A rapid medical image quality improvement system based on artificial intelligence

Abstract

This invention provides an artificial intelligence-based system for rapid improvement of medical image quality, relating to the field of medical image imaging technology. The system includes: controlling an ultrasound acquisition probe to acquire ultrasound signals from a current patient and reconstructing the image to obtain an ultrasound image; acquiring a standard ultrasound image as a reference image; using the ultrasound image as the image to be improved; inputting the reference image into a pre-trained feature extraction model to obtain multiple feature indicators; subsequently selecting multiple reference feature indicators from these indicators to perform a reference quality assessment of the image to be improved, resulting in an image quality difference map; extracting multiple regions to be improved from the image to be improved based on the image quality difference map; using an artificial intelligence analysis model to analyze the improvement indicators for each region; and then, for each region, performing image quality improvement based on the improvement indicators; finally, obtaining the improved image after all image quality improvements are completed. The beneficial effect is that it can achieve rapid improvement of ultrasound image imaging quality.

Description

Technical Field

[0001] This invention relates to the field of medical image imaging technology, and in particular to a medical image quality rapid improvement system based on artificial intelligence. Background Technology

[0002] Ultrasound imaging is a technique that uses the principle of reflection and scattering of ultrasound waves within the human body to obtain tomographic images of human tissues and organs. Ultrasound imaging has advantages such as being non-invasive, real-time, safe, and convenient, and is widely used in medical diagnosis and treatment. However, due to various factors affecting the propagation of ultrasound waves within the body, such as attenuation, interference, and speckle, the quality of ultrasound images is often low, making it difficult to clearly display details and features of the target area. The frequency, intensity, and waveform of ultrasound waves all affect the quality of ultrasound images. Generally, higher frequencies result in higher resolution but lower penetration; lower frequencies result in higher penetration but lower resolution. Therefore, selecting an appropriate frequency is crucial for improving ultrasound image quality. Furthermore, the intensity and waveform of ultrasound waves also affect the brightness and contrast of the image. The technical specifications and functions of the ultrasound diagnostic instrument also influence the quality of the ultrasound images. For example, a transducer is a component that converts electrical energy into ultrasonic energy and echo signals into electrical signals. Its performance directly determines the resolution and sensitivity of the ultrasound image. Signal processing involves filtering, enhancing, and restoring the echo signal to improve the contrast and detail of the image. Image display converts the processed signal into a visualized image to improve the brightness and clarity of the image.

[0003] In existing technologies, image quality improvement often involves adjusting the entire image uniformly. This not only causes areas that were originally of acceptable quality to be over-adjusted, leading to abnormal quality, but also takes a long time to adjust the overall image quality. Summary of the Invention

[0004] To address the storage problems in existing technologies, this invention provides an artificial intelligence-based system for rapidly improving the quality of medical images, comprising:

[0005] The image acquisition module is used to control the ultrasound acquisition probe to acquire ultrasound signals from the current patient according to the preset dynamic range and signal gain, and to reconstruct the ultrasound image to obtain the ultrasound image, as well as to acquire a standard ultrasound image as a reference image.

[0006] The quality assessment module, connected to the image acquisition module, is used to take the ultrasound image as the image to be improved, input the reference image into a pre-trained feature extraction model to obtain multiple feature indicators, and then select multiple reference feature indicators from each feature indicator to perform reference quality assessment on the image to be improved to obtain an image quality difference map.

[0007] The quality improvement module, connected to the quality assessment module, is used to extract multiple regions to be improved from the image to be improved based on the image quality difference map, analyze the improvement indicators of each region to be improved using an artificial intelligence analysis model, and then perform image quality improvement on each region to be improved based on each improvement indicator. After all image quality improvements are completed, the improved image is obtained.

[0008] Preferably, the quality assessment module includes:

[0009] The first storage unit is used to store the weights corresponding to each of the aforementioned feature indicators, as well as the weight thresholds corresponding to multiple evaluation levels.

[0010] A feature extraction unit is used to acquire a reference image, and then input the reference image into the feature extraction model to obtain multiple feature indicators;

[0011] The quality assessment unit, connected to the feature extraction unit and the first storage unit, is used to filter out each feature index whose weight exceeds the corresponding weight threshold as a reference feature index according to the assessment level of the external input, and then perform a reference quality assessment on the image to be improved according to each reference feature index to generate the corresponding image quality difference map.

[0012] Preferably, the quality assessment unit includes:

[0013] The full reference evaluation subunit is used to perform a full reference quality evaluation on the image to be improved based on each of the reference indicators when the number of reference feature indicators is equal to the number of feature indicators, to obtain the image quality difference map.

[0014] The downreference evaluation subunit is used to perform downreference quality evaluation on the image to be improved based on each of the reference indicators when the number of reference feature indicators is less than the number of feature indicators, to obtain the image quality difference map.

