Medical image noise suppression and enhancement method based on artificial intelligence

Abstract

The invention relates to the technical field of image processing, in particular to a medical image noise suppression and enhancement method based on artificial intelligence, and the method comprises the steps: S1, collecting original data of an input ultrasonic image, and carrying out the prediction of a noise type according to the ultrasonic image and the original data; and S2, calling a noise correction model of a corresponding type according to the noise type to process the input ultrasonic image so as to obtain a repaired image. In order to solve the problem that in the prior art, imaging degradation is easily caused by noise components in the imaging process of ultrasonic scanning equipment, according to the scheme, original data in the imaging process are predicted, and the noise components in the data and the corresponding noise types are obtained through recognition; and according to the noise type, the corresponding noise correction model is called to correct the corresponding area, so that noise removal is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a medical image noise suppression and enhancement method based on artificial intelligence. Background Art

[0002] Ultrasound scanning, or medical ultrasound examination, is a medical imaging diagnostic technology based on ultrasound. This technology can use ultrasound to visualize the patient's muscle tissue and internal organs outside the patient's body, including size, structure, and pathological lesions, to assist doctors in making effective diagnoses of the patient's disease. Ultrasound scanning equipment refers to medical equipment that uses the above-mentioned ultrasound scanning technology, which uses ultrasound to image the various tissue parts of the patient to assist in medical diagnosis. In order to achieve a better diagnostic effect on the disease, it is usually necessary to compile a corresponding ultrasound scanning diagnostic process and perform the scanning process according to a specific process. For example, for thyroid scanning, it is usually necessary to design the corresponding scanning direction, sequence, and judgment branches, such as how to deal with the nodule site when scanning.

[0003] In the prior art, there are already inspection schemes based on ultrasound. For example, Chinese patent CN202011641878.5 provides an ultrasonic scanning feedback system, including: an ultrasonic probe, used to scan the scanning part of the detection object to obtain an ultrasonic image of the scanning part of the detection object; a memory, storing a computer program; a processor, used to run the computer program, and the computer program executes an ultrasonic scanning feedback method when it is running; a display component, which displays accordingly according to the steps that need to be displayed in the ultrasonic scanning feedback method; the ultrasonic scanning feedback method includes: real-time acquisition and display of the ultrasonic image of the scanning part of the detection object; distinguishing and displaying the parts that have been scanned; identifying whether there is a lesion in the real-time acquired ultrasonic image, and if so, distinguishing and displaying the lesion area. The present invention provides feedback on the ultrasonic image and / or operator acquired by the ultrasonic probe to avoid missed detection and improve the efficiency of ultrasonic scanning.

[0004] However, during the actual implementation, the inventors found that during the actual scanning process, due to factors such as the presence of implants, active organs and environmental vibrations in the patient's body, a certain amount of noise is easily generated during the ultrasonic scanning process, resulting in a decrease in imaging quality. Summary of the invention

[0005] In view of the above problems existing in the prior art, a medical image noise suppression and enhancement method based on artificial intelligence is now provided.

[0006] The specific technical solutions are as follows:

[0007] A medical image noise suppression and enhancement method based on artificial intelligence, comprising:

[0008] Step S1: collecting raw data of an input ultrasound image, and predicting a noise type according to the ultrasound image and the raw data;

[0009] Step S2: according to the noise type, calling a noise correction model of a corresponding type to process the input ultrasound image, so as to obtain a repaired image.

[0010] On the other hand, the step S1 comprises:

[0011] Step S11: identifying at least one group of defect areas from the input ultrasonic image;

[0012] Step S12: extracting an echo signal interval from the original data according to the position of the defect area in the input ultrasonic image;

[0013] Step S13: predicting the noise type according to the echo signal interval and the defect area.

