Image noise evaluation method and apparatus, electronic device, and medium

Abstract

The application discloses an image noise evaluation method and device, electronic equipment and medium. The image noise evaluation method comprises the following steps: performing partial noise extraction on a target medical image to obtain a target partial noise image; performing noise evaluation on the target partial noise image to obtain a target noise level; obtaining a reference noise level, wherein a plurality of candidate noise levels are obtained by respectively performing partial noise evaluation on a plurality of reference medical images containing different degrees of noise, and the reference noise level is determined based on the plurality of candidate noise levels; and evaluating the target noise level based on the reference noise level to obtain a noise evaluation result of the target medical image.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an image noise assessment method, apparatus, electronic device, and medium. Background Technology

[0002] Noise assessment in medical images is a crucial aspect of medical imaging quality control. Its necessity stems from a deep-seated contradiction between the physical nature of imaging technology and clinical practice: X-ray imaging exhibits granular noise due to photon statistical fluctuations; MRI generates Gaussian background noise due to the thermal motion of human tissue; and ultrasound images display mottled textures due to sound wave scattering. Furthermore, clinical demands such as low-dose CT scans and rapid MRI sequences further amplify noise levels. This noise not only masks minute calcifications or low-contrast lesions in early-stage tumors but can also induce artifacts such as flow artifacts and truncation artifacts, interfering with diagnosis. Moreover, it can lead to systematic biases in quantitative analysis based on image grayscale values, severely impacting the implementation of precision medicine.

[0003] Some noise assessment methods require an ideal noise-free reference image as an evaluation reference when assessing noise in the image to be evaluated. However, this is limited in clinical applications due to the difficulty in obtaining a noise-free reference image of the same plane as the patient. Some noise assessment methods require completely separating the noise in the image to obtain a complete noise map, and then measuring the noise on the complete noise map. However, in practice, it is difficult to achieve complete noise separation, resulting in inaccurate noise assessment. Summary of the Invention

[0004] The embodiments of this application aim to at least partially solve one of the technical problems in the related art. Therefore, the first objective of the embodiments of this application is to provide an image noise assessment method, apparatus, electronic device, medium, and program product.

[0005] This application provides an image noise assessment method, comprising: extracting partial noise from a target medical image to obtain a target partially noisy image; assessing the noise in the target partially noisy image to obtain a target noise level; obtaining a baseline noise level, wherein multiple candidate noise levels are obtained by performing partial noise assessments on multiple baseline medical images containing different levels of noise, and a baseline noise level is determined based on the multiple candidate noise levels; and assessing the target noise level based on the baseline noise level to obtain a noise assessment result for the target medical image.

[0006] In other embodiments, the reference noise level includes a low reference noise level and a high reference noise level; obtaining the reference noise level includes: performing partial noise extraction on multiple reference medical images containing different levels of noise to obtain multiple reference partially noisy images; performing noise evaluation on the multiple reference partially noisy images to obtain multiple candidate noise levels; and determining the low reference noise level and the high reference noise level based on the multiple candidate noise levels.

[0007] In other embodiments, the target noise level is evaluated based on a reference noise level to obtain a noise evaluation result for the target medical image. This includes: determining a first noise level value as the noise evaluation result when the target noise level is less than or equal to a low reference noise level; determining a second noise level value as the noise evaluation result when the target noise level is greater than or equal to a high reference noise level, wherein the second noise level value is greater than the first noise level value; and determining a third noise level value as the noise evaluation result when the target noise level is greater than the low reference noise level and less than the high reference noise level, wherein the third noise level value includes the ratio between a first noise difference and a second noise difference, the first noise difference being the difference between the target noise level and the low reference noise level, the second noise difference being the difference between the high reference noise level and the low reference noise level, and the third noise level value being greater than the first noise level value and less than the second noise level value.

[0008] In other embodiments, the target noise level is evaluated based on a reference noise level to obtain a noise evaluation result for the target medical image, including: updating the reference noise level based on the texture intensity difference between the target medical image and the reference medical image to obtain an updated reference noise level; and evaluating the target noise level based on the updated reference noise level to obtain a noise evaluation result for the target medical image.

[0009] In other embodiments, the baseline noise level is updated based on the texture intensity difference between the target medical image and the baseline medical image to obtain an updated baseline noise level. This includes: determining a target difference image between the target medical image and a target partial noise image, and performing texture detection and evaluation on the target difference image to obtain a target texture intensity; determining a baseline difference image between the baseline medical image and a baseline partial noise image, and performing texture detection and evaluation on the baseline difference image to obtain a baseline texture intensity, wherein the baseline partial noise image is obtained by partially extracting noise from the baseline medical image; obtaining a texture intensity adjustment level based on the difference between the target texture intensity and the baseline texture intensity and an adjustment coefficient, and adding the texture intensity adjustment level to the baseline noise level to obtain the updated baseline noise level.

[0010] In other embodiments, the reference noise level includes a low reference noise level and a high reference noise level; the reference medical image includes a first reference medical image corresponding to the low reference noise level and a second reference medical image corresponding to the high reference noise level; determining a reference difference image between the reference medical image and the reference partial noise image, and performing texture detection and evaluation on the reference difference image to obtain a reference texture intensity, includes: determining a first reference difference image between the first reference medical image and the first reference partial noise image, and performing texture detection and evaluation on the first reference difference image to obtain a first reference texture intensity, wherein the first reference partial noise image is obtained by partially extracting noise from the first reference medical image; determining a second reference difference image between the second reference medical image and the second reference partial noise image, and performing texture detection and evaluation on the second reference difference image to obtain a second reference texture intensity, wherein the second reference partial noise image is obtained by partially extracting noise from the second reference medical image.