[0015] Preferably, the image quality difference map includes multiple difference regions, and the quality improvement module includes:

[0016] The second storage unit is used to store standard noise, standard sharpness, and standard distortion.

[0017] An analysis unit, connected to the second storage unit, is used to extract image blocks from corresponding positions in the image to be enhanced based on each difference region as the region to be enhanced. It then uses an artificial intelligence analysis model to analyze the noise, sharpness, and distortion of each region to be enhanced. Subsequently, for each region to be enhanced, it calculates the noise difference value between noise and standard noise, the sharpness difference value between sharpness and standard sharpness, and the distortion difference value between distortion and standard distortion. When the noise difference value is greater than a preset difference value threshold, noise is associated with the region to be enhanced as the enhancement indicator. When the sharpness difference value is greater than the difference value threshold, sharpness is associated with the region to be enhanced as the enhancement indicator. When the distortion difference value is greater than the difference value threshold, distortion is associated with the region to be enhanced as the enhancement indicator.

[0018] The quality improvement unit, connected to the analysis unit, is used to improve the image quality of each region to be improved based on each indicator to be improved, and to obtain the improved image after all image quality improvements are completed.

[0019] Preferably, the quality improvement unit includes:

[0020] A filtering subunit is used to perform image filtering processing on the region to be boosted when the region to be boosted is associated with a noisy index to be boosted.

[0021] An enhancement subunit is used to perform image enhancement processing on the region to be enhanced when the region to be enhanced is associated with the enhancement index of sharpness;

[0022] Complex atomic units are used to perform image restoration processing on the region to be enhanced when the region to be enhanced is associated with the enhancement index of distortion.

[0023] Preferably, the quality assessment module further includes an acquisition and adjustment unit, used to adjust the dynamic range and the signal gain according to the indicator to be improved, and then perform the next ultrasound signal acquisition.

[0024] This invention also provides a method for rapidly improving the quality of medical images based on artificial intelligence, applied to the aforementioned system for rapidly improving the quality of medical images. The method for rapidly improving the quality of medical images includes:

[0025] Step S1: The medical image quality rapid improvement system acquires a reference image and controls the ultrasound acquisition probe to acquire ultrasound signals from the patient and generate an ultrasound image as the image to be improved. The reference image is input into a pre-trained feature extraction model to obtain multiple feature indicators. Then, multiple reference feature indicators are selected from each feature indicator to perform reference quality assessment on the image to be improved to obtain an image quality difference map.

[0026] Step S2: The medical image quality rapid improvement system extracts multiple regions to be improved from the image to be improved based on the image quality difference map, analyzes the improvement indicators of each region using an artificial intelligence analysis model, and then improves the image quality of each region based on the improvement indicators. After all the image quality improvements are completed, the improved image is obtained.

[0027] Preferably, in the medical image quality rapid improvement system, the weights corresponding to each feature indicator and the weight thresholds corresponding to multiple evaluation levels are stored. Then, step S1 includes:

[0028] Step S11: The medical image quality rapid enhancement system controls the ultrasound acquisition probe to acquire ultrasound signals according to a preset dynamic range and signal gain, and performs image reconstruction to obtain the ultrasound image as the image to be enhanced.

[0029] Step S12: The medical image quality rapid improvement system acquires a reference image, and then inputs the reference image into the feature extraction model to obtain multiple feature indicators;

[0030] In step S13, the medical image quality rapid improvement system selects each feature index whose weight exceeds the corresponding weight threshold as reference feature index based on the externally input evaluation level. Then, it performs a reference quality evaluation on the image to be improved based on each reference feature index to generate the corresponding image quality difference map.

[0031] Preferably, step S13 includes:

[0032] Step S131: When the number of reference feature indicators is equal to the number of feature indicators, the medical image quality improvement system performs a full reference quality assessment on the image to be improved based on each of the reference indicators to obtain the image quality difference map.

[0033] In step S132, when the number of reference feature indicators is less than the number of feature indicators, the medical image quality improvement system performs a reference quality assessment on the image to be improved based on each reference indicator to obtain the image quality difference map.

[0034] Preferably, the image quality difference map includes multiple difference regions, and the medical image quality rapid improvement system saves standard noise, standard sharpness, and standard distortion. Then, step S2 includes:

[0035] Step S21: The medical image quality rapid improvement system extracts image blocks from corresponding positions in the image to be improved as the areas to be improved based on the differences between each area. It then uses an artificial intelligence analysis model to analyze the noise, sharpness, and distortion of each area to be improved. Subsequently, for each area to be improved, it calculates the noise difference between noise and standard noise, the sharpness difference between sharpness and standard sharpness, and the distortion difference between distortion and standard distortion. When the noise difference value is greater than a preset difference value threshold, noise is used as the indicator to be improved and associated with the area to be improved. When the sharpness difference value is greater than the difference value threshold, sharpness is used as the indicator to be improved and associated with the area to be improved. When the distortion difference value is greater than the difference value threshold, distortion is used as the indicator to be improved and associated with the area to be improved.