[0014] On the other hand, the step S11 comprises:

[0015] Step A11: traversing each pixel unit in the input ultrasound image using a sliding window unit to obtain an adjacent unit associated with the pixel unit;

[0016] Step A12: Calculate the target grayscale value of the pixel unit and the average grayscale value of the adjacent units respectively;

[0017] Step A13: comparing the target grayscale value and the average grayscale value with a pre-calibrated grayscale variation range, thereby selecting the pixel unit having a large deviation between the target grayscale value and the average grayscale value as the defective area.

[0018] On the other hand, the step S11 further includes:

[0019] A wavelet transform is performed on the input ultrasound image to extract artifact components, and the artifact components are marked as the defect area.

[0020] On the other hand, in step S13, a defect classification model is used to predict the noise type, and the defect classification model includes:

[0021] A signal analysis module, wherein the signal analysis module separates a noise signal from the echo signal interval according to the echo signal interval and an adjacent interval of the echo signal interval;

[0022] A convolutional network, wherein the convolutional network extracts the defect area to obtain high-dimensional image features;

[0023] A classification module, wherein the classification module is connected to the signal analysis module and the convolutional network respectively, and the classification module predicts the noise type according to the noise signal and the high-dimensional image features.

[0024] On the other hand, the signal analysis module comprises:

[0025] A neighboring area extraction module, wherein the neighboring area extraction module obtains the adjacent interval from the original data according to the echo signal interval;

[0026] A Fourier transform module, wherein the Fourier transform module is connected to the neighboring area extraction module, and the Fourier transform module performs Fourier transform on the adjacent interval and the echo signal interval to obtain time-frequency domain data;

[0027] A noise extraction module, wherein the noise extraction module is connected to the Fourier transform module, and the noise extraction module compares the time-frequency domain data of the adjacent interval with the time-frequency domain data of the echo signal interval in the frequency domain to obtain the noise signal.

[0028] On the other hand, in step S13, the classification module uses a random forest model to classify the noise signal to obtain the noise type.

[0029] On the other hand, when the noise type is a vibration artifact, the artifact noise elimination model is called in step S3 to eliminate the vibration artifact;

[0030] The artifact noise elimination model includes:

[0031] An artifact extraction module, wherein the artifact extraction module extracts the artifact noise corresponding to the defect area to obtain an artifact area;

[0032] An artifact removal module is connected to the artifact extraction module, and the artifact removal module removes artifacts from the input ultrasound image according to the artifact area to obtain the repaired image.

[0033] On the other hand, when the noise type is reflection noise, the reflection noise elimination model is called in step S3 to eliminate the reflection noise;

[0034] The reflection noise elimination model includes:

[0035] An echo separation module, wherein the echo separation module separates the reflection noise and the imaging echo from the echo signal interval according to the scanned tissue type;

[0036] A reconstruction module, the reconstruction module is connected to the echo separation module, and the reconstruction module obtains a reconstruction area according to the imaging echo reconstruction;

[0037] A repair module is connected to the reconstruction module, and the repair module corrects the input ultrasound image according to the reconstruction area to obtain the repaired image.

[0038] The above technical solution has the following advantages or beneficial effects:

[0039] In view of the problem that the ultrasonic scanning equipment in the prior art is prone to imaging degradation due to noise components during the imaging process, this solution predicts the original data during the imaging process, identifies the noise components and corresponding noise types in the data, and calls the corresponding noise correction model according to the noise type to correct the corresponding area, thereby achieving noise removal. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The embodiments of the present invention will be described more fully with reference to the attached drawings, which are provided for illustration and description only and are not intended to limit the scope of the present invention.

[0041] Figure 1 is an overall schematic diagram of an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of step S1 in an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of step A11 in an embodiment of the present invention;

[0044] Figure 4 Schematic diagram of a defect classification model in an embodiment of the present invention;

[0045] Figure 5 This is a schematic diagram of a signal analysis module in an embodiment of the present invention;

[0046] Figure 6 Schematic diagram of an artifact noise elimination model in an embodiment of the present invention;

[0047] Figure 7 Schematic diagram of a reflection noise elimination model in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0050] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0051] The present invention comprises:

[0052] A medical image noise suppression and enhancement method based on artificial intelligence, such as Figure 1 As shown, including:

[0053] Step S1: collecting raw data of an input ultrasound image, and predicting the noise type according to the ultrasound image and the raw data;

[0054] Step S2: according to the noise type, a noise correction model of a corresponding type is called to process the input ultrasound image, thereby obtaining a repaired image.

