Scanning image processing method, electronic device and readable medium

Abstract

The present disclosure provides a scanning image processing method, electronic device, and readable medium, the method comprising: acquiring a scanning image; inputting the scanning image into a lesion segmentation model to obtain a segmented lesion region; based on the similarity to the lesion region features, performing regional expansion outward from the lesion region to determine a region surrounding the lesion; and obtaining the lesion region and the region surrounding the lesion. The present disclosure can accurately extract the lesion region and other regions from CT and other scanning images, and effectively determine the region surrounding the lesion using an adaptive radial expansion method, and then effectively predict the lesion infiltration grade using the lesion determination results, thereby providing doctors and other relevant personnel with more accurate tumor infiltration analysis and prediction, improving the efficiency and accuracy of evaluation and prediction, and thereby effectively improving the doctor's work efficiency and differential diagnosis level.

Description

Technical Field

[0001] The present disclosure relates to the technical field of image processing, and in particular to a scanning image processing method, an electronic device, and a readable medium. Background Art

[0002] The incidence and mortality rates of lung cancer rank first in the world. Early detection and diagnosis of lung cancer are the key to lung cancer prevention and treatment and improving survival rates. The most common type of primary lung cancer is adenocarcinoma, which is divided into pre-invasive lesions, microinvasive and invasive adenocarcinoma according to the severity. Accurate assessment of the aggressiveness of adenocarcinoma directly determines the choice of treatment plan and prognosis assessment. The environment surrounding the tumor can secrete a large number of growth factors and cytokines, inducing hypoxia and angiogenesis, and plays an important role in the occurrence, development and metastasis of the tumor. The World Health Organization defines spread through airspaces (STAS) as a special form of metastasis of lung adenocarcinoma. STAS refers to the discovery of tumor cells in the airway cavity of the lung parenchyma in the area surrounding the tumor, and is an independent risk factor for postoperative recurrence and metastasis in patients with lung cancer.

[0003] While most studies focus on feature extraction or analysis of the tumor region, others use radial expansion of a fixed distance to determine the peritumoral extent, incorporating imaging features of the tumor and surrounding area to aid diagnosis. Integrating peritumoral imaging information can more comprehensively characterize tumor invasiveness, enabling intelligent identification of early-stage lung adenocarcinoma and prediction of invasiveness, providing a theoretical basis for surgical treatment selection and prognosis assessment.

[0004] Currently, differential diagnosis of lung cancer invasiveness is primarily performed through manual interpretation, with radiologists reviewing both plain and enhanced CT scans of patients. However, this requires repeated review of each CT image, which is time-consuming and tedious. The diagnostic results are highly subjective, resulting in low efficiency and accuracy in assessment and prediction. Summary of the Invention

[0005] The main purpose of the present disclosure is to provide a scanning image processing method, an electronic device and a readable medium to improve the above-mentioned defects in the prior art.

[0006] The present disclosure solves the above technical problems through the following technical solutions:

[0007] As one aspect of the present disclosure, a scanned image processing method is provided, comprising:

[0008] Acquire a scanned image;

[0009] Inputting the scanned image into a lesion segmentation model to obtain a segmented lesion area;

[0010] Based on the similarity with the features of the lesion area, expanding the area outward from the lesion area to determine the area surrounding the lesion; and

[0011] The lesion area and the area surrounding the lesion are obtained.

[0012] As an optional implementation manner, after the step of obtaining the lesion area and the area surrounding the lesion, the image processing method further includes:

[0013] The lesion infiltration grade is evaluated based on the image including the lesion area and the area surrounding the lesion to output a lesion infiltration grade result.

[0014] As an optional embodiment, the step of evaluating the lesion infiltration grade based on the image including the lesion area and the area surrounding the lesion to output the lesion infiltration grade result includes:

[0015] Extracting radiomic features of the lesion area and the area surrounding the lesion from an image including the lesion area and the area surrounding the lesion;

[0016] Inputting the image from which the radiomic features of the lesion area and the area surrounding the lesion are extracted into a lesion invasiveness grading model to predict the lesion invasiveness grading;

[0017] Output the lesion invasiveness grading results.

