulation sign exists based on the nodule mask processing image corresponding to each nodule.

[0226] In the embodiments of this disclosure and other possible embodiments, the step of determining the nodule morphology in the nodule mask processing image corresponding to each nodule includes: extracting multiple nodule mask position coordinates corresponding to the nodule mask lines in the nodule mask processing image corresponding to each nodule; fitting or reconstructing the multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology corresponding to each nodule; and determining whether the nodule morphology is regular based on the nodule morphology corresponding to each nodule.

[0227] In the embodiments disclosed herein and other possible embodiments, the step of processing the image based on the nodule mask corresponding to each nodule to determine whether there is nodule lobulation feature includes: extracting the nodule mask line corresponding to each nodule respectively; if the shape of the nodule mask line is an arc-shaped protrusion connected together, then the corresponding nodule has nodule lobulation feature; otherwise, nodule lobulation feature exists.

[0228] In the embodiments disclosed herein and other possible embodiments, the step of processing the image based on the nodule mask corresponding to each nodule to determine whether there is nodule lobulation feature further includes: segmenting the nodule mask line according to its concave part to obtain multiple arc-shaped convex line segments; determining whether the arc-shaped convex is a tip based on the slope corresponding to each pixel in the multiple arc-shaped convex line segments; if it is a tip, the corresponding nodule does not have nodule lobulation feature and the nodule has spur feature; if there is no tip, the corresponding nodule has nodule lobulation feature.

[0229] In the embodiments disclosed herein and other possible embodiments, determining whether the arc-shaped protrusion is a tip based on the slope corresponding to each pixel point in the multiple arc-shaped protrusion segments includes: mapping the multiple arc-shaped protrusion segments to a Cartesian coordinate system, wherein the coordinate point corresponding to the arc-shaped protrusion of the multiple arc-shaped protrusion segments is the highest pixel point of the arc-shaped protrusion segment in the Cartesian coordinate system; calculating the left derivative and right derivative corresponding to the highest pixel point; if the difference between the left derivative and the right derivative is less than or equal to a set difference, then the arc-shaped protrusion is determined to be a tip; otherwise, the arc-shaped protrusion is determined to be a non-tip.

[0230] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the set difference value as needed. For example, the set difference value can be configured as 1, 2, or other possible values.

[0231] In the embodiments of this disclosure and other possible embodiments, the step of fitting or reconstructing the multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology includes: under the optimal criterion, approximating multiple preset closed curve equations or multiple preset closed surface equations using the multiple nodule mask position coordinates corresponding to each nodule to determine the nodule morphology corresponding to each nodule.

[0232] In the embodiments disclosed herein and other possible embodiments, the step of approximating a preset closed curve equation or a preset closed surface equation using the multiple nodule mask position coordinates corresponding to each nodule under the optimal criterion to determine the nodule morphology corresponding to each nodule includes: under the optimal criterion, minimizing the error or absolute residual of the nodule mask position coordinates between the approximation equations and the multiple near-preset closed curve equations or multiple preset closed surface equations when the multiple nodule mask position coordinates corresponding to each nodule approximate the multiple preset closed curve equations or multiple preset closed surface equations, thereby determining the nodule morphology corresponding to each nodule.

[0233] In the embodiments of this disclosure and other possible embodiments, before determining the nodule mask processing image corresponding to the nodule in the nodule mask image of the chest medical image, the method includes: using a preset nodule segmentation model constructed based on a deep learning segmentation network to segment the chest medical image into nodules, and determining the nodule mask image of the chest medical image.

[0234] In the embodiments of this disclosure and other possible embodiments, before segmenting chest medical images into nodules using a preset segmentation model constructed based on a deep learning segmentation network, the method includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset nodule segmentation model constructed based on the deep learning segmentation network.

[0235] In this embodiment, determining the nodule vacuolation sign includes: performing nodule vacuolation sign target detection on each nodule in a chest medical image, configuring the nodules corresponding to detected nodule vacuolation signs as having nodule vacuolation signs; extracting the average pixel value of the nodule edge corresponding to the nodule for which nodule vacuolation signs were not detected; if the average pixel value corresponding to the nodule for which nodule vacuolation signs were not detected is less than the corresponding average pixel value of the nodule edge, then it is determined that the nodule for which nodule vacuolation signs were not detected has nodule vacuolation signs; otherwise, it is determined that the nodule for which nodule vacuolation signs were not detected does not have nodule vacuolation signs. This addresses at least one of the technical problems in existing methods, such as the need for manual determination of nodule vacuolation, which is too subjective, lacks intelligent determination of nodule vacuolation, resulting in a huge workload for doctors, and prevents further surgical procedures and prediction of lymph node metastasis in some solid lung adenocarcinomas.