[0011] In other embodiments, a texture intensity adjustment level is obtained based on the difference between the target texture intensity and the reference texture intensity and an adjustment coefficient. The texture intensity adjustment level is then added to a reference noise level to obtain an updated reference noise level. This includes: obtaining a first texture intensity adjustment level based on the difference between the target texture intensity and the first reference texture intensity and a first adjustment coefficient, and adding a low reference noise level to the first texture intensity adjustment level to obtain an updated low reference noise level; obtaining a second texture intensity adjustment level based on the difference between the target texture intensity and the second reference texture intensity and a second adjustment coefficient, and adding a high reference noise level to the second texture intensity adjustment level to obtain an updated high reference noise level.

[0012] In other embodiments, texture detection and evaluation are performed on the target difference image to obtain the target texture intensity, including: performing edge detection on the target difference image to obtain a target edge detection image, and performing partial noise extraction and evaluation on the target edge detection image to obtain the target texture intensity; texture detection and evaluation are performed on the reference difference image to obtain the reference texture intensity, including: performing edge detection on the reference difference image to obtain a reference edge detection image, and performing partial noise extraction and evaluation on the reference edge detection image to obtain the reference texture intensity.

[0013] This application provides an image noise assessment device, including an extraction module for extracting partial noise from a target medical image to obtain a target partially noisy image; a first assessment module for assessing the noise in the target partially noisy image to obtain a target noise level; an acquisition module for acquiring a baseline noise level, wherein multiple candidate noise levels are obtained by performing partial noise assessments on multiple baseline medical images containing different levels of noise, and a baseline noise level is determined based on the multiple candidate noise levels; and a second assessment module for assessing the target noise level based on the baseline noise level to obtain a noise assessment result for the target medical image.

[0014] This application provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the above embodiments.

[0015] This application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method of any of the above embodiments.

[0016] This application provides a computer program product that includes instructions that, when executed by a processor of a computer device, enable the computer device to perform the steps of the method described in any of the above embodiments. Attached Figure Description

[0017] Figure 1 A schematic diagram of the image noise evaluation method provided in the embodiments of this application.

[0018] Figure 2 A schematic diagram illustrating part of the noise extraction provided for an embodiment of this application.

[0019] Figure 3 This is a schematic diagram of an image noise evaluation method provided in another embodiment of this application.

[0020] Figure 4 This is a schematic diagram illustrating the establishment of a noise level reference for an embodiment of this application.

[0021] Figure 5 A schematic diagram of an image noise evaluation apparatus provided for an embodiment of this application.

[0022] Figure 6 A block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0023] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0024] In noise assessment methods, reference image-based assessment systems achieve precise quantification by constructing an ideal noise-free reference image and analyzing the differences between the image to be assessed. The mean squared error (MSE) reflects the overall distortion level by calculating the squared mean of pixel-level grayscale differences; the peak signal-to-noise ratio (PSNR) converts it to a logarithmic scale to conform to human visual perception characteristics; and the structural similarity index (SSIM) simulates the multi-channel decomposition mechanism of the human visual system from three dimensions: brightness, contrast, and structure. These methods achieve high assessment consistency in standard model validation, but their clinical application is limited by the difficulty in obtaining noise-free reference images of the same level as the patient.

[0025] For clinical scenarios without reference images, assessment methods have shifted towards in-depth analysis of the intrinsic statistical properties of images. Local variance analysis estimates noise intensity by calculating the pixel fluctuation range of homogeneous regions (such as cerebrospinal fluid or adipose tissue), but it is susceptible to partial volume effects. Spatial domain model fitting utilizes Rayleigh and gamma distributions to describe the noise probability density function, while frequency domain analysis separates signal and noise frequency bands using Fourier transform or wavelet transform, and assesses noise based on its energy proportion. Blind assessment methods based on deep learning automatically learn noise characteristics through neural networks to achieve noise separation and subsequently noise measurement. While these reference-free methods eliminate the dependence on ideal references, they face the dual challenges of tissue heterogeneity and the complexity of pathological structures, impacting their accuracy and applicability.

[0026] In some examples, a deep neural network model with CT image denoising capabilities is obtained through training a classification network. This denoising model is used to denoise the image to be evaluated, achieving separation between the denoised image and the image noise. Based on this, noise assessment with a reference image is approximated. The drawback is that the denoising capability of this model is affected by various factors such as the scanning equipment model, scanning parameters, reconstruction parameters, scanning location, and patient characteristics. High accuracy and generalization of the denoising model are required, and in practical applications, the generated pseudo-reference image may fail to denoise or fall short of expectations, thus affecting the accuracy and reliability of noise assessment. Furthermore, this method requires the use of two neural networks for related calculations, limiting computational speed and hardware performance. For noise assessment of large batches of image data, its computational performance needs improvement.