[0036] Step S22: The medical image quality rapid improvement system improves the image quality of each area to be improved based on each indicator to be improved, and obtains the improved image after all image quality improvements are completed.

[0037] The above technical solution has the following advantages or beneficial effects:

[0038] 1) Perform image quality assessment on the image to be improved, extract the difference area between the image to be improved and the reference image, and conduct targeted analysis on the indicators that need to be improved for each difference area. Then make targeted adjustments without affecting other image areas that meet the quality standards. Moreover, only the parts of the image that do not meet the quality standards need to be adjusted, which can save a lot of time.

[0039] 2) When analyzing image quality, physicians can dynamically adjust the quality assessment criteria according to the image quality requirements, saving quality assessment time. Attached Figure Description

[0040] Figure 1 A schematic diagram of a medical image quality rapid improvement system based on artificial intelligence is shown in a preferred embodiment of the present invention.

[0041] Figure 2 A flowchart illustrating a method for rapidly improving the quality of medical images based on artificial intelligence, as a preferred embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of a sub-process of step S1 in a preferred embodiment of the present invention.

[0043] Figure 4 This is a schematic diagram of a sub-process of step S13 in a preferred embodiment of the present invention.

[0044] Figure 5This is a schematic diagram of the sub-process of step S2 in a preferred embodiment of the present invention. Detailed Implementation

[0045] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0046] In a preferred embodiment of the present invention, based on the aforementioned problems stored in the prior art, an artificial intelligence-based medical image quality rapid improvement system is provided, comprising:

[0047] Image acquisition module 1 is used to control the ultrasound acquisition probe to acquire ultrasound signals from the current patient according to the preset dynamic range and signal gain, and to reconstruct the ultrasound image to obtain an ultrasound image, as well as to acquire a standard ultrasound image as a reference image.

[0048] Quality assessment module 2, connected to image acquisition module 1, is used to take ultrasound image as the image to be improved, input reference image into pre-trained feature extraction model to obtain multiple feature indicators, and then select multiple reference feature indicators from each feature indicator to perform reference quality assessment on the image to be improved to obtain image quality difference map.

[0049] The quality improvement module 3, connected to the quality assessment module 2, is used to extract multiple areas to be improved from the image to be improved based on the image quality difference map, analyze the improvement indicators of each area using an artificial intelligence analysis model, and then improve the image quality of each area based on each improvement indicator. After all the image quality improvements are completed, the improved image is obtained.

[0050] Specifically, in this embodiment, when acquiring images of a patient, scanning is performed according to a preset dynamic range and signal gain. The preset values ​​are selected based on previous scanning and imaging experience.

[0051] Dynamic range refers to the ratio of the maximum signal intensity to the minimum signal intensity that an ultrasonic system can handle, usually expressed in decibels (dB). A larger dynamic range means the system can distinguish more echo signals of varying intensities, thus displaying richer details and contrast. Adjusting the ultrasonic acquisition probe according to the preset dynamic range can avoid image distortion or blurring caused by excessively large or small dynamic ranges.

[0052] Signal gain refers to the amplification factor of an ultrasonic system on an input signal, also expressed in decibels (dB). A higher signal gain indicates greater sensitivity of the system to the input signal, resulting in a brighter image. Adjusting the ultrasonic acquisition probe according to the preset signal gain can avoid image noise or dimness caused by excessively high or low signal gain.

[0053] However, for different individual patients, the acquired images will vary, resulting in some areas of the image not meeting our expected imaging effect and quality. Therefore, image quality assessment is necessary. For ultrasound image quality assessment, we select an ultrasound image with better imaging effect for the current scanned area as a reference image. In other words, the image quality of the reference image is the image quality we want to achieve. After selecting the reference image, we need to extract feature indicators from it for quality assessment of the image to be improved. The feature extraction model here is not specifically limited; existing feature extraction models can be used. There are many feature extraction models, such as wavelet transform-based feature extraction models. This model utilizes the multi-scale analysis and direction selectivity of wavelet transform to extract features from ultrasound images. Features of different frequencies and directions, such as energy, entropy, and variance, can be extracted. A convolutional neural network-based feature extraction model utilizes the deep learning capabilities of convolutional neural networks to automatically learn and extract high-level abstract features from ultrasound images, such as shape, texture, and semantics. A fractional Brownian motion-based feature extraction model uses the fractal theory of fractional Brownian motion to extract fractal features reflecting image complexity and roughness from ultrasound images, such as fractal dimension, fractal length, and fractal spectrum. As long as the desired feature indicators can be extracted from ultrasound images, conventional training methods can be used. To achieve better feature extraction results, the training data and the final feature requirements need to be adjusted according to actual needs, which will not be elaborated here.