[0236] In the embodiments of this disclosure and other possible embodiments, unlike the prior art, the technical solution proposed in this disclosure, by performing target detection of nodule vacuolation sign on each nodule in chest medical images, and using the average pixel value of the nodule edge corresponding to the nodule where nodule vacuolation sign was not detected, further processing of the nodules where nodule vacuolation sign was not detected, can effectively solve the technical problem of missed detection of nodule vacuolation sign. By using the average pixel value of the nodule edge corresponding to the nodule where nodule vacuolation sign was not detected, the robustness of the method for determining whether nodule where nodule vacuolation sign was not detected is improved to a certain extent. At the same time, it avoids at least one of the technical problems of existing methods that require manual determination of the existence of nodule vacuolation, which is too subjective and lacks intelligent determination of nodule vacuolation, resulting in a huge workload for doctors and preventing further surgery, lymph node metastasis prediction in some solid lung adenocarcinomas.

[0237] In the embodiments of this disclosure and other possible embodiments, the nodule cavitation target detection for each nodule in the chest medical image includes: using a deep learning-based nodule cavitation target detection model to perform nodule cavitation target detection for each nodule in the chest medical image.

[0238] In the embodiments of this disclosure and other possible embodiments, before performing nodule cavitation target detection on each nodule in the chest medical image using a deep learning-based nodule cavitation target detection model, the method includes: training a preset deep learning target detection network to obtain a deep learning-based nodule cavitation target detection model.

[0239] In the embodiments of this disclosure and other possible embodiments, training a preset deep learning target detection network to obtain a deep learning-based target detection model for nodule cavitation includes: training the preset deep learning target detection network using a preset number of chest medical training images and their corresponding nodule cavitation target training images to obtain a deep learning-based target detection model for nodule cavitation.

[0240] In the embodiments disclosed herein and other possible embodiments, before training the preset deep learning target detection network, the method includes: using segmentation and annotation software to delineate nodular vacuolation sign bounding boxes on a preset number of chest medical images to obtain corresponding nodular vacuolation sign target training images. For example, annotation software such as CasiaLabeler or Labelme can be used to delineate nodular vacuolation sign bounding boxes on a preset number of chest medical images to obtain corresponding nodular vacuolation sign target training images. Those skilled in the art can configure the preset number according to actual needs. For example, the preset number can be configured as 300-500 or other possible values.

[0241] In the embodiments disclosed herein and other possible embodiments, the preset deep learning object detection network may be configured as a YOLO series or an object detection network based on YOLO.

[0242] In the embodiments of this disclosure and other possible embodiments, determining the total number of nodules corresponding to the chest medical image includes: performing nodule segmentation or nodule target detection on the chest medical image to determine the total number of nodules corresponding to the chest medical image; and / or, performing nodule segmentation on the chest medical image includes: performing nodule segmentation on the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network; and / or, performing nodule target detection on the chest medical image includes: performing nodule target detection on the chest medical image using a nodule target detection model based on deep learning.

[0243] In the embodiments of this disclosure and other possible embodiments, a preset nodule segmentation model based on a deep learning segmentation network is used to segment the chest medical image into nodules to obtain corresponding nodule mask images; the number of nodule masks corresponding to the nodule mask images is counted to determine the total number of nodules corresponding to the chest medical image.

[0244] In the embodiments of this disclosure and other possible embodiments, a nodule target detection model based on deep learning is used to detect nodules in the chest medical image to obtain corresponding nodule detection images; the number of nodule detection boxes on the nodule detection images is counted to determine the total number of nodules corresponding to the chest medical image.

[0245] In the embodiments of this disclosure and other possible embodiments, before performing nodule segmentation on the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network, the process includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding nodule mask training images to obtain a preset nodule segmentation model constructed based on the deep learning segmentation network.

[0246] In the embodiments of this disclosure and other possible embodiments, before training the preset deep learning segmentation network, the method includes: using segmentation annotation software to delineate nodules on a preset number of chest medical images to obtain corresponding nodule mask training images.