[0018] As an optional embodiment, the step of evaluating the lesion infiltration grade based on the image including the lesion area and the area surrounding the lesion to output the lesion infiltration grade result includes:

[0019] Input the image including the lesion area and the area surrounding the lesion into the deep neural network classification model to output the lesion invasiveness grading result;

[0020] The training steps of the deep neural network classification model specifically include:

[0021] Inputting a training image including the lesion area and the area surrounding the lesion into a deep neural network classification model to be trained;

[0022] Extracting an attention region from the training image and extracting the same attention region from the supervision image, wherein the attention region includes at least the lesion region and the lesion surrounding region;

[0023] Calculating a graded loss of lesion infiltration between the training image and the supervisory image based on the lesion infiltration of pixels in the attention area of ​​the training image and the supervisory image;

[0024] The deep neural network classification model to be trained is trained according to the hierarchical loss to obtain a trained deep neural network classification model.

[0025] As an optional implementation, the lesion region features include any one or more of the grayscale value, entropy, grayscale statistics, shape features, texture features, and features extracted by multiple filters of the lesion region.

[0026] As an optional implementation manner, the step of inputting the scanned image into the lesion segmentation model further includes:

[0027] Inputting the scanned image into a lesion segmentation model to obtain a segmented tissue structure;

[0028] The step of expanding the region outward from the lesion region based on the similarity with the lesion region features to determine the region surrounding the lesion includes:

[0029] Based on the similarity with the features of the lesion region, a region expansion is performed outward from the lesion region and along the direction of the tissue structure vector to determine the region surrounding the lesion.

[0030] As an optional implementation manner, the step of inputting the scanned image into the lesion segmentation model further includes:

[0031] Inputting the scanned image into a lesion segmentation model to obtain a segmented tissue anatomical structure;

[0032] Before the step of obtaining the lesion area and the area surrounding the lesion, the scan image processing method further includes:

[0033] Based on the range and boundary of the tissue anatomical structure, the determined range and boundary of the perilesion area are limited to confine the lesion area and the perilesion area within the range and boundary of the tissue anatomical structure.

[0034] As an optional implementation manner, the step of expanding the region outward from the lesion region based on the similarity with the lesion region features includes:

[0035] Determine whether the similarity between the target pixel point outside the lesion area and the reference point is less than a preset threshold, if so, stop expanding the target pixel point, if not, continue expanding the target pixel point;

[0036] Wherein, the reference point is determined based on the pixel points of the lesion area.

[0037] As another aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned scanned image processing method when executing the computer program.

[0038] As another aspect of the present disclosure, a computer-readable medium is provided, on which computer instructions are stored. When the computer instructions are executed by a processor, the scanned image processing method as described above is implemented.

[0039] Based on this disclosure, those skilled in the art can understand other aspects of the present disclosure.

[0040] The positive progress of this disclosure is:

[0041] The scanning image processing method, electronic device and readable medium provided by the present disclosure can accurately extract lesion areas and other areas from CT and other scanning images, and effectively determine the area around the lesion using an adaptive radial expansion method, and then use the lesion determination results to effectively predict the lesion infiltration grade, providing doctors and other relevant personnel with more accurate tumor infiltration analysis and prediction, thereby improving the evaluation and prediction efficiency and accuracy, and effectively improving the doctor's work efficiency and differential diagnosis level. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The features and advantages of the present disclosure will be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, components are not necessarily drawn to scale, and components with similar related properties or characteristics may have the same or similar reference numerals.

[0043] Figure 1 Schematic diagram of a flow chart of a scanned image processing method according to an optional embodiment of the present disclosure.

[0044] Figure 2 Schematic diagram of the flow of the scan image processing method applied to lung tumor analysis.

[0045] Figure 3 A schematic diagram illustrating the model training and inference process of the lesion segmentation model.

[0046] Figure 4 Schematic diagram showing the traditional equidistant expansion of the lesion area.

[0047] Figure 5 Schematic diagram showing an adaptive radial expansion of a lesion area according to an optional embodiment of the present disclosure.

[0048] Figure 6 Schematic diagram showing the expansion of the lesion area using the lung structure.

