Method and system for predicting enhanced region of liver cancer image

Abstract

The invention provides a liver cancer image enhancement region prediction method and system, and relates to the technical field of medical image processing. Comprising the following steps: acquiring a liver cancer medical image, and segmenting the liver cancer medical image based on a pre-trained neural network model to obtain an aorta region, a liver region and a liver cancer tumor region; screening a residual liver region from the liver region according to the liver cancer tumor region; calculating an intensified blood vessel intensity threshold value of the aorta region, and screening out a residual liver correction region greater than the intensified blood vessel intensity threshold value from the residual liver region; and calculating the average HU value of the residual liver correction area as a liver cancer tumor strengthening threshold value, and screening out an area greater than the liver cancer tumor strengthening threshold value from the liver cancer tumor area as a liver cancer tumor strengthening area. According to the invention, on the basis that the neural network segments the liver cancer medical image, the algorithm is adopted to extract the liver cancer tumor intensified area, manual identification measurement is not needed, errors are greatly reduced, and accurate identification of the liver cancer tumor intensified area is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image processing, and in particular relates to a method and system for predicting enhanced regions of liver cancer images. Background Art

[0002] Hepatocellular carcinoma (HCC) is a common primary liver malignancy. The evaluation of its treatment effect is an important step in the management of HCC patients and drug clinical trials.

[0003] At present, the mRECIST (modified Response Evaluation Criteria in Solid Tumors) standard is often used to evaluate the response to HCC treatment. This standard selects lesions suitable for repeated measurement as target lesions, and gives priority to measuring typical lesions, that is, intrahepatic HCC lesions that show non-marginal tumor enhancement on CT or MRI enhanced scans. To adapt to the particularity of HCC, mRECIST requires the measurement of the longest diameter of the enhanced part of the lesion, that is, the active part of the tumor, while avoiding the necrotic area. At present, the identification and measurement of target lesions rely on subjective manual evaluation. Since the layer where the longest diameter of the enhanced part is located may change with treatment, and the necrotic area in the tumor needs to be avoided, the reliable results require careful observation, measurement and comparison of images at each level, which is time-consuming and labor-intensive, and inevitably has errors. Summary of the invention

[0004] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for predicting liver cancer image enhancement areas. Based on the segmentation of liver cancer medical images by a neural network, an algorithm is used to extract the liver cancer tumor enhancement area therein. No manual identification measurement is required, which greatly reduces the error and realizes accurate identification of liver cancer tumor enhancement areas.

[0005] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0006] A first aspect of the present invention provides a method for predicting enhanced regions of liver cancer images.

[0007] The method for predicting liver cancer image enhancement area includes the following steps:

[0008] Obtain liver cancer medical images, segment the liver cancer medical images based on the pre-trained neural network model, and obtain the aorta area, liver area, and liver cancer tumor area;

[0009] According to the liver cancer tumor area, the residual liver area is screened from the liver area;

[0010] Calculate the enhanced vascular intensity threshold of the aorta region, and select the corrected remnant liver region with an intensity greater than the enhanced vascular intensity threshold from the remnant liver region;

[0011] The average HU value of the residual liver correction area was calculated as the liver cancer tumor enhancement threshold, and the area with a value greater than the liver cancer tumor enhancement threshold was screened out from the liver cancer tumor area as the liver cancer tumor enhancement area.

[0012] A second aspect of the present invention provides a liver cancer image enhancement region prediction system.

[0013] Liver cancer image enhancement region prediction system, including:

[0014] The image segmentation module is configured to: acquire a liver cancer medical image, and segment the liver cancer medical image based on a pre-trained neural network model to obtain an aorta region, a liver region, and a liver cancer tumor region;

[0015] The residual liver region screening module is configured to: screen the residual liver region from the liver region according to the liver cancer tumor region;

[0016] The residual liver correction region screening module is configured to: calculate the enhanced blood vessel intensity threshold of the aorta region, and screen out the residual liver correction region whose intensity is greater than the enhanced blood vessel intensity threshold from the residual liver region;

[0017] The liver cancer tumor enhancement area screening module is configured to: calculate the average HU value of the residual liver correction area as the liver cancer tumor enhancement threshold, and screen out the area greater than the liver cancer tumor enhancement threshold from the liver cancer tumor area as the liver cancer tumor enhancement area.