[0027] In some examples, the image to be evaluated is segmented to obtain foreground, background, and neglected regions, and noise is calculated and evaluated using the foreground and background regions. However, using the pixel values ​​of the foreground region to calculate noise intensity assumes that the pixel values ​​of the foreground region are basically uniformly distributed, but the foreground region is often not uniform, and its structural features may be misidentified as noise components, leading to noise estimation bias.

[0028] Figure 1 A schematic diagram of the image noise evaluation method provided in the embodiments of this application.

[0029] like Figure 1 As shown, the image noise evaluation method 100 provided in this application includes, for example, steps S110-S140.

[0030] Step S110: Extract partial noise from the target medical image to obtain a target partial noise image.

[0031] The target medical image, such as a CT image to be evaluated, is difficult to accurately determine due to the difficulty in obtaining its noise components. Therefore, it is challenging to effectively separate the structural content from the noise components in the image being evaluated. This application addresses this by extracting a portion of the noise from the target medical image to obtain a target partial noise image. The method for extracting the partial noise image can be expressed as follows: F (*), specifically, it can be based on Gaussian filtering, median filtering, wavelet decomposition, neural network noise learning, etc., to extract some noise components in the image to be evaluated, i.e., formula (1): I n = F ( I (1) in, I Represents the target medical image. I n This represents a noisy portion of the target image.

[0032] Since the extracted noise component is likely only a part of all the noise, it is called... I n This is a noisy image of the target area. For example... F (*) indicates that when using median high-pass filtering, the extraction is based on the high-pass filtering method. I n Includes target medical images I Part of the high-frequency noise, I n It may also contain targeted medical images. I Partial structural information in the text.

[0033] Figure 2This is a schematic diagram illustrating part of the noise extraction provided for an embodiment of this application. For example... Figure 2 As shown, for the target part of the noisy image obtained after high-pass filtering I n It contains target medical images I Partial boundary and texture information.

[0034] In another example, due to the target medical image I It may contain some interference areas, such as in Figure 2 Left-side target medical image I The background region outside the inscribed circle. This background region may interfere with noise assessment; therefore, image processing methods can be used to remove interfering regions such as bright or dark areas, thereby improving the generalization of the noise assessment method. For example, threshold segmentation can be used to remove interfering regions, since the target medical image... I The CT value for a neutral atmospheric background is around -1000. Figure 4 Left-side target medical image I The gray area within the circular region represents non-organ tissue and is therefore considered as atmospheric background. Therefore, the area below -1000 can be approximated as the interference region (i.e.,...). Figure 2 Left-side target medical image I The black area outside the inscribed circle is the interference area. This allows for the setting of a low threshold value. V min (-1000) can also be used to set a threshold for high-value regions. V max To exclude abnormally bright areas, thereby achieving the desired medical image I Perform region of interest extraction. For example, in a target medical image. I pixel value at any position in the region of interest V ROI(I) This can be expressed as formula (2): (2) in, ROI(I) Indicates the target medical image I The region of interest (ROI) image obtained by performing region of interest (ROI) identification. V I Represents target medical image I The pixel value at any position in the image is used to extract the target part of the noise image using formula (1). I n It can be written as formula (3): I n = F ( ROI(I) (3) Step S120: Perform noise assessment on the noisy image of the target portion to obtain the target noise level.

[0035] Obtain the noise image of the target part I n Then, the target portion of the noisy image can be processed. I n Incorporate noise assessment. For example, calculate the noise level of the target portion of the image. I n The approximate average noise level is used as the target noise level. It is called "approximate" because the target portion of the image is noisy. I n Includes target medical images I This includes structural information such as partial boundaries and textures. For example, in a target medical image... I The width, height, and number of layers can be expressed as follows: W, H, D The sampling step size in the three dimensions is... P W , P H , P D Using the sampling step size as the interval, in the noisy image of the target region... I n Up-extracted sampled image patch I P And calculate the image of each sampled block. I P noise level N IP Examples include standard deviation. The noise level of each sampling block is obtained. N IP Then, the average noise level of all sampled blocks is calculated as the target portion of the noise image. I n Target noise level N In .

[0036] Step S130: Obtain the baseline noise level, wherein multiple candidate noise levels are obtained by performing partial noise assessment on multiple baseline medical images containing different levels of noise, and the baseline noise level is determined based on the multiple candidate noise levels.

[0037] For example, multiple candidate noise levels are obtained by performing partial noise extraction and noise evaluation on multiple benchmark medical images containing different levels of noise. Then, a benchmark noise level is determined based on the multiple candidate noise levels. The benchmark noise level and the target noise level are obtained by the same principle, which is by extracting partial noise.

[0038] Step S140: Evaluate the target noise level based on the baseline noise level to obtain the noise evaluation result of the target medical image.

[0039] For example, the noise level of a target medical image can be assessed by comparing the target noise level with a baseline noise level.

[0040] The embodiments of this application perform noise assessment by extracting a portion of the noise level of an image. This eliminates the need for accurate separation of noise information in the image and the construction of an ideal noise-free reference image, greatly reducing the difficulty of noise assessment. Furthermore, this noise assessment method is computationally fast and can achieve full and rapid assessment of the image to be evaluated. The method has low requirements for noise extraction and does not necessarily require the completeness and purity of the extracted noise, thus exhibiting high applicability and reliability. This noise assessment method can conveniently, quickly, and stably achieve relative noise assessment of an image.