[0049] Figure 7 Schematic diagram showing the expansion of the area surrounding the lesion within a lung lobe or segment.

[0050] Figure 8 Schematic diagram of the structure of the deep neural network classification model based on the attention mechanism.

[0051] Figure 9 FIG2 is a schematic structural diagram of an electronic device for implementing a scanned image processing method according to another optional embodiment of the present disclosure. DETAILED DESCRIPTION

[0052] The present disclosure is further illustrated below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0053] It should be noted that references in the specification to "one embodiment," "an alternative embodiment," "another embodiment," etc. indicate that the described embodiment may include a particular feature, structure, or characteristic, but every embodiment may not necessarily include that particular feature, structure, or characteristic. Moreover, such phrases do not necessarily refer to the same embodiment. In addition, when a particular feature, structure, or characteristic is described in conjunction with an embodiment, whether or not explicitly described, it is within the knowledge of those skilled in the relevant art to implement such feature, structure, or characteristic in conjunction with other embodiments.

[0054] In the description of the present disclosure, it should be understood that the terms "center", "lateral", "up", "down", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present disclosure. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present disclosure, unless otherwise specified, "multiple" means two or more. In addition, the term "including" and any variations thereof are intended to cover non-exclusive inclusions.

[0055] In the description of the present disclosure, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in the present disclosure based on the specific circumstances.

[0056] The terms used herein are intended only to describe specific embodiments and are not intended to limit exemplary embodiments. Unless the context clearly indicates otherwise, the singular forms "a", "an", "an item" used herein are also intended to include the plural. It should also be understood that the terms "comprise" and / or "include" used herein specify the presence of stated features, integers, steps, operations, units and / or components, and do not preclude the presence or addition of one or more other features, integers, steps, operations, units, components and / or combinations thereof.

[0057] In order to overcome the above-mentioned defects that currently exist, this embodiment provides a scanning image processing method, including: acquiring a scanning image; inputting the scanning image into a lesion segmentation model to obtain a segmented lesion area; based on the similarity with the characteristics of the lesion area, expanding the area outward from the lesion area to determine the area around the lesion; and obtaining the lesion area and the area around the lesion.

[0058] In this embodiment, the lesion area can be accurately extracted from the scanned image, and the area around the lesion can be effectively determined using an adaptive radial expansion method, thereby providing doctors and other relevant personnel with more accurate disease analysis and prediction, thereby improving the efficiency and accuracy of evaluation and prediction, and effectively improving the work efficiency and differential diagnosis level of doctors.

[0059] Specifically, as an optional embodiment, Figure 1 As shown, the scanned image processing method provided in this embodiment mainly includes the following steps:

[0060] Step 101: Acquire a scanned image.

[0061] In this embodiment, the scanning image processing method is mainly used in the lung tumor analysis scenario, and the scanning image is preferably a CT image of the patient, but the application scenario and image type of the scanning image processing method are not specifically limited, and can be adjusted and selected accordingly according to actual needs or possible needs.

[0062] Step 102: Input the scanned image into a lesion segmentation model to obtain the segmented lesion area, tissue structure, and tissue anatomical structure.

[0063] In this step, the lesion segmentation model can be a 3D convolutional neural network segmentation model, refer to Figure 2 As shown in the figure, a three-dimensional convolutional neural network is used to segment the lung tissue anatomical structures such as lung lobes and lung segments, lung tissue structures such as trachea and blood vessels, and lesion areas from the CT image. Of course, the segmentation model can also perform segmentation at the same time and can be set according to actual needs.

[0064] Specifically, refer to Figure 3 As shown in the figure, the neural network segmentation model mainly includes five segmentation models: lung lobe, lung segment, trachea, blood vessel, and lesion. It is mainly divided into model training and inference. During the model training phase, the corresponding ROI (region of interest) annotated image data is first collected for separate model training, generating five neural network model files containing a large number of parameters obtained through deep learning algorithms.

[0065] The training processes of the five segmentation models are similar. Therefore, taking the lesion segmentation model as an example, the training process of the lesion segmentation model is described in detail below.