[0054] Ultrasound image quality assessment is a method for quantifying or classifying the sharpness, uniformity, structure, blood flow, and other characteristics of ultrasound images, aiming to improve the accuracy and efficiency of ultrasound diagnosis. Ultrasound image quality assessment can be divided into subjective and objective assessments. Objective assessment refers to using mathematical models or algorithms to evaluate ultrasound images, typically employing methods such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and No Reference Quality Index (NRQI). The advantages of objective assessment are its speed, convenience, and ease of standardization and automation.

[0055] In a preferred embodiment of the present invention, the quality assessment module 2 includes:

[0056] The first storage unit 21 is used to store the weights corresponding to each feature index and the weight thresholds corresponding to multiple evaluation levels.

[0057] The feature extraction unit 22 is used to acquire a reference image, and then input the reference image into the feature extraction model to obtain multiple feature indicators;

[0058] The quality assessment unit 23, which connects the feature extraction unit 22 and the first storage unit 21, is used to select each feature index whose weight exceeds the corresponding weight threshold as a reference feature index according to the assessment level of the external input. Then, the reference quality assessment is performed on the image to be improved according to each reference feature index to generate the corresponding image quality difference map.

[0059] Specifically, in this embodiment, feature indicators whose weights exceed the corresponding weight thresholds are selected as reference feature indicators based on the evaluation level of the external input. The evaluation level of the external input can be categorized as lenient, moderate, or strict. A lenient level indicates low image quality requirements, requiring only quality improvement of the more important parts of the ultrasound image; a moderate level indicates moderate image quality requirements, requiring quality improvement of the more important and some less important parts of the ultrasound image; and a strict level indicates very high image quality requirements, requiring quality improvement of the vast majority of the ultrasound image. Since each feature indicator has already been assigned weights according to its importance, the corresponding weight threshold decreases when the evaluation level increases and increases when the evaluation level decreases (higher requirements mean more feature indicators need to be improved; lowering the threshold increases the number of feature indicators exceeding the threshold). Feature indicators whose weights exceed the corresponding weight thresholds are then selected as reference feature indicators.

[0060] In a preferred embodiment of the present invention, the quality assessment unit 23 includes:

[0061] The full reference evaluation subunit 231 is used to perform a full reference quality evaluation on the image to be enhanced based on each reference indicator when the number of reference feature indicators is equal to the number of feature indicators, and obtain an image quality difference map.

[0062] The downreference evaluation subunit 232 is used to perform downreference quality evaluation on the image to be enhanced based on each reference indicator when the number of reference feature indicators is less than the number of feature indicators, and obtain an image quality difference map.

[0063] Specifically, in this embodiment, the aforementioned objective evaluation can be further categorized based on the proportion of source image reference information: Full Reference Image Quality Assessment (FR-IQA), Reduced Reference Image Quality Assessment (RR-IQA), and No Reference Image Quality Assessment (NR-IQA). This embodiment employs either full reference or reduced reference methods.

[0064] Full-reference evaluation can only be performed when the original, distortion-free images are available, making it relatively easy. Its core idea is to compare the information content or feature similarity between two images. Due to the abundance of information, research on this topic is relatively thorough, and various evaluation metrics are well-established.

[0065] Semi-reference evaluation uses only partial information from the original image or partial features extracted from a reference image; such methods fall between FR-IQA and NR-IQA. Semi-reference evaluation aims to extract the features that best reflect image quality while minimizing the amount of information required.

[0066] Generally, the main features of ultrasound images include:

[0067] Brightness: Brightness refers to the grayscale value of each pixel in an image, reflecting the intensity of the ultrasound echo signal. Brightness can be used to measure the contrast and clarity of tissues or structures in ultrasound images.

[0068] Contrast ratio: Contrast ratio refers to the degree of brightness difference between different areas or objects in an image, reflecting the attenuation and reflection of ultrasound echo signals in different media. Contrast ratio can be used to measure the discernibility and detail of tissues or structures in ultrasound images.

[0069] Structure: Structure refers to the spatial relationships and morphological characteristics between different regions or objects in an image, reflecting the propagation and scattering of ultrasound echo signals in different media. Structure can be used to measure the shape and location of tissues or structures in ultrasound images.

[0070] Blood flow: Blood flow refers to the dynamic blood flow shown in an ultrasound image, reflecting the frequency shift signal generated by the Doppler effect. Blood flow can be used to measure the function and condition of blood vessels or the heart in ultrasound images.