[0041] In one example, the reference noise level includes a low reference noise level. N min and high reference noise level N max Multiple benchmark medical images containing varying degrees of noise were analyzed. I 1 , I 2 , ……, I m Partial noise extraction was performed to obtain multiple baseline partial noise images. I n_1 , I n_2 , ……, I n_m .

[0042] Then, for multiple reference partial noise images respectively I n_1 , I n_2 , ……, I n_m Noise assessment was performed to obtain multiple candidate noise levels. N In_1 , N In_2 , ……, N In_m Then, based on multiple candidate noise levels N In_1 , N In_2 , ……, N In_m Determine the low reference noise level N min and high reference noise level N maxFor example, from multiple candidate noise levels N In_1 , N In_2 , ……, N In_m Two thresholds, high and low, are obtained as reference benchmarks for evaluating noise levels.

[0043] One approach is to select a low-noise (lowest or lowest) benchmark medical image. I i Corresponding candidate noise level N In_i As a low reference noise level N min Select a high-noise (highest or highest) benchmark medical image. I j Corresponding candidate noise level N In_j As a high reference noise level N max .

[0044] Another approach is to select multiple candidate noise levels corresponding to several low-noise benchmark medical images and calculate their average as the low benchmark noise level. N min Multiple candidate noise levels from several high-noise benchmark medical images are selected, and their average value is calculated as the high benchmark noise level. N max .

[0045] As can be seen, by extracting some noise and calculating the noise level from the benchmark medical image, a low benchmark noise level can be generated. N min and high reference noise level N max These serve as two benchmark thresholds for low noise and high noise.

[0046] It is understandable that, in order to reduce the bias of the noise benchmark and improve the accuracy of noise assessment, the method of extracting partial noise and calculating the noise level from the benchmark medical image differs from that used for the target medical image. I The method for extracting noise and calculating noise levels is the same.

[0047] In one example, the target noise level is evaluated based on a baseline noise level to obtain the noise assessment result for the target medical image. For the target medical image... I The target portion of the noisy image has been obtained. I n and the target part of the noisy image I n Target noise levelN In At this point, a low reference noise level is utilized. N min and high reference noise level N max As a noise level normalization parameter, for the target noise level N In Normalization is performed, as shown in formula (4): (4) in, N I Represents target medical image I The noise assessment results; N In Represents target medical image I Target noise level N In ; N min Indicates a low reference noise level; N max This indicates a high reference noise level.

[0048] At the target noise level N In Less than or equal to the low reference noise level N min In this case, the first noise level value is determined as the noise assessment result. N In In one example, the first noise level value can be 0, representing the target medical image. I The noise level is 0, which is a completely acceptable low noise level.

[0049] At the target noise level N In Greater than or equal to the high reference noise level N max In this case, the second noise level value is determined as the noise assessment result. N In In one example, the second noise level value is greater than the first noise level value. The second noise level value can be 1, representing the target medical image. I The noise level is 1, which contains unacceptably high noise.

[0050] At the target noise level N In Greater than the low reference noise level N min And less than the high reference noise level N max In this case, determine the third noise level value. As a result of the noise assessment, the third noise level value includes the first noise difference. Difference between the second and third noise levels The ratio between them. First noise difference. For target noise level N In With low reference noise level N min The difference between them. Second noise difference. For high reference noise level N max With low reference noise level N min The difference between them. Third noise level value. It is greater than the first noise level value of 0 and less than the second noise level value of 1.

[0051] According to embodiments of this application, the target medical image to be evaluated is... I Target noise level N In Reference low reference noise level N min and high reference noise level N max Normalization is performed (the noise assessment result is between 0 and 1) to obtain the target medical image to be evaluated. I Relative noise evaluation results N I .

[0052] In another example, the baseline noise level can be updated based on the difference in texture intensity between the target medical image and the baseline medical image to obtain the updated baseline noise level. Then, the target noise level is evaluated based on the updated baseline noise level to obtain the noise evaluation result of the target medical image.

[0053] When the target medical image and the baseline medical image belong to different modalities, or when multiple baseline medical images belong to different modalities, there are certain differences in structural or noise details between images of different modalities. Even when different images belong to the same modality, differences in image structure and noise details can occur due to scanning different human structures or using different scanning reconstruction parameters. To improve the accuracy of noise assessment, the baseline noise level can be updated and corrected.

[0054] For the target medical image to be evaluated I The target medical image can be analyzed using formula (1). I Partial noise extraction is performed to obtain the target partial noise image. I n and identify target medical imagesI and target part of the noisy image I n Target difference images I’ Alternatively, the target medical image can be processed using formula (3). I Region of Interest (ROI) Image Obtained by Region of Interest Identification ROI(I) and the image of the region of interest. ROI(I) Partial noise extraction is performed to obtain the target partial noise image. I n Then determine the region of interest in the image. ROI(I) and target part of the noisy image I n Target difference images I’ As shown in formula (5): I’ = I - I n or I’ = ROI(I) - I n (5) Because the intensity and density of texture gradient changes in structural regions vary in images from different locations or with different scanning parameters, if the target medical image... I The structure has dense texture or large gradient changes, such as in target medical images. I The texture intensity is higher than that in the benchmark medical image, in the noisy image of the target region. I n The more residual structural textures are in the image, the more noisy the target area becomes. I n The noise assessment level is higher than the actual level. Conversely, if the target medical image... I The texture is dense or the gradient variation is small, such as in target medical images. I The texture intensity is lower than that in the baseline medical image, in the noisy image of the target region. I n The less residual structural texture there is, the less noise in the target part of the image. I n The noise assessment level is lower than the actual level.