[0018] The third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for predicting liver cancer image enhancement regions as described in the first aspect of the present invention.

[0019] The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for predicting liver cancer image enhancement regions as described in the first aspect of the present invention are implemented.

[0020] One or more of the above technical solutions have the following beneficial effects:

[0021] The present invention provides a method and system for predicting enhanced regions of liver cancer images. Based on a pre-trained neural network model, liver cancer medical images are segmented to obtain an aorta region, a liver region, and a liver cancer tumor region. Then, the residual liver region and the residual liver correction region in the liver region are screened in turn according to the rules, and the average HU value of the residual liver correction region is calculated as the liver cancer tumor enhancement threshold. The region greater than the liver cancer tumor enhancement threshold is screened from the liver cancer tumor region as the liver cancer tumor enhancement region. Based on the segmentation of liver cancer medical images by a neural network, the present invention uses an algorithm to extract the liver cancer tumor enhancement region therein, without the need for manual identification measurement, which greatly reduces the error and realizes the accurate identification of the liver cancer tumor enhancement region.

[0022] After obtaining the liver cancer tumor enhancement area, the relevant parameters of the liver cancer tumor enhancement area are measured, including the area of ​​the enhancement area in the largest layer of the tumor, the maximum diameter of the enhancement area in the largest layer of the tumor, the volume of the enhancement area and the three-dimensional maximum diameter of the enhancement area, to achieve medical measurement and improve the work efficiency of medical workers.

[0023] When training the neural network model, the present invention marks the region of interest on the sample image of the liver cancer medical image, generates a corresponding mask image, pairs multiple sample images with the corresponding mask image to form a training data set, and uses the training data set to train the neural network model until the loss function is minimized, thereby realizing the automatic segmentation of the region of interest in the liver cancer medical image by the neural network model. The neural network model obtained by the training of the present invention through the above method can realize the accurate segmentation of medical images.

[0024] In the process of obtaining the region of interest, for the Ith region of interest, if I corresponds to the probability map where the liver or aorta is located, the connected domain with the largest area is retained as the output segmentation result of the liver or aorta; if I corresponds to the probability map where the liver cancer tumor is located, the number of output connected domains is not limited, that is, all connected domains are retained as the output segmentation results of several liver tumors. The above means are used to further improve the recognition accuracy.

[0025] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0027] Figure 1 This is a flow chart of the method of the first embodiment.

[0028] Figure 2This is a schematic diagram of the segmentation result of the arterial phase in enhanced CT in the first embodiment.

[0029] Figure 3 This is a schematic diagram of the tumor enhancement region R_en of liver cancer in the arterial phase of enhanced CT in the first embodiment. DETAILED DESCRIPTION

[0030] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0031] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0032] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0033] Embodiment 1

[0034] As mentioned above, relying on subjective manual evaluation to identify and measure target lesions is time-consuming and labor-intensive, and inevitably contains errors. The vigorous development of artificial intelligence is expected to solve the above problems well. Artificial intelligence algorithms can automatically extract key features in images, such as shape and texture, to achieve automatic image segmentation, which can be used for lesion identification. With its advantages of high efficiency and high accuracy, it has been applied in many aspects, including diagnosis, treatment and prognosis assessment. Using artificial intelligence to extract and calculate the size of the enhanced part is expected to become a feasible technology for accurate and automatic measurement of the enhanced diameter of HCC lesions.