[0071] In this embodiment, the physician's medical experience is used to determine the degree of influence of each feature index of the image on the image quality of the affected area, and this standard is used to assign corresponding weights to each feature of the image.

[0072] When assessing the quality of images to be upgraded, an assessment level is selected. The assessment level indicates the rigor of the quality assessment. For example, if a physician has high requirements for image quality, he will set the assessment level to be high and the weight threshold to be low. This increases the number of feature indicators that need to be judged, ensuring that as many differences as possible between the image to be upgraded and the reference image can be found. When the weight threshold is 0, it corresponds to full reference assessment.

[0073] Both full-reference evaluation and dereference evaluation generate image quality difference maps, which can quickly reflect the differences between the image to be upgraded and the reference image in graphical form, and extract these differences as difference regions.

[0074] In a preferred embodiment of the present invention, the image quality difference map includes multiple difference regions, and the quality improvement module 3 includes:

[0075] The second storage unit 31 is used to store standard noise, standard sharpness, and standard distortion.

[0076] Analysis unit 32, connected to second storage unit 31, is used to extract image blocks from corresponding positions in the image to be enhanced based on each difference region as the region to be enhanced. It uses an artificial intelligence analysis model to analyze the noise, sharpness, and distortion of each region to be enhanced. Then, for each region to be enhanced, it calculates the noise difference value between noise and standard noise, the sharpness difference value between sharpness and standard sharpness, and the distortion difference value between distortion and standard distortion. When the noise difference value is greater than a preset difference value threshold, noise is used as an enhancement indicator and associated with the region to be enhanced. When the sharpness difference value is greater than the difference value threshold, sharpness is used as an enhancement indicator and associated with the region to be enhanced. When the distortion difference value is greater than the difference value threshold, distortion is used as an enhancement indicator and associated with the region to be enhanced.

[0077] The quality improvement unit 33 is connected to the analysis unit and is used to improve the image quality of each area to be improved based on each indicator to be improved. After all the image quality improvements are completed, the improved image is obtained.

[0078] Specifically, in this embodiment, image quality improvement is carried out by targeted improvement of the indicators to be improved in each difference region of the image;

[0079] Therefore, before proceeding, it is necessary to analyze the differences between each difference area and the corresponding position of the reference image through an artificial intelligence model. In this embodiment, noise, sharpness, and distortion are mainly used as the main analysis indicators. Specifically, in practical applications, the analysis indicators of the artificial intelligence model can be adjusted according to actual needs. The artificial intelligence model analyzes which aspects of each difference area are not up to standard and associates the indicator with the difference area as the indicator to be improved, and then makes targeted adjustments.

[0080] The following are some possible ways to use artificial intelligence analysis models to analyze the noise, sharpness, and distortion of each area to be improved:

[0081] Image quality assessment-based approach: This method utilizes Image Quality Assessment (IQA) to extract features reflecting image quality from ultrasound images, such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Laplacian Gradient Sum (LGS). Then, it calculates numerical values ​​for noise, sharpness, and distortion based on these features. The advantage of this approach is its relative simplicity and speed, leveraging existing IQA algorithms and datasets. The disadvantage is that it may not fully consider the unique characteristics and diversity of ultrasound images, as well as the perceptual features of the Human Visual System (HVS).

[0082] The deep learning-based approach: This method utilizes deep learning (DL) technology to automatically learn and extract high-level abstract features from ultrasound images, such as shape, texture, and semantics. Then, an artificial intelligence analysis model is built based on these features to predict values ​​for metrics such as noise, sharpness, and distortion. The advantage of this approach is its adaptability to different types and scenarios of ultrasound images, improving the accuracy and robustness of the analysis. The disadvantages are the need for large amounts of labeled data and computational resources, as well as complex model design and optimization.

[0083] The adversarial network-based approach: This method utilizes Generative Adversarial Networks (GANs) to generate optimized 2D grayscale images from ultrasound images, thus creating high-quality ultrasound images. Then, it calculates metrics such as noise, sharpness, and distortion based on the differences between the generated and original images. The advantage of this approach is that it can generate high-quality and realistic ultrasound images, improving the credibility of the analysis and the visualization effect. The disadvantage is the need to design appropriate loss functions and evaluation metrics, as well as balance the adversarial relationship between the generator and the discriminator.

[0084] In addition to the methods mentioned above, there are other relatively mature artificial intelligence analysis models in the existing technology for analyzing the noise, sharpness, and distortion of ultrasound images. Any model that can analyze the noise, sharpness, and distortion of each area to be improved can be selected, without any specific limitations.

[0085] In a preferred embodiment of the present invention, the quality improvement unit 33 includes:

[0086] The filtering subunit 331 is used to perform image filtering processing on the region to be lifted when the region to be lifted is associated with noisy indicators to be lifted.