[0055] To extend the application of a baseline noise level obtained from a specific region to other regions, a target difference image after noise extraction is introduced. I’ Corresponding target texture intensity S I’ This achieves adaptive adjustment of the reference noise level, enabling it to automatically scale to images with different scanning locations and parameters. Target texture intensity.S I’ This can be represented as a target difference image. I’ The intensity of structural edge variation.

[0056] For example, to obtain a target difference image I’ Next, the target difference image I’ Perform texture detection and evaluation to obtain the target texture intensity. S I’ For example, for target difference images I’ Perform edge detection to obtain the target edge detection image. I’ b The edge detection operator is represented as F b (*), edge detection operator F b (*) For example, operators such as Canny and Sobel; then, based on methods for extracting parts of the noisy image. F (*) Target edge detection image I’ b Partial noise extraction is performed to obtain the target texture intensity image. I’ bp ， As shown in formula (6). Then... I’ bp As a partially noisy image, the noise intensity, i.e. the target texture intensity, is calculated based on the same partially noisy evaluation method as described above. S I’ .

[0057] I’ bp = F ( F b ( I’ )) = F ( I’ b (6) Based on the same principle, after partially extracting noise from the benchmark medical image to obtain the benchmark partial noise image, the benchmark difference image between the benchmark medical image and the benchmark partial noise image is determined, and texture detection and evaluation are performed on the benchmark difference image to obtain the benchmark texture intensity. For example, edge detection is performed on the benchmark difference image to obtain the benchmark edge detection image, and partial noise extraction and evaluation are performed on the benchmark edge detection image to obtain the benchmark texture intensity. The specific principle is similar to that of formula (6).

[0058] In one example, the reference texture intensity includes the first reference texture intensity. S min Second reference texture intensity Smax The reference noise level includes the low reference noise level. N min and high reference noise level N max First baseline texture intensity S min With low reference noise level N min Correspondingly, the second baseline texture intensity S max With high reference noise level N max correspond.

[0059] For example, benchmark medical images include those with low benchmark noise levels. N min Corresponding first benchmark medical image I i And including with high reference noise levels N max Corresponding second benchmark medical image I j .

[0060] Determine the first benchmark medical image I i The first reference difference image is obtained between the first reference noise image and the first reference partial noise image. Texture detection and evaluation are performed on the first reference difference image to obtain the first reference texture intensity. S min The first reference portion of the noisy image is a first reference medical image. I i Obtained by partial noise extraction.

[0061] Determine the second benchmark medical image I j The second reference difference image is obtained by comparing the second reference noise image with the second reference part, and texture detection and evaluation are performed on the second reference difference image to obtain the second reference texture intensity. S max The second reference portion of the noisy image is a subset of the second reference medical image. I j Obtained by partial noise extraction.

[0062] Among them, the first reference texture intensity S min and low reference noise level N min With the same reference image source, the second reference texture intensity S max and high reference noise level N maxThey share the same baseline image source.

[0063] In one example, after obtaining the target texture intensity and the reference texture intensity, the first reference texture intensity is used as the basis for... S min Second reference texture intensity S max For the original low reference noise level N min and the original high reference noise level N max Make corrections. For example, the texture intensity adjustment level can be obtained based on the difference between the target texture intensity and the reference texture intensity and the adjustment coefficient, and the reference noise level can be added to the texture intensity adjustment level to obtain the updated reference noise level, as shown in formulas (7) and (8): N ' min = N min + α * ( S I’ - S min (7) N ' max = N max + β * ( S I’ - S max (8) As shown in formula (7), based on the target texture intensity S I’ and the first reference texture intensity S min The difference between them ( S I’ - S min ) and the first adjustment factor α The first texture intensity adjustment level is obtained. α * ( S I’ - S min And adjust the first texture intensity level. α * ( S I’ - S min Plus low reference noise level N min The updated low reference noise level was obtained. N ' minFirst adjustment coefficient α These are fine-tuning parameters used to adjust the strength of the threshold.

[0064] As shown in formula (8), based on the target texture intensity S I’ Second reference texture intensity S max The difference between them ( S I’ - S max ) and second adjustment coefficient β The second texture intensity adjustment level is obtained. β * ( S I’ - S max And adjust the second texture intensity level. β * ( S I’ - S max In addition to a high reference noise level N max The updated high reference noise level was obtained. N ' max Second adjustment coefficient β These are fine-tuning parameters used to adjust the strength of the threshold.

[0065] In one example, the updated low baseline noise level is obtained. N min 'and updated high reference noise level N max 'Then, based on the updated low baseline noise level' N ' min and the updated high reference noise level N ' max To evaluate the noise level of the target, the noise evaluation result of the target medical image is obtained, as shown in formula (9): (9) in, N I Represents target medical image I The noise assessment results; N In Represents target medical image I Target noise level N In .