[0055] Specifically, in order to address the problem that ultrasonic scanning equipment in the prior art is prone to imaging degradation due to noise components during the imaging process, this solution predicts the original data during the imaging process, identifies the noise components and corresponding noise types in the data, and calls the corresponding noise correction model according to the noise type to correct the corresponding area, thereby achieving noise removal.

[0056] In one embodiment, Figure 2 As shown, step S1 includes:

[0057] Step S11: identifying at least one group of defect areas from the input ultrasonic image;

[0058] Step S12: intercepting the echo signal interval of the original data according to the position of the defect area in the input ultrasonic image;

[0059] Step S13: predicting the noise type according to the echo signal interval and defect area.

[0060] Specifically, in order to achieve better extraction and prediction of noise components in the image, in this embodiment, the input ultrasound image is first traversed to extract the defective area in the input ultrasound image that affects the image quality, and then the echo sequence collected during the imaging process is reversely searched according to the position of the defective area in the input ultrasound image, that is, the pixel coordinates, so as to intercept the echo signal interval at the corresponding position in the original data. By combining the two, the original data features in the original imaging process can be extracted, so that the model can better predict the type, cause and intensity of the noise component, which is convenient for subsequent correction.

[0061] In one embodiment, Figure 3 As shown, step S11 includes:

[0062] Step A11: traverse each pixel unit in the input ultrasound image using a sliding window unit to obtain an adjacent unit associated with the pixel unit;

[0063] Step A12: Calculate the target grayscale value of the pixel unit and the average grayscale value of the adjacent units respectively;

[0064] Step A13: Compare the target grayscale value and the average grayscale value with the pre-calibrated grayscale variation range, so as to select the pixel units with a large deviation between the target grayscale value and the average grayscale value as defective areas.

[0065] Specifically, in order to effectively extract the local defect area in the image, in this embodiment, for the input ultrasound image, a sliding window unit with a certain size is used to traverse each pixel unit to obtain an adjacent unit associated with the pixel unit, such as a 7*7 sliding window, the center position of the sliding window is the currently traversed pixel unit, and the other parts of the sliding window are adjacent units. Then, the target gray value of the pixel unit and the average gray value of the adjacent unit are calculated respectively.

[0066] Normally, the noise caused by noise components in the imaging process will be reflected in the image as a grayscale shift of a single pixel or a small number of pixels, while the grayscale change process of a normal pixel unit is continuous. Therefore, by calculating the average grayscale value of adjacent units and setting the grayscale change range, and then judging whether the target grayscale value deviates from the grayscale change range based on the average grayscale value, the pixel unit corresponding to the noise part is marked.

[0067] In one embodiment, step S11 further includes:

[0068] Wavelet transform is performed on the input ultrasound image to extract artifact components, and the artifact components are marked as defect areas.

[0069] Specifically, in order to achieve better extraction of noise in the artifact part, in this embodiment, the input ultrasound image is also subjected to wavelet transformation to extract the artifact component, and the artifact component is marked as a defective area, thereby extracting the artifact caused by factors such as vibration in the image.

[0070] In one embodiment, in step S13, a defect classification model is used to predict the noise type, such as Figure 4 As shown, the defect classification model includes:

[0071] A signal analysis module 11, the signal analysis module 11 separates the noise signal from the echo signal interval according to the echo signal interval and the adjacent interval of the echo signal interval;

[0072] Convolutional network 12, the convolutional network 12 extracts the defect area to obtain high-dimensional image features;

[0073] The classification module 13 is connected to the signal analysis module 11 and the convolution network 12 respectively. The classification module 13 predicts the noise type according to the noise signal and the high-dimensional image features.