[0087] Enhancement subunit 332 is used to perform image enhancement processing on the region to be enhanced when the region to be enhanced is associated with a clear indicator to be enhanced;

[0088] Complex atomic unit 333 is used to perform image restoration processing on the region to be lifted when the region to be lifted is associated with a distortion index to be lifted.

[0089] Specifically, in this embodiment, the methods for improving the quality of ultrasound images based on the aforementioned indicators can be mainly carried out from three aspects: filtering, enhancement, and restoration.

[0090] Filtering: Filtering refers to performing mathematical transformations or operations on ultrasound images to eliminate or reduce noise and artifacts, and improve the signal-to-noise ratio and resolution. There are many filtering methods, such as linear filtering, nonlinear filtering, adaptive filtering, and frequency domain filtering. Linear filtering involves weighted averaging or convolution operations on the gray values ​​of each pixel in the image, such as mean filtering, Gaussian filtering, and Wiener filtering. Nonlinear filtering involves nonlinear transformations or operations on the gray values ​​of each pixel in the image, such as median filtering, bilateral filtering, and anisotropic diffusion. Adaptive filtering adjusts the filter parameters or coefficients based on local features or statistical information of the image, such as adaptive median filtering and adaptive Wiener filtering. Frequency domain filtering involves converting the image to the frequency domain through Fourier transform or other transformations, and then designing appropriate filters based on frequency characteristics, such as low-pass filters, high-pass filters, and band-pass filters.

[0091] Enhancement: Enhancement refers to processing ultrasound images to emphasize or highlight certain features or information of interest, improving image contrast and recognizability. There are many enhancement methods, such as histogram equalization, homomorphic filtering, retinal enhancement, and pyramid fusion. Histogram equalization adjusts the grayscale distribution of an image to make it more uniform and balanced, thereby increasing the image's dynamic range and brightness. Homomorphic filtering separates the reflection and illumination components of an image and processes them separately, simultaneously increasing both brightness and detail. Retinal enhancement uses the processing mechanisms of the human retina to adaptively enhance and restore the image's color. Pyramid fusion decomposes images of different scales or modalities into multi-layered pyramid structures and recombines these layers into a new image according to certain rules.

[0092] Restoration: Restoration refers to performing inverse operations or inferences on ultrasound images to restore or reconstruct the original or ideal state of the image, improving its realism and accuracy. Restoration methods are mainly divided into two categories: model-based methods and learning-based methods. Model-based methods establish the relationship between image degradation and restoration based on the physical process or mathematical model of ultrasound image imaging, and then solve for the restored image through optimization algorithms, such as deconvolution restoration, linear algebraic restoration, and blind deconvolution. Learning-based methods utilize large amounts of labeled or unlabeled image data, employing machine learning or deep learning techniques to learn the mapping function or network model between image degradation and restoration, and then obtain the restored image through inference or prediction, such as dictionary learning, sparse coding, and convolutional neural networks.

[0093] In a preferred embodiment of the present invention, the quality assessment module 2 further includes an acquisition and adjustment unit 24, which is used to adjust the dynamic range and signal gain according to the index to be improved, and then perform the next ultrasound signal acquisition.

[0094] This invention also provides an artificial intelligence-based method for rapidly improving the quality of medical images, applied to the aforementioned system for rapidly improving the quality of medical images, such as... Figure 2 As shown, methods for rapidly improving the quality of medical images include:

[0095] Step S1: The medical image quality rapid improvement system acquires a reference image and controls the ultrasound acquisition probe to acquire ultrasound signals from the patient and generate an ultrasound image as the image to be improved. The reference image is input into a pre-trained feature extraction model to obtain multiple feature indicators. Then, multiple reference feature indicators are selected from each feature indicator to perform reference quality assessment on the image to be improved and obtain an image quality difference map.

[0096] Step S2: The medical image quality rapid improvement system extracts multiple areas to be improved from the image to be improved based on the image quality difference map, uses an artificial intelligence analysis model to analyze the improvement indicators of each area, and then improves the image quality of each area based on the improvement indicators. After all the image quality improvements are completed, the improved image is obtained.

[0097] In a preferred embodiment of the present invention, the medical image quality rapid improvement system stores the weights corresponding to each feature index and the weight thresholds corresponding to multiple evaluation levels, as follows: Figure 3 As shown, step S1 includes:

[0098] Step S11: The medical image quality rapid enhancement system controls the ultrasound acquisition probe to acquire ultrasound signals according to the preset dynamic range and signal gain, and performs image reconstruction to obtain an ultrasound image as the image to be enhanced.