[0066] The training phase of the lesion segmentation model is to input the CT images and annotated images into a three-dimensional convolutional neural network for training to obtain a segmentation model file. For example, the three-dimensional convolutional neural network can adopt a V-Net (a network architecture) network, and the training strategy can adopt a combination of two different scale segmentation models, thereby effectively improving the processing efficiency and accuracy of the segmentation model. Of course, the deep learning algorithm model used for model training and inference can also include a DenseNet (a network architecture) network, etc., which can be selected and adjusted accordingly according to actual needs or possible needs.

[0067] Specifically, the input raw volume data image is first resampled into two images of different resolution sizes (for example, using the sampling operators [3,3,3] and [1,1,1]). The entire image is normalized, and image blocks of the same size (for example, 64X64X64) are randomly extracted from the full image. The image blocks of the two resolution sizes are then input into the corresponding convolutional neural network for training. The training is repeated until the size of the loss function (for example, dice) is reduced to a preset threshold. The trained model files are saved to obtain segmentation model files based on two resolution sizes, namely the coarse segmentation network model file and the fine segmentation network model file.

[0068] The following describes the inference process for the lesion segmentation model. A coarse segmentation network model and a fine segmentation network model are cascaded. The coarse segmentation network model file is used to locate the lesion, while the fine segmentation network model file is used to finely segment the lesion margin. Specifically, the CT image is input into the deep learning segmentation algorithm. The segmentation algorithm first resamples the image to a specified resolution, then normalizes it and feeds the preprocessed image into the segmentation model to automatically extract the lesion region.

[0069] Step 103: Based on the similarity with the features of the lesion region, the region is expanded outward from the lesion region and along the direction of the tissue structure vector to determine the region surrounding the lesion.

[0070] In this step, based on the lesion segmentation result obtained in step 102, an adaptive radial expansion algorithm is used to expand and extend along lung structures such as the trachea and blood vessels to calculate the area around the lesion.

[0071] Specifically, the traditional expansion of the perilesional area such as the tumor is generally done by gradient isometric expansion. Figure 4 As shown in the example, using equidistant expansion of 0-5mm and 5-10mm, radiomic features are extracted from these equidistant regions and their relationship to benign and malignant classification is further analyzed. However, this equidistant expansion method does not consider whether there are disease-related or irrelevant areas in the peritumoral region; it is simply an equidistant expansion, thus reducing accuracy.

[0072] In this embodiment, it is proposed to utilize the grayscale information in the omics features to non-equidistantly expand and determine the peritumoral area.

[0073] refer to Figure 5 As shown, the middle inner area ① is the lesion area, and the annular area ② is based on the correlation of grayscale information. As an optional implementation method, different distances are expanded based on the entropy (entrop) of the lesion area. Upward expansion indicates that the entropy is close, and downward expansion indicates that the entropy is not close. Therefore, the similar area will expand a greater distance than the non-similar area. The annular area ③ becomes lighter in color, indicating that the similarity is decreasing, so the expanded area gradually becomes smaller until the similarity reaches the set threshold, and then it will no longer expand. Therefore, the greater the similarity of the grayscale information (entropy), the larger the expanded area. The non-equidistant expansion method determines the peritumoral area based on the characteristics of the area, and the boundary will be more accurate, which is more conducive to the assessment of the tumor.

[0074] As another embodiment, in this step, it is determined whether the similarity between the target pixel point outside the lesion area and the reference point is less than a preset threshold value. If so, the expansion of the target pixel point is stopped, and if not, the expansion of the target pixel point is performed; wherein the reference point is determined based on the pixel point of the lesion area. In this embodiment, the lesion area is determined, and the expanded area is used as the area around the lesion, that is, the area around the lesion changes dynamically, and the lesion area does not change dynamically. Of course, the lesion area can also change dynamically, and an iteration of the expanded area may occur at this time, that is, the last expanded area will change the selection of the pixel point of the lesion area, so during the iteration process, the lesion area can also change dynamically.

[0075] The basic principle of the non-equidistant expansion method is to group pixels with similar properties to form a region. Using all pixels in the lesion area as a reference point, the extended region is then radially expanded to include pixels whose proximity to the reference point falls within a set threshold. In this embodiment, similar properties include not only grayscale value and entropy, but also grayscale statistics, shape features, texture features, and features extracted by various filters.

