Ultrasound system with improved noise performance resulting from continuous processing

By using continuous processors and noise bias coefficient weighting technology in ultrasonic imaging systems, the noise substrate is reduced by pixel level, solving the problem of difficulty in reducing noise substrates, and improving the signal-to-noise ratio of ultrasonic images and the sensitivity and clarity of images .

CN111656218BActive Publication Date: 2024-11-26KONINKLIJKE PHILIPS NV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201980010060.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2018-01-24
Filing Date
2019-01-15
Publication Date
2024-11-26
Estimated Expiration
2039-01-15

AI Technical Summary

Technical Problem

There is a problem that noise substrates are difficult to reduce in existing ultrasound imaging systems, resulting in poor signal-to-noise ratio of the image, affecting the sensitivity and time clarity of the image.

Method used

Using a persistent processor that can analyze the content of each pixel signal, the noise substrate is reduced by weighting the noise bias coefficient, and the pixel-by-pixel-level reduction of noise is achieved using a FIR or IIR filter network.

Benefits of technology

Effectively reduce the noise base, improve the signal-to-noise ratio of ultrasonic images, and enhance the sensitivity and time clarity of anatomical motion.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111656218B_ABST
    Figure CN111656218B_ABST
Patent Text Reader

Abstract

The ultrasound system generates a persistence image in response to both a persistence coefficient and a noise bias coefficient. The persistence coefficient controls the degree of persistence, which reduces noise variance through persistence processing. The noise bias coefficient is generated corresponding to the probability that an image pixel is noise, and operates to reduce the noise floor of the persistence image. A single user control enables the user to collaboratively control both the noise variance and the noise level of the image, or alternatively, dedicated controls may be made available to separately adjust the noise variance and the noise level.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to ultrasound systems, and in particular to ultrasound imaging systems that process images with persistence and reduce image noise. Background Art

[0002] There are many sources of noise that appear in ultrasound images as unwanted image artifacts. One source of noise is speckle noise that appears due to the coherent nature of ultrasound imaging. Speckle is caused by the intermodulation of signals from different signal paths in the image field, resulting in a speckled appearance of what should appear as uniform smooth tissue. Two methods widely used to reduce image speckle are frequency compounding as described in U.S. Pat. No. Re. 35148 (Lizzi et al.) and spatial compounding as described in U.S. Pat. No. 6126598 (Entrekin et al.). Another source of image noise is radio frequency radiation from nearby electrical equipment. This noise is reduced by shielding the ultrasound system and transducer from radio frequency interference and by wire filtering. Yet another source of noise is out-of-band noise in the transducer. This noise is reduced by shaping the passband used to receive and process the desired ultrasound signal. Another source of image noise is noise sources formed in the circuits and electrical components used in the ultrasound system, such as noise sources formed in amplifiers used to amplify received signals. These components and circuits have an inherent noise floor that is expected to be reduced to as low a level as possible. These latter noise sources can be reduced by combining continuously acquired images. Since much of this noise is actually random, combining pixels of consecutive images on a pixel-by-pixel basis will average out some of this noise. One image processing technique introduced for another purpose and which will reduce noise through image combination is called persistence processing. Many biological functions in the body will produce functions that are to be imaged only momentarily, and are often too rapid to be easily discerned in diagnosis. One such function is the maximum blood flow velocity in blood turbulence at peak systole, which occurs when blood flow pressure is at its maximum. Persistence is formed to assist in discerning such moments of peak blood flow, and does so by extending or persisting the appearance of such peak blood flow in color Doppler imaging. Several consecutive image frames of blood flow are continuously combined so that the peak flow appears in the image for several displayed frames, thereby increasing the likelihood that the clinician will discover the occurrence of the peak flow velocity. Each frame used in the combination has a persistence weighting factor, which causes its effect on the combination to decrease after several displayed images. Therefore, the blood flow peak will not appear in only one frame, but will persist, reducing the effect for several frames, making it easier to discern in the image sequence. Because persistence techniques perform their image combination on a pixel-by-pixel basis, the process will inherently average out random noise in the combined image as a function of the square root of the number of images combined. Speckle artifacts will also be reduced by averaging. Unfortunately, due to the inherent nature of the temporal averaging involved in persistence processing, this noise reduction only results in a reduction in the variation of noise artifacts; it does not reduce the average noise level (i.e., the "noise floor") itself. Therefore, it would be desirable to implement a noise reduction technique that has the effect of reducing the noise floor, thereby increasing the signal-to-noise ratio of the image to produce a more noise-free ultrasound image. Summary of the invention