[0099] Step S12: The medical image quality rapid improvement system acquires a reference image, and then inputs the reference image into the feature extraction model to obtain multiple feature indicators;

[0100] In step S13, the medical image quality rapid improvement system selects each feature index whose weight exceeds the corresponding weight threshold as reference feature index based on the external input evaluation level. Then, it performs a reference quality evaluation on the image to be improved based on each reference feature index to generate the corresponding image quality difference map.

[0101] In a preferred embodiment of the present invention, such as Figure 4 As shown, step S13 includes:

[0102] Step S131: When the number of reference feature indicators is equal to the number of feature indicators, the medical image quality rapid improvement system performs a full reference quality assessment on the image to be improved based on each reference indicator to obtain an image quality difference map.

[0103] Step S132: When the number of reference feature indicators is less than the number of feature indicators, the medical image quality rapid improvement system performs a reference quality assessment on the image to be improved based on each reference indicator to obtain an image quality difference map.

[0104] In a preferred embodiment of the present invention, the image quality difference map includes multiple difference regions, and the medical image quality rapid improvement system saves standard noise, standard sharpness, and standard distortion, then... Figure 5 As shown, step S2 includes:

[0105] Step S21: The medical image quality rapid improvement system extracts image blocks from the corresponding positions in the image to be improved as areas to be improved based on each difference area. It uses an artificial intelligence analysis model to analyze the noise, sharpness, and distortion of each area to be improved. Then, for each area to be improved, it calculates the noise difference value between noise and standard noise, the sharpness difference value between sharpness and standard sharpness, and the distortion difference value between distortion and standard distortion. When the noise difference value is greater than the preset difference value threshold, the noise is used as an indicator to be improved and associated with the area to be improved. When the sharpness difference value is greater than the difference value threshold, the sharpness is used as an indicator to be improved and associated with the area to be improved. When the distortion difference value is greater than the difference value threshold, the distortion is used as an indicator to be improved and associated with the area to be improved.

[0106] Step S22: The medical image quality rapid improvement system improves the image quality of each area to be improved based on each indicator to be improved, and obtains the improved image after all image quality improvements are completed.

[0107] The above are merely 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 recognize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. A medical image quality rapid improvement system based on artificial intelligence, characterized in that, include: The image acquisition module is used to control the ultrasound acquisition probe to acquire ultrasound signals from the current patient according to the preset dynamic range and signal gain, and to reconstruct the ultrasound image to obtain the ultrasound image, as well as to acquire a standard ultrasound image as a reference image. The quality assessment module, connected to the image acquisition module, is used to take the ultrasound image as the image to be improved, input the reference image into a pre-trained feature extraction model to obtain multiple feature indicators, and then select multiple reference feature indicators from each of the feature indicators to perform reference quality assessment on the image to be improved to obtain an image quality difference map; the feature indicators include at least one of brightness, contrast, structure and blood flow. The quality improvement module, connected to the quality assessment module, is used to extract multiple regions to be improved from the image to be improved based on the image quality difference map, analyze the improvement indicators of each region to be improved using an artificial intelligence analysis model, and then perform image quality improvement on each region to be improved based on each improvement indicator. After all image quality improvements are completed, the improved image is obtained. The quality assessment module includes: The first storage unit is used to store the weights corresponding to each of the feature indicators, and the weight thresholds corresponding to multiple evaluation levels, wherein the weights are pre-allocated based on the degree of influence of each feature indicator on image quality. A feature extraction unit is used to acquire a reference image, and then input the reference image into the feature extraction model to obtain multiple feature indicators; The quality assessment unit, connected to the feature extraction unit and the first storage unit, is used to filter out each feature index whose weight exceeds the corresponding weight threshold as a reference feature index according to the assessment level of the external input, and then perform a reference quality assessment on the image to be improved according to each reference feature index to generate the corresponding image quality difference map.

2. The medical image quality rapid improvement system according to claim 1, characterized in that, The quality assessment unit includes: The full reference evaluation subunit is used to perform a full reference quality evaluation on the image to be improved based on each of the reference feature indicators to obtain the image quality difference map when the number of reference feature indicators is equal to the number of feature indicators. The downreference evaluation subunit is used to perform downreference quality evaluation on the image to be improved based on each of the reference feature indicators to obtain the image quality difference map when the number of reference feature indicators is less than the number of feature indicators.

3. The medical image quality rapid improvement system according to claim 1, characterized in that, The image quality difference map includes multiple difference regions, and the quality improvement module includes: The second storage unit is used to store standard noise, standard sharpness, and standard distortion. An analysis unit, connected to the second storage unit, is used to extract image blocks from corresponding positions in the image to be enhanced based on each difference region as the region to be enhanced. It then uses an artificial intelligence analysis model to analyze the noise, sharpness, and distortion of each region to be enhanced. Subsequently, for each region to be enhanced, it calculates the noise difference value between noise and standard noise, the sharpness difference value between sharpness and standard sharpness, and the distortion difference value between distortion and standard distortion. When the noise difference value is greater than a preset difference value threshold, noise is associated with the region to be enhanced as the enhancement indicator. When the sharpness difference value is greater than the difference value threshold, sharpness is associated with the region to be enhanced as the enhancement indicator. When the distortion difference value is greater than the difference value threshold, distortion is associated with the region to be enhanced as the enhancement indicator. The quality improvement unit, connected to the analysis unit, is used to improve the image quality of each region to be improved based on each indicator to be improved, and to obtain the improved image after all image quality improvements are completed.