[0035] Based on this, this embodiment discloses a method for predicting liver cancer image enhancement areas. Figure 1 As shown, the liver cancer image enhancement region prediction method may include the following steps:

[0036] Obtain liver cancer medical images, segment the liver cancer medical images based on the pre-trained neural network model, and obtain the aorta area, liver area, and liver cancer tumor area;

[0037] According to the liver cancer tumor area, the residual liver area is screened from the liver area;

[0038] Calculate the enhanced vascular intensity threshold of the aorta region, and select the corrected remnant liver region with an intensity greater than the enhanced vascular intensity threshold from the remnant liver region;

[0039] The average HU value of the residual liver correction area was calculated as the liver cancer tumor enhancement threshold, and the area with a value greater than the liver cancer tumor enhancement threshold was screened out from the liver cancer tumor area as the liver cancer tumor enhancement area.

[0040] In some embodiments, the complement of the liver cancer tumor region relative to the liver region is taken as the residual liver region.

[0041] In some embodiments, the average HU value of the aorta region is used as the enhanced vessel intensity threshold of the aorta region.

[0042] In some embodiments, after obtaining the liver cancer tumor enhancement region, it also includes measurement of parameters related to the liver cancer tumor enhancement region, including the area of ​​the enhancement region in the largest layer of the tumor, the maximum diameter of the enhancement region in the largest layer of the tumor, the volume of the enhancement region and the three-dimensional maximum diameter of the enhancement region.

[0043] As you can understand, HU value is a unit of CT value, which is used to express tissue density. CT value is a relative value obtained by measuring the X-ray attenuation coefficient and undergoing a certain conversion mode, which is used to express the density difference of different tissues. HU value is widely used in medical imaging, especially in CT examination, which helps doctors evaluate the density of tissues and possible diseases. In CT examination, HU value is used to measure the density of different tissues.

[0044] In this embodiment, the calculation methods of the average HU value of the aorta region and the average HU value of the residual liver correction region are common knowledge in the art and will not be elaborated herein.

[0045] In some embodiments, the neural network model is further trained, specifically:

[0046] Acquire multiple sample images of liver cancer medical images, annotate the regions of interest therein, generate corresponding mask images, and pair the multiple sample images with the corresponding mask images to form a training data set;

[0047] Establish a neural network model, train the neural network model based on the training data set until the loss function is minimized, and realize the automatic segmentation of the region of interest in liver cancer medical images by the neural network model;

[0048] The regions of interest include the aorta region, the liver region and the liver cancer tumor region. The neural network model includes a fully convolutional neural network, a convolutional neural network in a coding-decoding symmetric form, or a Trans-Unet structure that combines a traditional convolutional neural network and a Transformer. The loss function includes a set similarity measurement function and a focal loss function.

[0049] In some embodiments, the method further includes preprocessing the acquired liver cancer medical images, including:

[0050] Pixel standardization, by adjusting the brightness and contrast information of the image, makes the liver cancer medical images collected by different scanning schemes have similar feature range and distribution;

[0051] Voxel normalization resamples the actual physical space range corresponding to each voxel in liver cancer medical images to a uniform scale.

[0052] In some embodiments, the specific process of obtaining the region of interest is:

[0053] The output dimension of the neural network model is 1×(I+1), which represents the probability map corresponding to I regions of interest in one input channel and one background probability map respectively;

[0054] Binarizing the probability map according to a preset probability threshold to obtain a binary mask image;

[0055] For the Ith region of interest, if i corresponds to the probability map of the liver or aorta, the connected domain with the largest area is retained as the output segmentation result of the liver or aorta;

[0056] If I corresponds to the probability map where the liver cancer tumor is located, the number of output connected domains is not limited, that is, all connected domains are retained as the output segmentation results of several liver tumors.

[0057] The technical solution of this embodiment will be explained in detail below in conjunction with the accompanying drawings. The process of the liver cancer image enhancement area prediction method provided by this embodiment mainly includes:

[0058] 1) Generate training dataset

[0059] Using annotation software, for a given N sample images, mark I region of interest contours Ri (1≤i≤I, and I is a natural number) for each sample image, and generate its corresponding mask image. The mask image here consists of I binary mask rectangles with the same size as the original sample image, which assign different representative values ​​i to the coordinate points in each region of interest contour Ri.