[0003] According to the principles of the present invention, an ultrasound system employs a persistence processor that is capable of reducing the noise floor and thus improving the signal-to-noise ratio of an ultrasound image. The signal content of each pixel in the ultrasound image is analyzed relative to a signal-to-noise model to determine the likelihood that it is a signal or noise. The results of this analysis are used to generate a noise bias coefficient or weighting factor, which is applied to each pixel on a pixel-by-pixel basis during persistence processing. The result is an image with a noise floor that is reduced in proportion to the persistence applied. The inventive system utilizes lower persistence to produce enhanced noise reduction, which improves sensitivity and temporal clarity of ultrasound images in the presence of anatomical motion. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] In the attached picture:

[0005] Figure 1 The figure shows how noise can be reduced by conventional persistence processing or other image combining techniques.

[0006] Figure 2 An ultrasound system constructed in accordance with the principles of the present invention is illustrated in block diagram form.

[0007] Figure 3a and Figure 3b Illustrated is a technique for classifying image pixels according to the likelihood that they are noise.

[0008] Figure 4a , Figure 4b and Figure 4c Several conversion curves that may be used to create noise bias coefficients in accordance with the present invention are illustrated.

[0009] Figure 5 The persistence processor of the present invention is illustrated as being implemented as a finite impulse response (FIR) filter network.

[0010] Figure 6a and Figure 6b The persistence processor of the present invention is illustrated as being implemented as an infinite impulse response (IIR) filter network. DETAILED DESCRIPTION

[0011] Reference is made below to one or more processors and memory associated with the processors, wherein the processors execute functions according to instructions. It should be understood that a processor associated with a particular function as described herein may be the same or a different processor than another processor associated with the particular function. For example, one skilled in the art will appreciate that a processor or processors may include the processors described herein, such as, for example, an image processor, a noise offset coefficient processor, and a persistence processor.

[0012] Figure 1 The figure shows how noise can be reduced by continuous processing of the image or other image combination techniques. t , 50 t-1 and 50 t-2 50 represents three image pixels or groups of pixels at a common location in three consecutive ultrasound images. In this example, the pixels are all noise, the amplitude of which is shown by the height of the circles 52, 54 and 56 in the boxes and indicated by the numerical scale adjacent to each box. The values ​​of the noise in the three images are 2, 4 and 6 respectively. The average of the three noise values ​​is four, with a variance of four. If the persistence processor were to combine the first two images 50 t and 50 t-1 Similarly, if the persistence processor were to combine the second and third images 50 t-1 and 50 t-2 If there are 10 pixels in the image, the average will be five. Therefore, the noise values ​​will span the range of three to five, with a variance of two, but their average will still be four. Therefore, the variance of the noise values ​​has been reduced from a span of four to a span of two, but the noise level remains at four. Conventional persistence processors reduce the noise variance, not the noise level of the image.

[0013] Reference now Figure 2 , an ultrasound imaging system constructed according to the principles of the present invention is shown in block diagram form. A transducer array 12 is provided in an ultrasound probe 10 for transmitting ultrasound waves and receiving ultrasound echo information. The transducer array 12 can be a one or two-dimensional array of transducer elements that can be scanned in two or three dimensions (for example, in both elevation (in 3D) and azimuth). The two-dimensional array probe will include a microwave beamformer coupled to the array elements, which controls the transmission and reception of signals through the array elements. The microwave beamformer is capable of at least partially beamforming the signals received by the group or "tile" of transducer elements, as described in U.S. Patents US 5997479 (Savord et al.), US 6013032 (Savord) and US6623432 (Powers et al.). The microwave beamformer of the transducer element is coupled to the beamformer 14 via a probe cable. The transmission of ultrasound beams from the transducer array 12 is directed by a transmitter in the beamformer, either under the control of a microbeamformer when so equipped or directly in the absence of a microbeamformer, which receives input from operation of a user of the ultrasound system's user interface or control panel 28. Among the transmission characteristics controlled by the transmitter are the frequency, amplitude, phase, and polarity of the transmitted beam. The beam formed in the direction of pulsed transmission can be steered straight ahead (orthogonal to) from the transducer array or at different angles to scan a wider field of view.