[0066] At the target noise level N In Less than or equal to the updated low reference noise level N 'min In this case, the first noise level value is determined as the noise assessment result. N In In one example, the first noise level value can be 0, representing the target medical image. I The noise level is 0, which is a completely acceptable low noise level.

[0067] At the target noise level N In Greater than or equal to the updated high reference noise level N ' max In this case, the second noise level value is determined as the noise assessment result. N In In one example, the second noise level value is greater than the first noise level value. The second noise level value can be 1, representing the target medical image. I The noise level is 1, which contains unacceptably high noise.

[0068] At the target noise level N In Greater than the updated low reference noise level N ' min And less than the updated high reference noise level N ' max In this case, a third noise level value is determined as the noise assessment result. The third noise level value includes the ratio between the first noise difference and the second noise difference, where the first noise difference is the target noise level. N In Compared with the updated low reference noise level N ' min The difference between them, the second noise difference is the updated high reference noise level. N ' max Compared with the updated low reference noise level N ' min The difference between the two values ​​is such that the third noise level value is greater than the first noise level value (0) and less than the second noise level value (1).

[0069] This application corrects the noise baseline threshold (baseline noise level) based on the texture intensity of the image structure to achieve adaptive noise estimation, thereby enabling adaptive noise estimation for images with different structures and scanning parameters to adapt to different scenarios.

[0070] Figure 3 This is a schematic diagram of an image noise evaluation method provided in another embodiment of this application.

[0071] like Figure 3As shown, the image to be evaluated (target medical image) is acquired, and a portion of the noise in the image to be evaluated is extracted to obtain a partially noisy image (target partially noisy image). The average noise level (target noise level) of the partially noisy image (target partially noisy image) is calculated. Then, based on the established noise level benchmark (benchmark noise level), the relative noise level of the average noise level (target noise level) is calculated to obtain the noise evaluation result of the image to be evaluated (target medical image).

[0072] Figure 4 This is a schematic diagram illustrating the establishment of a noise level reference for an embodiment of this application.

[0073] like Figure 4 As shown, images containing different levels of noise (multiple benchmark medical images) are acquired in batches. Each benchmark medical image is used as the image to be evaluated, and part of the noise in the image to be evaluated is extracted to obtain a partially noisy image (benchmark partially noisy image). The average noise level (candidate noise level) of the partially noisy image (benchmark partially noisy image) is calculated. The average noise of a certain low-level noise image is selected as the lowest noise threshold (low benchmark noise level), and the average noise of a certain high-level noise image is selected as the highest noise threshold (high benchmark noise level) to complete the establishment of the noise level benchmark.

[0074] In another example, the image noise assessment method can be executed by an image noise assessment system, which includes an input module, a noise extraction module, a noise calculation module, a noise benchmark acquisition module, and a noise assessment module.

[0075] The input module is used to acquire the target medical image.

[0076] The noise extraction module is used to extract partial noise from the target medical image to obtain a target partial noise image.

[0077] The noise calculation module is used to evaluate the noise level of the target portion of the image and obtain the target noise level.

[0078] The noise baseline acquisition module acquires multiple baseline medical images. A noise extraction module then extracts partial noise from each of these baseline medical images to obtain multiple baseline noisy images. A noise calculation module then evaluates the noise in each of these baseline noisy images to obtain multiple candidate noise levels. Finally, a baseline noise level is determined based on these candidate noise levels. The baseline noise level includes a minimum noise threshold and a maximum noise threshold, which serve as standards for measuring the severity of image noise.

[0079] The noise assessment module is used to assess the target noise level based on the baseline noise level, and obtain the noise assessment result of the target medical image. The obtained noise assessment result serves as an indicator reflecting the noise level of the target medical image.

[0080] Figure 5 A schematic diagram of an image noise evaluation apparatus provided for an embodiment of this application.

[0081] Please see Figure 5 The image noise assessment device 500 includes...

[0082] The extraction module 510 is used to extract some noise from the target medical image to obtain a target part noise image.

[0083] The first evaluation module 520 is used to evaluate the noise in the target part of the noisy image to obtain the target noise level.

[0084] The acquisition module 530 is used to acquire the baseline noise level, wherein multiple candidate noise levels are obtained by performing partial noise assessment on multiple baseline medical images containing different levels of noise, and the baseline noise level is determined based on the multiple candidate noise levels.

[0085] The second evaluation module 540 is used to evaluate the target noise level based on the baseline noise level to obtain the noise evaluation result of the target medical image.

[0086] For example, the reference noise level includes a low reference noise level and a high reference noise level; the acquisition module 530 is further configured to: extract partial noise from multiple reference medical images containing different levels of noise to obtain multiple reference partial noise images; evaluate the noise of the multiple reference partial noise images to obtain multiple candidate noise levels; and determine the low reference noise level and the high reference noise level based on the multiple candidate noise levels.

[0087] For example, the second evaluation module 540 is further configured to: determine a first noise level value as a noise evaluation result when the target noise level is less than or equal to the low reference noise level; determine a second noise level value as a noise evaluation result when the target noise level is greater than or equal to the high reference noise level, wherein the second noise level value is greater than the first noise level value; and determine a third noise level value as a noise evaluation result when the target noise level is greater than the low reference noise level and less than the high reference noise level, wherein the third noise level value includes the ratio between the first noise difference and the second noise difference, the first noise difference being the difference between the target noise level and the low reference noise level, the second noise difference being the difference between the high reference noise level and the low reference noise level, and the third noise level value being greater than the first noise level value and less than the second noise level value.