[0060] In this way, N mask images are paired with N sample images to form a training data set.

[0061] 2) Establish a convolutional neural network model

[0062] Construct a convolutional neural network model and set the hyperparameters of the convolutional neural network model, where the input channel of the network is 1 and the output channel is 1×(I+1), corresponding to the detection probability maps of the contours of I regions of interest and the detection probability map of the background, respectively.

[0063] The training data set is divided into training set X1, validation set X2 and test set X3. The training set, validation set and test set are independent of each other, and their numbers are n1, n2 and n3 respectively, which are natural numbers, and n1+n2+n3=N, n1≥1 / 2N.

[0064] 3) Training Convolutional Neural Network

[0065] The convolutional neural network established in step 2) is trained using the training data set generated in step 1). The training set X1 is used to train the convolutional neural network model, the validation set X2 is used to evaluate the current performance of the model, and the test set X3 is used to test the generalization performance of the model.

[0066] During the training process, the training set will be divided into multiple batches and repeatedly input into the convolutional neural network model for multiple rounds of training. At the same time, the loss function is used to calculate the difference between the detection probability map output by the convolutional neural network segmentation model and the mask image as the training error and performance test indicator.

[0067] Optionally, the loss function includes but is not limited to a set similarity measurement function (Dice loss) or a focal loss function (Focol loss), etc. The difference between the predicted result and the actual result is calculated by a predefined loss function, and finally the model parameters are adjusted according to the difference to minimize the loss.

[0068] Through the optimization algorithm, the fitting process of minimizing the error in the training process of the convolutional neural network model is realized. Optionally, commonly used optimization algorithms include but are not limited to: Stochastic gradient descent (SGD), Adaptive Moment Estimation (Adam), and Momentum.

[0069] After each batch of training is completed, the validation set is used to perform a performance test on the convolutional neural network model. When the performance test indicators are stable, the training of the convolutional neural network model is considered complete and the trained network model is saved. Optional neural network models for implementing region of interest segmentation include but are not limited to: fully convolutional neural network (FCN), convolutional neural network with encoding and decoding symmetry (such as U-net, V-net), or Trans-Unet, a structure that combines traditional convolutional neural network and Transformer.

[0070] 4) Divide the enhanced area according to the segmentation output

[0071] 4-1 Obtaining segmentation output:

[0072] The original image is preprocessed as necessary. Common preprocessing operations include pixel normalization and voxel normalization.

[0073] Pixel normalization is to adjust the brightness, contrast and other information of the image so that different scanning schemes have similar feature ranges and distributions, which is conducive to the model to better learn effective features. Optionally, pixel normalization methods include but are not limited to: absolute value normalization, maximum-minimum normalization, Z-score normalization, etc.; voxel normalization is to resample the actual physical space range corresponding to each voxel in the original image to a uniform scale to avoid the impact of physical space differences on the training process. Optionally, the resampling methods of voxel normalization include but are not limited to: nearest neighbor interpolation, bilinear interpolation, spline interpolation, etc.

[0074] The pre-processed image is used as the input of the network model to segment the imaging area of ​​liver cancer lesions. Medical images of various modalities can be used, such as CT, MRI, etc. If necessary, multiple sequence combinations can be included at the same time to build a multi-channel neural network, and the liver cancer tumor area can be finally confirmed by learning the signal differences in different influencing sequences, such as the arterial phase and venous phase in enhanced images.

[0075] In this embodiment, the arterial phase and venous phase images of the abdominal enhanced CT are selected as the original images. The pre-trained network is used to perform forward reasoning calculations on the input image for the region of interest to complete automatic segmentation. The output dimension of the network model is 1×(I+1), and this output represents a probability map corresponding to I regions of interest in one input channel and a background probability map, wherein the pixel value in the probability map is the probability of the corresponding key point or background.