[0014] The echoes received by a successive set of transducer elements are beamformed in a beamformer 14 by delaying them appropriately and then combining them. Analog beamformers are known, but modern ultrasound systems perform beamforming in the digital domain by converting the received echo signals into digital signal samples prior to beamforming. The partially beamformed signals produced by the microbeamformer are digitized by the beamformer and combined into fully beamformed coherent echo signals.

[0015] The coherent echo signal is coupled to a quadrature bandpass filter (QBP) 14. The QBP performs three functions: band limiting the RF echo signal data, generating in-phase and quadrature pairs (I and Q) of the echo signal data, and decimating the digital sampling rate. The QBP includes two separate filters, one filter generating in-phase samples and the other filter generating quadrature samples, wherein each filter is formed by a plurality of multipliers-accumulators implementing FIR filters. The quadrature signal samples undergo signal processing through a signal processor 18, which includes filtering caused by a digital filter and speckle reduction such as caused by spatial or frequency compounding. The signal processor can also shift the frequency band to a lower or baseband frequency range, such as can be a QBP. The signal processor can also distinguish signals in harmonic frequency bands by filtering or pulse inversion. For example, the digital filter of the signal processor 18 can be a filter of the type disclosed in U.S. Pat. No. 5,833,613 (Averkiou et al.).

[0016] The beamformed and processed coherent echo signals are coupled to one or more image processors. The B-mode processor 22 generates a B-mode image of structures (such as tissue) in the body. The B-mode processor generates a B-mode image of the structure (such as tissue) in the body by (I 2 +Q 2 ) 1 / 2The amplitude (envelope) detection of the I and Q signal components of the orthogonal demodulation is performed by calculating the echo signal amplitude in the form of . The B-mode processor also applies logarithmic compression to the B-mode image values. The orthogonal echo signal component is also coupled to the Doppler processor 24. The Doppler processor stores an ensemble of echo signals from discrete points in the image field, which are then used to estimate the Doppler shift at the point in the image by fast Fourier transform (FFT) processing. The Doppler shift is proportional to the motion at the point in the image field, such as blood flow and tissue motion. For color Doppler images, the estimated Doppler flow value at each point in the blood vessel is wall filtered and converted to a color value using a lookup table. The B-mode image signals and Doppler flow values ​​are coupled to the scan converter 20, which converts the B-mode and Doppler samples from the R-θ coordinates it acquires to Cartesian (x, y) coordinates for display in a desired display format (e.g., a straight line display format or a sector display format). Either the B-mode image or the Doppler image can be displayed alone, or both can be shown together in anatomical registration, with the color Doppler overlay showing blood flow in the tissue and vessels at the location in the image. The succession of received and processed ultrasound images is stored in the image memory 26 .

[0017] According to the principles of the present invention, image noise is reduced by a persistence processor 30 before display. Noise reduction uses a noise model 32 with data representing the noise level of pixels in the image, which can be configured in various ways. One way is to base the noise model on knowledge of the noise power versus gain of the system's receive amplifier. The noise power of the system amplifier as a function of gain is measured at the factory and aggregated in multiple modes for different probe apertures and gain levels. The noise model is stored in a memory in the ultrasound system and is accessed during imaging depending on the probe aperture used and the setting of the user gain control. Another way to configure the noise model is by cross-correlating consecutive received images from a fixed image field (such as a phantom or air). In the absence of noise, co-located pixels in the two images will be perfectly correlated, indicating the absence of noise. A low degree of correlation indicates a strong presence of noise. Therefore, a table is constructed and stored in a memory of correlation values ​​representing the degree of noise presence in the image field. The correlated consecutive images can be two-dimensional or one-dimensional (line) images.

[0018] Another technique for configuring the noise model is from the reception of an image without an ultrasound signal being transmitted. Ideally, the received image would present a complete lack of signal, but if any is present, it is assumed to be due to noise. A table of measured levels of pixels in the image that are assumed to be due to noise is stored in memory and used as a noise model. Yet another technique for configuring the noise model is to set the receiver filter to the noise portion of the received signal spectrum, frequencies where noise signal energy should not be present. A table of measured levels of pixels in the image corresponding to the noise-centered receive filter is then stored in memory as a noise model.