[0247] In the embodiments of this disclosure and other possible embodiments, the step of extracting the pixel values ​​of the nodule edges corresponding to nodules for which nodule cavitation was not detected includes: performing edge detection on the nodule mask image corresponding to the nodule for which nodule cavitation was not detected to obtain a nodule edge image; performing pixel-level multiplication on the nodule edge image and its corresponding chest medical image to obtain a nodule edge image; and extracting the pixel values ​​corresponding to the nodule edges in the nodule edge image to obtain the corresponding nodule edge pixel values.

[0248] In the embodiments of this disclosure and other possible embodiments, the step of performing edge detection on the nodule mask image corresponding to the nodule where nodule cavitation was not detected to obtain a nodule edge image includes: performing an erosion operation on the nodule mask image corresponding to the nodule where nodule cavitation was not detected to obtain a nodule mask erosion image; and subtracting the corresponding nodule mask erosion image from the nodule mask image to obtain a nodule edge image.

[0249] In the embodiments of this disclosure and other possible embodiments, the step of extracting the average pixel value of the nodule edge corresponding to the nodule for which nodule cavitation was not detected includes: performing edge detection on the nodule mask image corresponding to the nodule for which nodule cavitation was not detected to obtain a nodule edge image; performing pixel-level multiplication on the nodule edge image and its corresponding chest medical image to obtain a nodule edge image; extracting the pixel value corresponding to the nodule edge in the nodule edge image to obtain the corresponding nodule edge pixel value; and calculating the average value of the nodule edge pixel value to obtain the average pixel value of the nodule edge.

[0250] In the embodiments of this disclosure and other possible embodiments, the step of performing edge detection on the nodule mask image corresponding to the nodule where nodule cavitation was not detected to obtain a nodule edge image includes: performing an erosion operation on the nodule mask image corresponding to the nodule where nodule cavitation was not detected to obtain a nodule mask erosion image; and subtracting the corresponding nodule mask erosion image from the nodule mask image to obtain a nodule edge image.

[0251] In the embodiments of this disclosure and other possible embodiments, determining whether a nodule without detected nodule cavitation exists based on the average pixel value corresponding to the nodule without detected nodule cavitation and the corresponding average pixel value of the nodule edge includes: if the average pixel value corresponding to the nodule without detected nodule cavitation is less than the corresponding average pixel value of the nodule edge, then it is determined that the nodule without detected nodule cavitation exists; otherwise, it is determined that the nodule without detected nodule cavitation does not exist.

[0252] In this embodiment of the disclosure, the method further includes: acquiring preset clinical information corresponding to the chest medical image at the first time point; using preset quantitative imaging features of nodules corresponding to partial solid lung adenocarcinoma in the chest medical image at the first time point, the preset clinical information, and the pathological results of lymph nodes in partial solid lung adenocarcinoma corresponding to metastasis or non-metastasis at the second time point after the first time point, training a preset predictor to construct a lymph node metastasis prediction model; and using the lymph node metastasis prediction model, the preset quantitative imaging features of nodules corresponding to partial solid lung adenocarcinoma in the chest medical image to be predicted, and the preset clinical information, predicting lymph node metastasis in partial solid lung adenocarcinoma.

[0253] In this embodiment of the disclosure, determining the preset clinical information corresponding to the chest medical image at the first moment includes: acquiring candidate clinical information corresponding to the chest medical image at the first moment; and using a univariate logistic regression analysis algorithm, the candidate clinical information and its corresponding lymph node pathology results in the partial solid lung adenocarcinoma, performing univariate logistic regression on each candidate clinical information feature to determine the preset clinical information related to the lymph node pathology results in the partial solid lung adenocarcinoma. In this embodiment of the disclosure, the clinical information to be selected is configured as one or more of age, gender, smoking history, and history of lung cancer. Prior to training the preset predictor, the process includes: standardizing or normalizing the numerical clinical information to be selected, and encoding the categorized clinical information to be selected.

[0254] In the embodiments of this disclosure and other possible embodiments, for example, before training the preset predictor, the age corresponding to the clinical information to be selected is standardized or normalized, and one or more of the gender, smoking history, and history of lung cancer corresponding to the categorized clinical information to be selected are encoded.

[0255] In this disclosure and other possible embodiments, for example, encoding processing is performed on one or more of the following clinical information to be selected by category: gender, smoking history, and presence or absence of a history of lung cancer. This includes: configuring a first value for males; configuring a second value for females; configuring a first value for those with a smoking history; configuring a second value for those without a smoking history; configuring a first value for those with a history of lung cancer; and configuring a second value for those with a history of lung cancer.