[0076] In this embodiment, there are three regions of interest, namely, the liver, liver tumor, and aorta region in enhanced CT. The following steps can be used to obtain the region of interest for each probability map:

[0077] ① Binarize the probability map according to the preset probability threshold to obtain a binary mask image;

[0078] ② For the i-th region of interest except the background probability map, if i corresponds to the probability map where the liver or aorta is located, the connected domain with the largest area is retained as the output liver or aorta segmentation result;

[0079] If i corresponds to the probability map where the liver cancer tumor is located, the number of output connected domains is not limited, that is, all connected domains are retained as the output segmentation results of several liver tumors.

[0080] Schematic diagram of the segmentation results of the arterial phase in enhanced CT Figure 2 As shown, the pink area represents the liver; the gray area represents the liver cancer tumor; and the green area represents the aorta.

[0081] 4-2 Acquisition of liver cancer tumor enhancement area:

[0082] In this embodiment, the enhanced blood vessels and tumor areas are first removed during the enhanced CT arterial phase to obtain the residual liver area, and the average HU value of the pixels in the residual liver area is calculated as the enhancement threshold.

[0083] Based on the automatic segmentation output of the liver, liver cancer tumor and aorta in enhanced CT in step 4-1, firstly, the complement Set1 of the liver cancer tumor relative to the liver in the enhanced CT arterial phase sequence is taken as the residual liver region R1.

[0084] To further remove the influence of the enhanced vascular region in the residual liver, only the residual liver correction region R2 in the residual liver region that is greater than the enhanced vascular intensity threshold T1 is retained. In this embodiment, T1 is defined as the average HU value of the aorta region. The aorta region is the aorta segmentation region automatically output in the enhanced CT arterial phase in step 4-1.

[0085] Afterwards, the average HU value of the residual liver correction region R2 retained in the above steps is calculated as the liver cancer tumor enhancement threshold T2.

[0086] Finally, the area of ​​the liver cancer tumor that is automatically output in the enhanced CT arterial phase in step 4-1 and is larger than the liver cancer tumor enhancement threshold T2 is retained as the liver cancer tumor enhancement area R_en.

[0087] like Figure 3 The figure shows a schematic diagram of R_en in the arterial phase of enhanced CT in this embodiment, wherein the red area represents the liver cancer tumor; and the yellow area represents the liver cancer tumor enhancement area.

[0088] Based on the liver cancer tumor enhancement region R_en obtained in step 4-2, combined with the patient's pixel space (pixelspacing) information in the medical image header file, a variety of R_en related parameter measurements can be completed. These include but are not limited to: the area of ​​the enhanced region in the largest layer of the tumor, the maximum diameter of the enhanced region in the largest layer of the tumor, the volume of the enhanced region, and the maximum three-dimensional diameter of the enhanced region. Among them:

[0089] The area of ​​enhanced area in the largest section of the tumor:

[0090] Based on the liver cancer tumor area automatically output in the enhanced CT arterial phase in step 4-1, the area of ​​the liver cancer tumor at that level is calculated layer by layer, and the largest level s of the liver cancer tumor in the three-dimensional CT sequence is selected by layer by layer comparison. Based on the liver cancer tumor enhancement area output in step 4-2, its area in level s is calculated.

[0091] The maximum diameter of the enhanced area in the largest layer of the tumor:

[0092] Based on the liver cancer tumor area automatically output in the enhanced CT arterial phase in step 4-1, the area of ​​the liver cancer tumor at that level is calculated layer by layer, and the largest level s of the liver cancer tumor in the three-dimensional CT sequence is selected by layer-by-layer comparison. Based on the liver cancer tumor enhancement area output in step 4-2, the distance between the two farthest pixels in level s is calculated, which is regarded as the maximum diameter of the enhancement area in the largest level of the tumor.

[0093] Enhanced area volume:

[0094] Based on the liver cancer tumor enhancement region outputted in step 4-2, its volume in three-dimensional space is calculated.