[0074] Specifically, in order to achieve a better defect classification effect, in this embodiment, for the located defect area, a series of convolution kernels of different sizes in the convolution network 12 are first used to extract the defect area to obtain high-dimensional image features; at the same time, the extracted echo signal interval corresponding to the defect area is compared with its adjacent intervals in time sequence to separate the noise signal. Finally, the classification module 13 predicts the noise type according to the noise signal and the high-dimensional image features.

[0075] In order to achieve a better classification effect, a random forest model can be used as a classification module 13 to implement the classification process. Since both the noise signal and the high-dimensional image features are integrated in the prediction process, a better prediction effect can be achieved.

[0076] In one embodiment, Figure 5 As shown, the signal analysis module 11 includes:

[0077] A neighboring area extraction module 111, the neighboring area extraction module 111 extracts the neighboring area from the original data according to the echo signal interval;

[0078] Fourier transform module 112, the Fourier transform module 112 is connected to the neighboring area extraction module 111, and the Fourier transform module 112 performs Fourier transform on the adjacent interval and the echo signal interval to obtain time-frequency domain data;

[0079] The noise extraction module 113 is connected to the Fourier transform module 112. The noise extraction module 113 compares the time-frequency domain data of the adjacent interval with the time-frequency domain data of the echo signal interval in the frequency domain to obtain a noise signal.

[0080] Specifically, to achieve better noise signal extraction effect, in this embodiment, the neighboring area extraction module 111 first intercepts specific lengths forward and backward in the time direction according to the position of the echo signal interval in the original data, thereby intercepting the adjacent interval.

[0081] Then, the Fourier transform module 112 performs Fourier transform on the adjacent interval and the echo signal interval to obtain time-frequency domain data.

[0082] For the time-frequency domain data obtained by the transformation, the noise extraction module 113 compares the time-frequency domain data of the adjacent interval with the time-frequency domain data of the echo signal interval in the frequency domain to obtain a noise signal.

[0083] In one embodiment, when the noise type is a vibration artifact, in step S3, an artifact noise elimination model is called to eliminate the vibration artifact;

[0084] like Figure 6 As shown, the artifact noise removal model includes:

[0085] An artifact extraction module 21, the artifact extraction module 21 extracts the artifact noise corresponding to the defect area to obtain an artifact area;

[0086] The artifact removal module 22 is connected to the artifact extraction module 21 . The artifact removal module 22 removes artifacts from the input ultrasound image according to the artifact area to obtain a restored image.

[0087] Specifically, when it is determined that the noise type is an artifact caused by vibration, the present embodiment calls an artifact noise elimination model to eliminate vibration artifacts. The artifact noise elimination model includes an artifact extraction module 21, which identifies and segments the artifact noise portion corresponding to the defect area, thereby marking the artifact area in the image. Subsequently, the artifact elimination module 22 removes artifacts from the input ultrasound image according to the artifact area to obtain a repaired image.

[0088] In one embodiment, when the noise type is reflection noise, a reflection noise elimination model is called in step S3 to eliminate the reflection noise;

[0089] like Figure 7 As shown, the reflection noise elimination model includes:

[0090] An echo separation module 31, the echo separation module 31 separates the reflection noise and the imaging echo from the echo signal interval according to the scanned tissue type;

[0091] A reconstruction module 32, the reconstruction module 32 is connected to the echo separation module 31, and the reconstruction module 32 obtains a reconstruction area according to the imaging echo reconstruction;

[0092] The repair module 33 is connected to the reconstruction module 32, and the repair module 33 corrects the input ultrasound image according to the reconstruction area to obtain a repaired image.

[0093] Specifically, in order to effectively remove the noise part, in this embodiment, for the frequency bands of reflected clutter that are easily generated by different scanned tissue types, the echo signal interval is further analyzed in the time-frequency domain to obtain the reflected noise and imaging echo. Then the reconstruction module 32 reconstructs the imaging echo to obtain the reconstructed area, and the repair module 33 replaces the corresponding position in the input ultrasound image according to the reconstructed area to obtain the repaired image.