[0076] From an imaging perspective, there are five main types of relationships between bronchi and pulmonary nodules: A (bronchial truncation by tumor); B (bronchial encapsulation by tumor); C (bronchial compression by tumor); D (uniform thickening of the bronchial wall leading to the tumor, with a smooth and narrow lumen); and E (irregular thickening and twisting of the bronchial wall leading to the tumor, with a narrow lumen). Histopathologically, adenocarcinoma in situ often originates from the bronchiolar epithelium or alveolar epithelium, resulting in bronchial stiffness, stretching, narrowing, and truncation. The pulmonary artery runs alongside the bronchi and often exhibits morphological changes similar to those of the bronchi.

[0077] Therefore, in this embodiment, based on the segmentation results of the trachea, blood vessels, and lesion regions, the expansion boundary of the peritumoral region is suppressed. Figure 6 As shown in the figure, taking the trachea as an example, the mask of the lesion area, the mask of the trachea, the expanded lesion area, and the arrow indicating the direction of the trachea (clinically for the direction of the tracheal vector). The expanded area expands more pixels along the tracheal direction and fewer pixels along the perpendicular direction of the tracheal vector. Of course, the method of obtaining the tracheal direction vector is not limited to using PCA (principal component analysis).

[0078] Step 104: Based on the range and boundary of the tissue anatomical structure, limit the range and boundary of the determined area around the lesion.

[0079] In this step, the determined range and boundary of the area surrounding the lesion are limited based on the range and boundary of the tissue anatomical structure, so as to confine the lesion area and the area surrounding the lesion to the range and boundary of the tissue anatomical structure.

[0080] Specifically, the segmentation results of the lung anatomical structure (lobes and segments) are used to limit the scope and boundaries of the expansion. Figure 7 As shown in the figure, after initially expanding the surrounding tissue, the segmentation results of the lung anatomy, including lobes and segments, are used to further restrict the expansion of the surrounding tissue, ensuring that the expansion area and the lesion area are within the same lung lobe. Pneumonia is located between lung segments, and lesions generally do not extend across lung segments. Therefore, the expansion area is limited to the same lung segment, unless the lesion extends across lung segments.

[0081] Step 105: Evaluate the lesion infiltration grade based on the image including the lesion area and the area surrounding the lesion to output a lesion infiltration grade result.

[0082] In this step, based on the tumor and peritumoral regions, radiomics analysis and a three-dimensional convolutional neural classification network model are used to classify, evaluate, and predict the invasiveness of the lesions.

[0083] In this embodiment, the diagnostic information obtained by automatic classification includes but is not limited to lung tumor, non-infiltration, microinfiltration, infiltration, etc., and can be selected and adjusted accordingly according to actual needs or possible needs.

[0084] As an optional embodiment, in this step, the imaging genomics features of the lesion area and the surrounding area are extracted from the image including the lesion area and the surrounding area; the image from which the imaging genomics features of the lesion area and the surrounding area are extracted is input into the lesion infiltration grading model to predict the lesion infiltration grading; and the lesion infiltration grading result is output.

[0085] Specifically, the following describes the method of using omics analysis. First, radiomic features are extracted from the tumor and peritumoral regions. Then, feature dimensionality reduction and feature selection are performed using methods such as LASSO (regression model). Furthermore, classification algorithms such as SVM (support vector machine), logistic regression, and random forest are used to construct a tumor invasiveness grading model, i.e., a lesion invasiveness grading model. This step has two main outputs: one is the selected features, which can be fed back to the peritumoral region expansion algorithm to provide more features for the adaptive radial expansion algorithm; the other is a classification model based on a classifier.

[0086] As another optional embodiment, in this step, an image including the lesion area and the area surrounding the lesion is input into a deep neural network classification model to output a lesion infiltrativeness grading result.