[0256] In the embodiments disclosed herein and other possible embodiments, those skilled in the art can configure the first value and the second value according to actual needs. For example, the first value and the second value can be configured as 1 and 0, respectively.

[0257] The execution entity for the pleural mask image determination and intelligent medical image analysis method can be an image processing device. 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, in-vehicle device, wearable device, etc. In some possible implementations, the pleural mask image determination and intelligent medical image analysis method can be implemented by a processor calling computer-readable instructions stored in memory.

[0258] Those skilled in the art will understand that in the above-described method for determining pleural mask images and intelligent analysis of medical images 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 by its function and possible internal logic.

[0259] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: an expansion unit, configured to expand a lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain an expanded lung field mask image; and a subtraction unit, configured to perform pixel-level subtraction operations on the expanded lung field mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura.

[0260] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: 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 above-described pleural mask image determination method.

[0261] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: a computer-readable storage medium storing a computer program / instructions and a bit stream thereon, wherein the computer program / instructions, when executed by a processor, implement the above-described pleural mask image determination method to generate the bit stream.

[0262] According to one aspect of this disclosure, a pleural mask image determination system is provided, comprising: a computer program product configured with a computer program / instruction that, when executed by a processor, implements the aforementioned pleural mask image determination method.

[0263] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: an expansion unit, configured to expand the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain an expanded lung field mask image; a subtraction unit, configured to perform pixel-level subtraction operations on the expanded lung field mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura; a segmentation unit, configured to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image; and a first determination unit, configured to determine the nearest pleural region image to each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image.

[0264] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: an expansion unit, configured to expand the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain a lung field expansion mask image; a subtraction unit, configured to perform pixel-level subtraction operations on the lung field expansion mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura; a segmentation unit, configured to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image; a first determination unit, configured to determine the nearest pleural region image of each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image; and a second determination unit, configured to determine whether the chest medical image contains a nodular pleural traction sign using the nearest nodular pleural region image of each nodule.

[0265] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: an expansion unit, configured to expand the lung field in a lung field mask image corresponding to a chest medical image using a second preset edge width to obtain a lung field expansion mask image; a subtraction unit, configured to perform pixel-level subtraction operations on the lung field expansion mask image and the lung field mask image to determine a pleural mask image corresponding to the pleura; a segmentation unit, configured to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image; a first determination unit, configured to determine the nearest pleural region image of each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image; a second determination unit, configured to determine whether the chest medical image contains a nodular pleural traction sign using the nearest nodular pleural region image of each nodule; and a prediction unit, configured to predict nodule-related preset chest diseases based on quantitative imaging features of preset nodules corresponding to the presence or absence of nodular pleural traction sign.

[0266] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: 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 aforementioned medical image intelligent analysis method.

[0267] According to one aspect of this disclosure, a medical image intelligent analysis system is provided, comprising: a computer-readable storage medium storing a computer program / instructions and a bit stream thereon, wherein the computer program / instructions, when executed by a processor, implement the above-described medical image intelligent analysis method to generate the bit stream.

[0268] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the pleural mask image determination and medical image intelligent analysis method described in the above method embodiments. The specific implementation can be referred to the description of the above pleural mask image determination and medical image intelligent analysis method embodiments, which will not be repeated here for the sake of brevity.

[0269] This disclosure also proposes a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the aforementioned pleural mask image determination and intelligent medical image analysis method. The computer-readable storage medium can be a non-volatile computer-readable storage medium.

[0270] This disclosure also proposes an electronic device, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured for the aforementioned pleural mask image determination and medical image intelligent analysis method. The electronic device can be provided as a terminal, a server, or other form of device.

[0271] This disclosure also proposes a computer program product, including a computer program / instruction, which, when executed by a processor, implements the aforementioned pleural mask image determination and medical image intelligent analysis method.

[0272] Figure 3 This is a block diagram illustrating an electronic device 800 according to an exemplary embodiment. For example, the electronic device 800 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, or other terminal.

[0273] Reference Figure 3 The electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0274] Processing component 802 typically controls the overall operation of electronic device 800, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 802 may include one or more modules to facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.

[0275] Memory 804 is configured to store various types of data to support the operation of electronic device 800. Examples of such data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, etc. Memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0276] Power supply component 806 provides power to various components of electronic device 800. Power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800.