[0094] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A medical image noise suppression and enhancement method based on artificial intelligence, characterized in that: include: Step S1: collecting raw data of an input ultrasound image, and predicting a noise type according to the ultrasound image and the raw data; Step S2: according to the noise type, calling a noise correction model of a corresponding type to process the input ultrasound image, so as to obtain a repaired image.

2. The medical image noise suppression and enhancement method according to claim 1, characterized in that: The step S1 comprises: Step S11: identifying at least one group of defect areas from the input ultrasonic image; Step S12: extracting an echo signal interval from the original data according to the position of the defect area in the input ultrasonic image; Step S13: predicting the noise type according to the echo signal interval and the defect area.

3. The medical image noise suppression and enhancement method according to claim 2, characterized in that: The step S11 comprises: Step A11: traversing each pixel unit in the input ultrasound image using a sliding window unit to obtain an adjacent unit associated with the pixel unit; Step A12: Calculate the target grayscale value of the pixel unit and the average grayscale value of the adjacent units respectively; Step A13: comparing the target grayscale value and the average grayscale value with a pre-calibrated grayscale variation range, thereby selecting the pixel unit having a large deviation between the target grayscale value and the average grayscale value as the defective area.

4. The medical image noise suppression and enhancement method according to claim 2, characterized in that: The step S11 further includes: A wavelet transform is performed on the input ultrasound image to extract artifact components, and the artifact components are marked as the defect area.

5. The medical image noise suppression and enhancement method according to claim 2, characterized in that: In the step S13, a defect classification model is used to predict the noise type, and the defect classification model includes: A signal analysis module, wherein the signal analysis module separates a noise signal from the echo signal interval according to the echo signal interval and an adjacent interval of the echo signal interval; A convolutional network, wherein the convolutional network extracts the defect area to obtain high-dimensional image features; A classification module, wherein the classification module is connected to the signal analysis module and the convolutional network respectively, and the classification module predicts the noise type according to the noise signal and the high-dimensional image features.

6. The medical image noise suppression and enhancement method according to claim 5, characterized in that: The signal analysis module comprises: A neighboring area extraction module, wherein the neighboring area extraction module obtains the adjacent interval from the original data according to the echo signal interval; A Fourier transform module, wherein the Fourier transform module is connected to the neighboring area extraction module, and the Fourier transform module performs Fourier transform on the adjacent interval and the echo signal interval to obtain time-frequency domain data; A noise extraction module, wherein the noise extraction module is connected to the Fourier transform module, and the noise extraction module compares the time-frequency domain data of the adjacent interval with the time-frequency domain data of the echo signal interval in the frequency domain to obtain the noise signal.

7. The medical image noise suppression and enhancement method according to claim 5, characterized in that: In the step S13, the classification module uses a random forest model to classify the noise signal to obtain the noise type.

8. The medical image noise suppression and enhancement method according to claim 2, characterized in that: When the noise type is a vibration artifact, in step S3, an artifact noise elimination model is called to eliminate the vibration artifact; The artifact noise elimination model includes: An artifact extraction module, wherein the artifact extraction module extracts the artifact noise corresponding to the defect area to obtain an artifact area; An artifact removal module is connected to the artifact extraction module, and the artifact removal module removes artifacts from the input ultrasound image according to the artifact area to obtain the repaired image.

9. The medical image noise suppression and enhancement method according to claim 2, characterized in that: When the noise type is reflection noise, in step S3, a reflection noise elimination model is called to eliminate the reflection noise; The reflection noise elimination model includes: An echo separation module, wherein the echo separation module separates the reflection noise and the imaging echo from the echo signal interval according to the scanned tissue type; A reconstruction module, the reconstruction module is connected to the echo separation module, and the reconstruction module obtains a reconstruction area according to the imaging echo reconstruction; A repair module is connected to the reconstruction module, and the repair module corrects the input ultrasound image according to the reconstruction area to obtain the repaired image.

Images