[0095] Maximum three-dimensional diameter of the enhanced area:

[0096] Based on the liver cancer tumor enhancement region outputted in step 4-2, the distance between the two voxel points that are farthest apart in the three-dimensional space is calculated and regarded as the three-dimensional maximum diameter of the enhancement region.

[0097] This embodiment provides a method for predicting enhanced regions of liver cancer images, which segments liver cancer medical images based on a pre-trained neural network model to obtain the aorta region, liver region, and liver cancer tumor region; then, the residual liver region and residual liver correction region in the liver region are screened in turn according to the rules, and the average HU value of the residual liver correction region is calculated as the liver cancer tumor enhancement threshold, and the region greater than the liver cancer tumor enhancement threshold is screened from the liver cancer tumor region as the liver cancer tumor enhancement region. Based on the segmentation of liver cancer medical images by the neural network, an algorithm is used to extract the liver cancer tumor enhancement region therein, without the need for manual identification and measurement, which greatly reduces the error and achieves accurate identification of the liver cancer tumor enhancement region;

[0098] After obtaining the liver cancer tumor enhancement area, the relevant parameters of the liver cancer tumor enhancement area are measured, including the area of ​​the enhancement area in the largest layer of the tumor, the maximum diameter of the enhancement area in the largest layer of the tumor, the volume of the enhancement area and the three-dimensional maximum diameter of the enhancement area, to achieve medical measurement and improve the work efficiency of medical workers.

[0099] When training the neural network model, this embodiment marks the region of interest on the sample image of the liver cancer medical image, generates a corresponding mask image, pairs multiple sample images with the corresponding mask image to form a training data set, and uses the training data set to train the neural network model until the loss function is minimized, thereby realizing the automatic segmentation of the region of interest in the liver cancer medical image by the neural network model. The neural network model obtained by the training method of the present invention can realize the accurate segmentation of medical images;

[0100] In the process of obtaining the region of interest, for the Ith region of interest, if I corresponds to the probability map where the liver or aorta is located, the connected domain with the largest area is retained as the output segmentation result of the liver or aorta; if I corresponds to the probability map where the liver cancer tumor is located, the number of output connected domains is not limited, that is, all connected domains are retained as the output segmentation results of several liver tumors. The above means are used to further improve the recognition accuracy.

[0101] Embodiment 2

[0102] This embodiment discloses a liver cancer image enhancement region prediction system.

[0103] Liver cancer image enhancement region prediction system, including:

[0104] The image segmentation module is configured to: acquire a liver cancer medical image, and segment the liver cancer medical image based on a pre-trained neural network model to obtain an aorta region, a liver region, and a liver cancer tumor region;

[0105] The residual liver region screening module is configured to: screen the residual liver region from the liver region according to the liver cancer tumor region;

[0106] The residual liver correction region screening module is configured to: calculate the enhanced blood vessel intensity threshold of the aorta region, and screen out the residual liver correction region whose intensity is greater than the enhanced blood vessel intensity threshold from the residual liver region;

[0107] The liver cancer tumor enhancement area screening module is configured to: calculate the average HU value of the residual liver correction area as the liver cancer tumor enhancement threshold, and screen out the area greater than the liver cancer tumor enhancement threshold from the liver cancer tumor area as the liver cancer tumor enhancement area.

[0108] Embodiment 3

[0109] The purpose of this embodiment is to provide a computer-readable storage medium.

[0110] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the method for predicting liver cancer image enhancement areas as described in Example 1 of the present disclosure.

[0111] Embodiment 4

[0112] The purpose of this embodiment is to provide an electronic device.

[0113] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for predicting liver cancer image enhancement areas as described in Example 1 of the present disclosure are implemented.