[0277] Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 808 includes a front-facing camera and / or a rear-facing camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0278] Audio component 810 is configured to output and / or input audio signals. For example, audio component 810 includes a microphone (MIC) configured to receive external audio signals when electronic device 800 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 804 or transmitted via communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

[0279] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0280] Sensor assembly 814 includes one or more sensors for providing state assessments of various aspects of electronic device 800. For example, sensor assembly 814 can detect the on / off state of electronic device 800, the relative positioning of components such as the display and keypad of electronic device 800, changes in position of electronic device 800 or a component of electronic device 800, the presence or absence of user contact with electronic device 800, orientation or acceleration / deceleration of electronic device 800, and temperature changes of electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 814 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0281] Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices. Electronic device 800 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 816 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 816 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0282] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0283] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions that can be executed by a processor 820 of an electronic device 800 to perform the above-described method.

[0284] Figure 4 This is a block diagram illustrating an electronic device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. (Refer to...) Figure 4 The electronic device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions, such as application programs, that can be executed by the processing component 1922. The application programs stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 1922 is configured to execute instructions to perform the methods described above.

[0285] Electronic device 1900 may also include a power supply component 1926 configured to perform power management of electronic device 1900, a wired or wireless network interface 1950 configured to connect electronic device 1900 to a network, and an input / output (I / O) interface 1958. Electronic device 1900 can operate on an operating system stored in memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0286] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions that can be executed by a processing component 1922 of an electronic device 1900 to perform the above-described method.

[0287] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0288] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example, but not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0289] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0290] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0291] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0292] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0293] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0294] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0295] 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 determining a pleural mask image, characterized in that, include: The lung field in the lung field mask image corresponding to the chest medical image is expanded using the second preset edge width to obtain the lung field expanded mask image; A pixel-level subtraction operation is performed on the lung field expansion mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura.

2. The method for determining a pleural mask image according to claim 1, characterized in that, Determining the lung field mask image corresponding to the chest medical image includes: segmenting the lung field in the chest medical image to obtain the lung field mask image corresponding to the chest medical image; and / or, segmenting the lung field in the chest medical image to obtain the lung field mask image corresponding to the chest medical image includes: acquiring a preset lung field segmentation model constructed based on a deep learning segmentation network; segmenting the lung field in the chest medical image using the preset lung field segmentation model to obtain the lung field mask image corresponding to the chest medical image; and / or, before segmenting the lung field in the chest medical image using the preset lung field segmentation model, it includes: training a preset deep learning segmentation network using a preset number of chest medical images and their corresponding lung field mask training images to obtain a preset lung field segmentation model constructed based on the deep learning segmentation network.

3. A medical image intelligent analysis method, characterized in that, include: Using the pleural mask image determination method as described in any one of claims 1 or 2, a pleural mask image corresponding to the pleura in the chest medical image is determined; Using a dynamic lung field region segmentation method, the chest medical image is segmented into lung fields to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image. Based on the lung field region segmentation network corresponding to each nodule and the pleural mask image, the nearest pleural region image for each nodule is determined.

4. The intelligent medical image analysis method according to claim 3, characterized in that, Determining each nodule in the chest medical image includes: segmenting the nodules in the chest medical image to determine each nodule in the chest medical image; and / or, segmenting the nodules in the chest medical image to determine each nodule in the chest medical image includes: segmenting the nodules in the chest medical image using a preset nodule segmentation model constructed based on a deep learning segmentation network to obtain a nodule mask image corresponding to the chest medical image; and determining each nodule in the chest medical image based on the nodule mask image.

5. The intelligent medical image analysis method according to any one of claims 3 or 4, characterized in that, It also includes: using the nearest nodular pleural region image for each nodule to determine whether the chest medical image contains a nodular pleural traction sign; and / or, the step of using the nearest nodular pleural region image for each nodule to determine whether the chest medical image contains a nodular pleural traction sign includes: using a preset feature target detection model based on deep learning to perform feature target detection on the nodular pleural region image to determine whether there are cord-like or linear features between the pleura adjacent to the nodule at the preset position in the nodular pleural region image; if so, it is determined that the chest medical image contains a nodular pleural traction sign; otherwise, it is determined that the chest medical image does not contain a nodular pleural traction sign.