4. The medical image quality rapid improvement system according to claim 3, characterized in that, The quality improvement unit includes: A filtering subunit is used to perform image filtering processing on the region to be boosted when the region to be boosted is associated with a noisy index to be boosted. An enhancement subunit is used to perform image enhancement processing on the region to be enhanced when the region to be enhanced is associated with the enhancement index of sharpness; Complex atomic units are used to perform image restoration processing on the region to be enhanced when the region to be enhanced is associated with the enhancement index of distortion.

5. The medical image quality rapid improvement system according to claim 1, characterized in that, The quality assessment module also includes an acquisition and adjustment unit, which is used to adjust the dynamic range and the signal gain according to the index to be improved, and then perform the next ultrasound signal acquisition.

6. A method for rapidly improving the quality of medical images based on artificial intelligence, characterized in that, The medical image quality rapid enhancement system as described in any one of claims 1-5, wherein the medical image quality rapid enhancement method comprises: Step S1: The medical image quality rapid improvement system acquires a reference image and controls the ultrasound acquisition probe to acquire ultrasound signals from the patient and generate an ultrasound image as the image to be improved. The reference image is input into a pre-trained feature extraction model to obtain multiple feature indicators. Then, multiple reference feature indicators are selected from each feature indicator to perform reference quality assessment on the image to be improved to obtain an image quality difference map. Step S2: The medical image quality rapid improvement system extracts multiple regions to be improved from the image to be improved based on the image quality difference map, analyzes the improvement indicators of each region using an artificial intelligence analysis model, and then improves the image quality of each region based on the improvement indicators. After all the image quality improvements are completed, the improved image is obtained.

7. The method for rapidly improving the quality of medical images according to claim 6, characterized in that, The medical image quality rapid improvement system stores the weights corresponding to each feature indicator and the weight thresholds corresponding to multiple evaluation levels. Step S1 then includes: Step S11: The medical image quality rapid enhancement system controls the ultrasound acquisition probe to acquire ultrasound signals according to a preset dynamic range and signal gain, and performs image reconstruction to obtain the ultrasound image as the image to be enhanced. Step S12: The medical image quality rapid improvement system acquires a reference image, and then inputs the reference image into the feature extraction model to obtain multiple feature indicators; In step S13, the medical image quality rapid improvement system selects each feature index whose weight exceeds the corresponding weight threshold as reference feature index based on the externally input evaluation level. Then, it performs a reference quality evaluation on the image to be improved based on each reference feature index to generate the corresponding image quality difference map.

8. The method for rapidly improving the quality of medical images according to claim 7, characterized in that, Step S13 includes: Step S131: When the number of reference feature indicators is equal to the number of feature indicators, the medical image quality rapid improvement system performs a full reference quality assessment on the image to be improved based on each of the reference feature indicators to obtain the image quality difference map. Step S132: When the number of reference feature indicators is less than the number of feature indicators, the medical image quality rapid improvement system performs a reference quality assessment on the image to be improved based on each of the reference feature indicators to obtain the image quality difference map.

9. The method for rapidly improving the quality of medical images according to claim 6, characterized in that, The image quality difference map includes multiple difference regions. The medical image quality rapid improvement system saves standard noise, standard sharpness, and standard distortion. Then, step S2 includes: Step S21: The medical image quality rapid improvement system extracts image blocks from corresponding positions in the image to be improved as the areas to be improved based on the differences between each area. It then uses an artificial intelligence analysis model to analyze the noise, sharpness, and distortion of each area to be improved. Subsequently, for each area to be improved, it calculates the noise difference between noise and standard noise, the sharpness difference between sharpness and standard sharpness, and the distortion difference between distortion and standard distortion. When the noise difference value is greater than a preset difference value threshold, noise is used as the indicator to be improved and associated with the area to be improved. When the sharpness difference value is greater than the difference value threshold, sharpness is used as the indicator to be improved and associated with the area to be improved. When the distortion difference value is greater than the difference value threshold, distortion is used as the indicator to be improved and associated with the area to be improved. Step S22: The medical image quality rapid improvement system improves the image quality of each area to be improved based on each indicator to be improved, and obtains the improved image after all image quality improvements are completed.

Images