[0114] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0115] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0116] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for predicting liver cancer image enhancement regions, characterized in that: The following steps are involved: Obtain liver cancer medical images, segment the liver cancer medical images based on the pre-trained neural network model, and obtain the aorta area, liver area, and liver cancer tumor area; According to the liver cancer tumor area, the residual liver area is screened from the liver area; Calculate the enhanced vascular intensity threshold of the aorta region, and select the corrected remnant liver region with an intensity greater than the enhanced vascular intensity threshold from the remnant liver region; The average HU value of the residual liver correction area was calculated as the liver cancer tumor enhancement threshold, and the area with a value greater than the liver cancer tumor enhancement threshold was screened out from the liver cancer tumor area as the liver cancer tumor enhancement area.

2. The method for predicting liver cancer image enhancement area according to claim 1, characterized in that: The complement of the liver cancer tumor area relative to the liver area is taken as the residual liver area.

3. The method for predicting liver cancer image enhancement area according to claim 1, characterized in that: The average HU value of the aorta region was used as the threshold of enhanced vascular intensity in the aorta region.

4. The method for predicting liver cancer image enhancement area according to claim 1, characterized in that: After obtaining the liver cancer tumor enhancement region, it also includes measurement of parameters related to the liver cancer tumor enhancement region, including the area of ​​the enhancement region in the largest layer of the tumor, the maximum diameter of the enhancement region in the largest layer of the tumor, the volume of the enhancement region and the three-dimensional maximum diameter of the enhancement region.

5. The method for predicting liver cancer image enhancement area according to claim 1, characterized in that: It also includes training the neural network model, specifically: Acquire multiple sample images of liver cancer medical images, annotate the regions of interest therein, generate corresponding mask images, and pair the multiple sample images with the corresponding mask images to form a training data set; Establish a neural network model, train the neural network model based on the training data set until the loss function is minimized, and realize the automatic segmentation of the region of interest in liver cancer medical images by the neural network model; The regions of interest include the aorta region, the liver region and the liver cancer tumor region. The neural network model includes a fully convolutional neural network, a convolutional neural network in a coding-decoding symmetric form, or a Trans-Unet structure that combines a traditional convolutional neural network and a Transformer. The loss function includes a set similarity measurement function and a focal loss function.

6. The method for predicting liver cancer image enhancement area according to claim 1, characterized in that: It also includes pre-processing of acquired liver cancer medical images, including: Pixel standardization, by adjusting the brightness and contrast information of the image, makes the liver cancer medical images collected by different scanning schemes have similar feature range and distribution; Voxel normalization resamples the actual physical space range corresponding to each voxel in liver cancer medical images to a uniform scale.

7. The method for predicting liver cancer image enhancement area according to claim 5, characterized in that: The specific process of obtaining the region of interest is as follows: The output dimension of the neural network model is 1×(I+1), which represents the probability map corresponding to I regions of interest in one input channel and one background probability map respectively; Binarizing the probability map according to a preset probability threshold to obtain a binary mask image; For the Ith region of interest, if I corresponds to the probability map of the liver or aorta, the connected domain with the largest area is retained as the output segmentation result of the liver or aorta; If I corresponds to the probability map where the liver cancer tumor is located, the number of output connected domains is not limited, that is, all connected domains are retained as the output segmentation results of several liver tumors.

8. Liver cancer image enhancement region prediction system, characterized by: include: The image segmentation module is configured to: acquire a liver cancer medical image, and segment the liver cancer medical image based on a pre-trained neural network model to obtain an aorta region, a liver region, and a liver cancer tumor region; The residual liver region screening module is configured to: screen the residual liver region from the liver region according to the liver cancer tumor region; The residual liver correction region screening module is configured to: calculate the enhanced blood vessel intensity threshold of the aorta region, and screen out the residual liver correction region whose intensity is greater than the enhanced blood vessel intensity threshold from the residual liver region; The liver cancer tumor enhancement area screening module is configured to: calculate the average HU value of the residual liver correction area as the liver cancer tumor enhancement threshold, and screen out the area greater than the liver cancer tumor enhancement threshold from the liver cancer tumor area as the liver cancer tumor enhancement area.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the liver cancer image enhancement region prediction method according to any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the liver cancer image enhancement region prediction method according to any one of claims 1 to 7 are implemented.

Images