6. The intelligent medical image analysis method according to claim 5, characterized in that, Based on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of the nodular pleural traction sign, a preset chest disease related to nodules is predicted; and / or, the prediction of preset chest diseases related to nodules based on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of the nodular pleural traction sign includes: performing feature selection on the quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of the nodular pleural traction sign to obtain preset nodule quantitative imaging features corresponding to preset chest diseases; and predicting preset chest diseases related to nodules based on the preset nodule quantitative imaging features corresponding to preset chest diseases.

7. The intelligent medical image analysis method according to claim 6, characterized in that, The method of predicting nodule-related pre-defined chest diseases based on the pre-defined quantitative imaging features of pre-defined nodules corresponding to the pre-defined chest diseases includes: training a pre-defined predictor and constructing a pre-defined chest disease model using the pre-defined quantitative imaging features of pre-defined nodules corresponding to the pre-defined chest diseases in the chest medical image at a first time and the results of the pre-defined chest diseases at a second time after the first time; and / or, the pre-defined quantitative imaging features to be selected further include one or more of the following: nodule long axis, whether the nodule edge is clear, nodule morphology, nodule spinous process sign, nodule lobulation sign, and nodule vacuolation sign; and / or, detecting the nodule spinous process sign includes: calculating the ratio of the long axis to the short axis of each nodule based on the long axis and short axis of each nodule in the nodule mask image corresponding to the chest medical image; wherein, the chest medical image is configured as a two-dimensional medical image; if the ratio of the long axis to the short axis is greater than or equal to a first pre-defined ratio of the long axis to the short axis, then it is determined that the corresponding nodule has a spinous process sign.

8. A pleural mask image determination system, characterized in that, include: The expansion unit is used to expand the lung field in the lung field mask image corresponding to the chest medical image using a second preset edge width to obtain the lung field expansion mask image; the subtraction unit is used to perform pixel-level subtraction operation on the lung field expansion mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura. 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 pleural mask image determination method of claim 1 or 2; or, a computer-readable storage medium having a computer program / instructions and a bitstream stored thereon, wherein the computer program / instructions, when executed by a processor, implement the pleural mask image determination method of claim 1 or 2 to generate the bitstream; or, a computer program product having a computer program / instructions configured to implement the pleural mask image determination method of claim 1 or 2 when executed by a processor.

9. A medical image intelligent analysis system, characterized in that, include: The expansion unit is used to expand the lung field in the lung field mask image corresponding to the chest medical image using the second preset edge width to obtain the lung field expansion mask image. The subtraction unit is used to perform pixel-level subtraction operations on the lung field expansion mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura. A segmentation unit is used to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image. The first determining unit is used to determine the nearest pleural region image for each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image. or, The expansion unit is used to expand the lung field in the lung field mask image corresponding to the chest medical image using a second preset edge width to obtain the lung field expansion mask image; the subtraction unit is used to perform pixel-level subtraction operation on the lung field expansion mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura. A segmentation unit is used to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image. The first determining unit is used to determine the nearest pleural region image for each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image. The second determining unit is used to determine whether the chest medical image contains a nodular pleural traction sign by using the nearest nodular pleural region image for each nodule.

10. A medical image intelligent analysis system, characterized in that, include: The expansion unit is used to expand the lung field in the lung field mask image corresponding to the chest medical image using the second preset edge width to obtain the lung field expansion mask image. The subtraction unit is used to perform pixel-level subtraction operations on the lung field expansion mask image and the lung field mask image to determine the pleural mask image corresponding to the pleura. A segmentation unit is used to segment the lung field region of the chest medical image to obtain a lung field region segmentation network corresponding to each nodule in the chest medical image. The first determining unit is used to determine the nearest pleural region image for each nodule based on the lung field region segmentation network corresponding to each nodule and the pleural mask image. The second determining unit is used to determine whether the chest medical image contains a nodular pleural traction sign by using the image of the nearest nodular pleural region to each nodule; The prediction unit is used to predict nodule-related preset chest diseases based on quantitative imaging features of the candidate preset nodules corresponding to the presence or absence of nodule pleural traction signs. 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 intelligent medical image analysis method according to any one of claims 3-7; or, it comprises: 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 intelligent medical image analysis method according to any one of claims 3-7 to generate the bit stream; or, it comprises: a computer program product having a computer program / instructions configured thereon, wherein the computer program / instructions, when executed by a processor, implement the intelligent medical image analysis method according to any one of claims 3-7.

Images