[0019] The pixels of the image to be processed according to the present invention are compared with the values ​​of the noise model to assess the likelihood that the pixel is noise. An example of this comparison is shown in Figure 3. Reference numeral 60 indicates the noise level at a pixel position for the noise model, and 62 and 64 are pixels of the image received at that position. Figure 3a Received image pixel 62 in is seen as being well above noise floor 60 and therefore has a low likelihood of being noise. For example, its noise likelihood index value may be 10%. Another received image pixel 64 has an amplitude very close to noise floor 60 and therefore has a high likelihood of being noise. For example, the pixel may be assigned a noise likelihood index (NLI) value of 95%.

[0020] Figure 3b Two other examples of noise likelihood classification are illustrated. Pixel 66 is relatively close to the noise floor 60 and may be assigned an NLI value of 25%. Pixel 68 has an amplitude below the noise floor 60 and its NLI value is 100%.

[0021] These noise likelihood estimates for pixels are generated in the noise model module 32 and used to create noise bias coefficients (N 偏置 Coeff). The NLI value is converted to a range of values ​​having the desired characteristic, an example of which is shown in FIG4. This is accomplished, for example, by using a mathematical formula or applying the NLI value as an input (address) to a lookup table containing output values ​​exhibiting the desired characteristic. The lookup table resides at Figure 2 N of the ultrasound system 偏置 In processor 34. Figure 4a The conversion characteristic is illustrated as a smooth curve 72. For high NLI values ​​(a strong possibility of noise), N 偏置 The processor generates high N 偏置 Coeff values, which start to decrease around the middle of the NLI value range and approach zero for low NLI values. Figure 4b Shows N 偏置The relationship between the Coeff value and the NLI value is a perfectly linear example, as can be seen by the straight line 74 . Figure 4c Given N 偏置 An example of how Coeff values ​​decrease in a stepwise manner as NLI values ​​decrease. Once the desired conversion characteristic is selected, it can be implemented as a lookup table where an input NLI value selects the corresponding NLI from the table. 偏置 Coeff value. These or other conversion characteristics may be used in embodiments of the present invention.

[0022] The noise offset coefficient is applied to the persistence processor 30 where it is used to reduce image noise during persistence processing of ultrasound images. Figure 2 The persistence processor 30 in the image processing apparatus also receives a persistence coefficient (PersistCoeff) from the persistence controller 36 in response to the user setting the persistence control on the control panel 28. The user can turn off persistence, in which case there is no persistence processing. The user can turn persistence on to a low or high range of persistence settings, which will increase the persistence coefficient generated by the persistence controller and produce a greater degree of persistence. A single value of PersistCoeff generated by the persistence controller in response to the control setting can be used to process the entire image and can be used for each image processed. For example, if there is little or no desired motion in the image (e.g., pulsating blood flow), there is no desired motion effect maintained by persistence and the user can decide to use low or no persistence. Alternatively, the lack of motion in the image means that there will be no image blur when a high degree of persistence is employed, and the user can then decide to use a high persistence setting to increase the reduction of noise in the image without worrying about image blur. Preferably, the persistence value for the processing of each pixel position varies both in space and time. For example, when a user is imaging the heart and trying to discern peak blood flow velocity, a low persistence degree may be applied to color Doppler images acquired at and only below peak systole to avoid blurring the systolic peak, and a higher persistence degree applied at other times and during diastole to achieve maximum noise reduction. At the same time, a lower persistence degree may be used in B-mode regions of the image where no flow exists to reduce the probability of blurring moving myocardial tissue. Other instances where persistence may be beneficially varied in space or time will readily occur to those skilled in the art.