[0088] For example, the second evaluation module 540 further includes: an update submodule, used to update the baseline noise level based on the texture intensity difference between the target medical image and the baseline medical image, to obtain an updated baseline noise level; and an evaluation submodule, used to evaluate the target noise level based on the updated baseline noise level, to obtain a noise evaluation result for the target medical image.

[0089] For example, the update submodule is further configured to: determine a target difference image between the target medical image and the target partial noise image, and perform texture detection and evaluation on the target difference image to obtain the target texture intensity; determine a benchmark difference image between the benchmark medical image and the benchmark partial noise image, and perform texture detection and evaluation on the benchmark difference image to obtain the benchmark texture intensity, wherein the benchmark partial noise image is obtained by partially extracting noise from the benchmark medical image; and obtain a texture intensity adjustment level based on the difference between the target texture intensity and the benchmark texture intensity and an adjustment coefficient, and add the texture intensity adjustment level to the benchmark noise level to obtain the updated benchmark noise level.

[0090] For example, the reference noise level includes a low reference noise level and a high reference noise level; the reference medical image includes a first reference medical image corresponding to the low reference noise level and a second reference medical image corresponding to the high reference noise level; determining a reference difference image between the reference medical image and the reference partial noise image, and performing texture detection and evaluation on the reference difference image to obtain a reference texture intensity, includes: determining a first reference difference image between the first reference medical image and the first reference partial noise image, and performing texture detection and evaluation on the first reference difference image to obtain a first reference texture intensity, wherein the first reference partial noise image is obtained by partially extracting noise from the first reference medical image; determining a second reference difference image between the second reference medical image and the second reference partial noise image, and performing texture detection and evaluation on the second reference difference image to obtain a second reference texture intensity, wherein the second reference partial noise image is obtained by partially extracting noise from the second reference medical image.

[0091] For example, a texture intensity adjustment level is obtained based on the difference between the target texture intensity and the reference texture intensity and an adjustment coefficient, and a reference noise level is added to the texture intensity adjustment level to obtain an updated reference noise level, including: obtaining a first texture intensity adjustment level based on the difference between the target texture intensity and the first reference texture intensity and a first adjustment coefficient, and adding a low reference noise level to the first texture intensity adjustment level to obtain an updated low reference noise level; obtaining a second texture intensity adjustment level based on the difference between the target texture intensity and the second reference texture intensity and a second adjustment coefficient, and adding a high reference noise level to the second texture intensity adjustment level to obtain an updated high reference noise level.

[0092] For example, performing texture detection and evaluation on a target difference image to obtain target texture intensity includes: performing edge detection on the target difference image to obtain a target edge detection image, and performing partial noise extraction and evaluation on the target edge detection image to obtain target texture intensity; performing texture detection and evaluation on a reference difference image to obtain reference texture intensity includes: performing edge detection on the reference difference image to obtain a reference edge detection image, and performing partial noise extraction and evaluation on the reference edge detection image to obtain reference texture intensity.

[0093] It is understood that a detailed description of the image noise evaluation device 500 can be found in the description of the image noise evaluation method above, and will not be repeated here.

[0094] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0095] One embodiment of this application provides a computer program product including instructions that, when executed by a processor of a computer device, enable the computer device to perform the steps of the method described in any of the above embodiments.

[0096] Figure 6 A block diagram of an electronic device provided in an embodiment of this application.

[0097] This application provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method in any of the above embodiments.

[0098] like Figure 6 As shown, for ease of understanding, an embodiment of this application illustrates a specific electronic device 600.

[0099] Electronic device 600 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 600 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0100] like Figure 6As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of electronic device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0101] Multiple components in electronic device 600 are connected to I / O interface 605. These components include: input unit 606, such as a keyboard or mouse; output unit 607, such as various types of displays or speakers; storage unit 608, such as a disk or optical disk; and communication unit 609, such as a network interface card (NIC), modem, or wireless transceiver. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0102] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods described above. For example, in some embodiments, any one or more of the methods described above can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of any one or more of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform any one or more of the methods described above by any other suitable means (e.g., by means of firmware).

[0103] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this application, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0104] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0105] In the description of this application, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this application, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0106] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0107] Furthermore, the terms "first," "second," etc., used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this application can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this application, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly and specifically defined in the embodiments.

[0108] In this application, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing" appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication between two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific implementation.

[0109] In this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

Claims

1. A method for evaluating image noise, characterized in that, The method includes: Partial noise extraction is performed on the target medical image to obtain a target partially noisy image; The noise level of the target portion of the noise image is obtained by performing a noise assessment. A baseline noise level is obtained, which includes a low baseline noise level and a high baseline noise level; wherein, obtaining the baseline noise level includes: performing partial noise extraction on multiple baseline medical images containing different levels of noise to obtain multiple baseline partially noisy images; performing noise evaluation on the multiple baseline partially noisy images to obtain multiple candidate noise levels; and determining the low baseline noise level and the high baseline noise level based on the multiple candidate noise levels; The target noise level is evaluated based on the baseline noise level to obtain the noise evaluation result of the target medical image; The step of evaluating the target noise level based on the baseline noise level to obtain the noise evaluation result of the target medical image includes: If the target noise level is less than or equal to the low reference noise level, a first noise level value is determined as the noise assessment result. If the target noise level is greater than or equal to the high reference noise level, a second noise level value is determined as the noise assessment result, wherein the second noise level value is greater than the first noise level value; When the target noise level is greater than the low reference noise level and less than the high reference noise level, a third noise level value is determined as the noise assessment result. The third noise level value includes the ratio between a first noise difference and a second noise difference. The first noise difference is the difference between the target noise level and the low reference noise level, and the second noise difference is the difference between the high reference noise level and the low reference noise level. The third noise level value is greater than the first noise level value and less than the second noise level value.