[0087] Among them, the training steps of the deep neural network classification model specifically include: inputting a training image including the lesion area and the area around the lesion into the deep neural network classification model to be trained; extracting the attention area from the training image and extracting the same attention area from the supervision image, where the attention area at least includes the lesion area and the area around the lesion; based on the lesion infiltrative properties of the pixel points in the attention areas of the training image and the supervision image, calculating the graded loss of the lesion infiltrative properties between the training image and the supervision image; training the deep neural network classification model to be trained according to the graded loss to obtain the trained deep neural network classification model.

[0088] Specifically, the deep neural network classification model is described below. The obtained tumor and peritumoral regions are input into the classification network using the attention mechanism, focusing on learning the depth information of the tumor and peritumoral regions. That is, the invasiveness result (i.e., a value of 0 to 1) of each pixel point in the attention region (i.e., containing the tumor and peritumoral regions) in the supervised image is used to calculate the hierarchical loss between it and the invasiveness result of each pixel point in the corresponding region in the training image, so as to train the deep neural network classification model based on the calculated hierarchical loss, thereby allowing the classification model to focus on learning the tumor and peritumoral regions, such as Figure 8 As shown in the figure, the stability and accuracy of classification are improved.

[0089] During the classification model training phase, the CT image containing the tumor and peritumoral regions is input into a three-dimensional convolutional neural network for training. For example, a dense convolutional network (DenseNet) can be used for the three-dimensional convolutional neural network. The classification results of the tumor and peritumoral regions are input into the network at the same time. The attention mechanism is used to allow the model to focus on learning this region. The graded loss (MSE Loss) is calculated to finally obtain the classification model file.

[0090] During the classification model inference phase, the segmentation results of the tumor and peritumor are first read and input into the classification model together with the original image containing the lesion to obtain multiple graded probability values. The maximum probability value is set as the final predicted invasion level, thereby realizing automatic assessment of the invasiveness of lung cancer.

[0091] In this embodiment, the parameters involved in the above steps can be arbitrarily set according to the actual characteristics of the medical image.

[0092] The scanning image processing method provided in this embodiment can accurately extract areas such as lung lobes, lung segments, trachea, blood vessels, and lesions from CT and other scanning images through a deep neural network segmentation model. Based on the boundary of the lesion, an adaptive radial expansion algorithm is used, and combined with the structural characteristics of the lungs (trachea and blood vessels, etc.), the peritumoral area can be effectively calculated. The segmentation results of the lung anatomical structure are then used to further limit and accurately locate the peritumoral boundary. Imaging genomics analysis and deep neural network classification model training are then performed based on the tumor and peritumoral area, thereby effectively predicting the lesion infiltration grade, providing doctors and other relevant personnel with more accurate tumor infiltration analysis and prediction, improving the efficiency and accuracy of evaluation and prediction, and thereby effectively improving the work efficiency and differential diagnosis level of doctors.

[0093] Figure 9 The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the scanned image processing method in the above embodiment is implemented. Figure 9 The electronic device 30 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0094] like Figure 9 As shown, the electronic device 30 may be a general-purpose computing device, such as a server device. Components of the electronic device 30 may include, but are not limited to, the at least one processor 31, the at least one memory 32, and a bus 33 connecting different system components (including the memory 32 and the processor 31).

[0095] The bus 33 includes a data bus, an address bus, and a control bus.

[0096] The memory 32 may include a volatile memory, such as a random access memory (RAM) 321 and / or a cache memory 322 , and may further include a read-only memory (ROM) 323 .

[0097] The memory 32 may also include a program / utility 325 having a set (at least one) of program modules 324, such program modules 324 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0098] The processor 31 executes computer programs stored in the memory 32 to perform various functional applications and data processing, such as the scan image processing method in the above embodiment of the present disclosure.

[0099] The electronic device 30 may also communicate with one or more external devices 34 (e.g., a keyboard, a pointing device, etc.). Such communication may be performed via an input / output (I / O) interface 35. Furthermore, the model generating device 30 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 36. Figure 9 As shown, the network adapter 36 communicates with the other modules of the model-generated device 30 via the bus 33. It should be understood that, although not shown in the figures, other hardware and / or software modules may be used in conjunction with the model-generated device 30, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0100] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0101] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the scanned image processing method in the above embodiment are implemented.

[0102] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0103] In a possible implementation, the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product is executed on a terminal device, the program code is used to enable the terminal device to execute the steps in the scan image processing method in the above embodiment.