2. The method according to claim 1, characterized in that, The step of evaluating the target noise level based on the baseline noise level to obtain the noise evaluation result of the target medical image further includes: Based on the texture intensity difference between the target medical image and the benchmark medical image, the benchmark noise level is updated to obtain the updated benchmark noise level. The target noise level is evaluated based on the updated baseline noise level to obtain the noise evaluation result of the target medical image.

3. The method according to claim 2, characterized in that, The step of updating the baseline noise level based on the texture intensity difference between the target medical image and the baseline medical image to obtain the updated baseline noise level includes: A target difference image is determined between the target medical image and the target partial noise image, and texture detection and evaluation are performed on the target difference image to obtain the target texture intensity; A baseline difference image is determined between the baseline medical image and the baseline partial noise image, and texture detection and evaluation are performed on the baseline difference image to obtain the baseline texture intensity, wherein the baseline partial noise image is obtained by partially extracting noise from the baseline medical image; Based on the difference between the target texture intensity and the reference texture intensity and the adjustment coefficient, a texture intensity adjustment level is obtained, and the reference noise level is added to the texture intensity adjustment level to obtain an updated reference noise level.

4. The method according to claim 3, characterized in that, The benchmark medical image includes a first benchmark medical image corresponding to the low benchmark noise level, and a second benchmark medical image corresponding to the high benchmark noise level; determining a benchmark difference image between the benchmark medical image and the benchmark partial noise image, and performing texture detection and evaluation on the benchmark difference image to obtain the benchmark texture intensity, includes: A first benchmark difference image is determined between the first benchmark medical image and the first benchmark partial noise image, and texture detection and evaluation are performed on the first benchmark difference image to obtain the first benchmark texture intensity, wherein the first benchmark partial noise image is obtained by partially extracting noise from the first benchmark medical image; A second benchmark difference image is determined between the second benchmark medical image and the second benchmark partial noise image, and texture detection and evaluation are performed on the second benchmark difference image to obtain the second benchmark texture intensity, wherein the second benchmark partial noise image is obtained by partially extracting noise from the second benchmark medical image.

5. The method according to claim 4, characterized in that, The process of obtaining a texture intensity adjustment level based on the difference between the target texture intensity and the reference texture intensity, and an adjustment coefficient, and then adding the reference noise level to the texture intensity adjustment level to obtain an updated reference noise level, includes: Based on the difference between the target texture intensity and the first reference texture intensity and the first adjustment coefficient, a first texture intensity adjustment level is obtained, and the first texture intensity adjustment level is added to the low reference noise level to obtain an updated low reference noise level. Based on the difference between the target texture intensity and the second reference texture intensity and the second adjustment coefficient, a second texture intensity adjustment level is obtained, and the second texture intensity adjustment level is added to the high reference noise level to obtain an updated high reference noise level.

6. The method according to claim 3, characterized in that, The step of performing texture detection and evaluation on the target difference image to obtain the target texture intensity includes: performing edge detection on the target difference image to obtain a target edge detection image, and performing partial noise extraction and evaluation on the target edge detection image to obtain the target texture intensity; The step of performing texture detection and evaluation on the benchmark difference image to obtain benchmark texture intensity includes: performing edge detection on the benchmark difference image to obtain benchmark edge detection image, and performing partial noise extraction and evaluation on the benchmark edge detection image to obtain the benchmark texture intensity.

7. An image noise assessment device, characterized in that, The device includes: The extraction module is used to extract some noise from the target medical image to obtain a target part noise image; The first evaluation module is used to evaluate the noise level of the target portion of the noise image to obtain the target noise level. An acquisition module is used to acquire a reference noise level, which includes a low reference noise level and a high reference noise level. Specifically, the acquisition module is used to: extract partial noise from multiple reference medical images containing different levels of noise to obtain multiple reference partially noisy images; evaluate the noise of the multiple reference partially noisy images to obtain multiple candidate noise levels; and determine the low reference noise level and the high reference noise level based on the multiple candidate noise levels. The second evaluation module is configured to: determine a first noise level value as a noise evaluation result when the target noise level is less than or equal to the low reference noise level; determine a second noise level value as the noise evaluation result when the target noise level is greater than or equal to the high reference noise level, wherein the second noise level value is greater than the first noise level value; and determine a third noise level value as the noise evaluation result when the target noise level is greater than the low reference noise level and less than the high reference noise level, wherein the third noise level value includes the ratio between a first noise difference and a second noise difference, the first noise difference being the difference between the target noise level and the low reference noise level, the second noise difference being the difference between the high reference noise level and the low reference noise level, and the third noise level value being greater than the first noise level value and less than the second noise level value.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images