[0104] The program code for executing the present disclosure may be written in any combination of one or more programming languages, and may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0105] While specific embodiments of the present disclosure have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of protection of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications are intended to fall within the scope of protection of the present disclosure.

Claims

1. A scanned image processing method, characterized in that: include: Acquire a scanned image; Inputting the scanned image into a lesion segmentation model to obtain a segmented lesion area; Based on the similarity with the characteristics of the lesion area, the area surrounding the lesion is expanded outward from the lesion area, including: Determine whether the similarity between the target pixel point outside the lesion area and the reference point is less than a preset threshold; if not, expand the target pixel point; The reference point is determined based on the pixel points of the lesion area, and when the target pixel points are expanded, the expansion distance of the area is positively correlated with the proximity; as well as, The lesion area and the area surrounding the lesion are obtained.

2. The scanned image processing method according to claim 1, wherein: After the step of obtaining the lesion area and the area surrounding the lesion, the image processing method further includes: The lesion infiltration grade is evaluated based on the image including the lesion area and the area surrounding the lesion to output a lesion infiltration grade result.

3. The scanned image processing method according to claim 2, wherein: The step of evaluating the lesion infiltration grade based on the image including the lesion area and the area surrounding the lesion to output the lesion infiltration grade result includes: Extracting radiomic features of the lesion area and the area surrounding the lesion from an image including the lesion area and the area surrounding the lesion; Inputting the image from which the radiomic features of the lesion area and the area surrounding the lesion are extracted into a lesion invasiveness grading model to predict the lesion invasiveness grading; Output the lesion invasiveness grading results.

4. The scanned image processing method according to claim 2, wherein: The step of evaluating the lesion infiltration grade based on the image including the lesion area and the area surrounding the lesion to output the lesion infiltration grade result includes: Input the image including the lesion area and the area surrounding the lesion into the deep neural network classification model to output the lesion invasiveness grading result; The training steps of the deep neural network classification model specifically include: Inputting a training image including the lesion area and the area surrounding the lesion into a deep neural network classification model to be trained; Extracting an attention region from the training image and extracting the same attention region from the supervision image, wherein the attention region includes at least the lesion region and the lesion surrounding region; Calculating a graded loss of lesion infiltration between the training image and the supervisory image based on the lesion infiltration of pixels in the attention area of ​​the training image and the supervisory image; The deep neural network classification model to be trained is trained according to the hierarchical loss to obtain a trained deep neural network classification model.

5. The scanned image processing method according to any one of claims 1 to 4, wherein: The lesion region features include any one or more of the grayscale value, entropy, grayscale statistics, shape features, texture features, and features extracted by multiple filters of the lesion region.

6. The scanned image processing method according to claim 5, wherein: The step of inputting the scanned image into the lesion segmentation model further includes: Inputting the scanned image into a lesion segmentation model to obtain a segmented tissue structure; The step of expanding the region outward from the lesion region based on the similarity with the lesion region features to determine the region surrounding the lesion includes: Based on the similarity with the features of the lesion region, a region expansion is performed outward from the lesion region and along the direction of the tissue structure vector to determine the region surrounding the lesion.

7. The scanned image processing method according to claim 6, wherein: The step of inputting the scanned image into the lesion segmentation model further includes: Inputting the scanned image into a lesion segmentation model to obtain a segmented tissue anatomical structure; Before the step of obtaining the lesion area and the area surrounding the lesion, the scan image processing method further includes: Based on the range and boundary of the tissue anatomical structure, the determined range and boundary of the perilesion area are limited to confine the lesion area and the perilesion area within the range and boundary of the tissue anatomical structure.

8. The scanned image processing method according to claim 1, wherein: The step of expanding the region outward from the lesion region based on the similarity with the lesion region features includes: It is determined whether the similarity between the target pixel point outside the lesion area and the reference point is less than a preset threshold. If so, the expansion of the target pixel point is stopped.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the scanned image processing method according to any one of claims 1 to 8 is implemented.

10. A computer-readable medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the scanned image processing method according to any one of claims 1 to 8 is implemented.

Images