[0023] According to the principles of the present invention, persistence processing uses persistence control (PersistCoeff, or in short notation, Pc) and noise level control (N 偏置 Coeff, or in short notation, Nb) both are performed. Figure 5 An example of a persistence processor using these two types of control is shown in FIG. Figure 5The diagram shows a continuous processor configured as an FIR filter. n ,I n-1 ,I n-2 , and I n-3 Pixels at common positions in a sequence of images I collected at are applied to the filter input. The pixel values ​​are weighted by persistence coefficients Pc0, Pc1, Pc2, and Pc3, respectively, and are also weighted by noise bias coefficients Nb0, Nb1, Nb2, and Nb3. Thus, each pixel of each image is processed using both persistence and noise bias coefficients, which may vary in time (across filter inputs) and space (from one FIR filter for one pixel position to another FIR filter for another pixel position). The doubly weighted pixel values ​​are summed at the summing node to produce a pixel value at the input image pixel I. n The output pixel O is generated at the time n Although the noise bias weight is shown applied as a product function, it may alternatively be applied as a summation value. Thus, the FIR filter implementation may take the following form when the noise bias coefficient is applied as a product function:

[0024]

[0025] Or when the noise offset factor is applied as a summation value it can take the following form:

[0026]

[0027] Figure 6a The continuous processor of the present invention is shown configured as an IIR filter. This example is a first-order IIR filter, where the input image I n The pixels are first multiplied by the weighting function 1-Pc(I n ). The weighted value is summed with the weighted delayed output value to generate the output value O n The previous output value is delayed by a delay D through one processing interval (one frame interval) to produce the previous output value O for the current calculation. n-1 . The previous output value O n-1 Weighted by f{Pc,Nb}, which is a value or expression that is a function of both the persistence coefficient (Pc) and the noise bias coefficient (Nb). That is, f{Pc,Nb} may be a product function, a sum function, or may combine the persistence and noise bias coefficients by some other mathematical operation. For example, a simple choice for f{Pc,Nb} may be Pc-Nb, where Nn is the noise bias coefficient, and Nb=0 corresponds to a default persistence algorithm that does not include a noise bias component.

[0028] The weighted previous output value is summed with the weighted input value at the summing node "+". Mathematically, this IIR filter performs the following algorithm:

[0029] O n =I n (1-Pc(In))+O n-1 (f{Pc,Nb})

[0030] The output of the persistence processor is coupled to a display processor 42 which suitably conditions the image for display on the image display 40 .

[0031] Figure 6b The diagram shows the persistence processor of the present invention configured as a second order IIR filter. n The pixels of I are weighted by the coefficient β0 and applied to the summing node. The pixels of the previous input image at the same position in the image have been delayed by a delay D (one image frame interval) and the I n-1 The pixels are weighted by the coefficient β1 and applied to the summing node. n-1 The frame is again delayed by a delay of D, and the I n-2 The pixel is weighted by the coefficient β2 and applied to the summing node. The output O generated at the summing node n Delay D to produce the previous output O n-1 , and delayed by a second delay D to produce output O before that n-2 These previous outputs are weighted by coefficients α1 and α2 which are functions of the noise coefficient f{Pc,Nb}, and the two weighted previous outputs are applied to the summing node. The algorithm performed by this second-order IIR filter has the following form:

[0032] O n =I n β0+I n-1 β1+I n-2 β2+O n-1 α1+O n-2 α2

[0033] Where β0, β1 and β2 are functions of the persistence coefficient(s), and α1 and α2 are functions of both the persistence coefficient(s) and the noise bias coefficient(s).

[0034] In a constructed embodiment, adjustment of the persistence control on the control panel 28 varies the noise variance by controlling PersistCoeff, and varies the noise variance by controlling N 偏置Coeff varies the noise floor. The persistence control can be implemented as a knob, a slider switch, or a virtual control on a touch screen. PersistCoeff is variable between 0.0 and 1.0, where 0.0 implements no persistence (persistence off) and 1.0 is the most aggressive persistence. Increasing the PersistCoeff value increases the weighting applied to the previous image, the number of images persisted, or both. The PersistCoeff value is also coupled to N 偏置 Processor 34, so that when more aggressive persistence is selected, more aggressive noise floor processing is. For example, when the user selects a relatively low persistence with a PersistCoeff value of 0.3, N 偏置 Processor 34 may implement Figure 4b However, when the user dials in a stronger persistence with a PersistCoeff setting of 0.8, N 偏置 The processor switches to a more aggressive noise floor curve (such as Figure 4a The use of the noise floor curve) produces a more aggressive N 偏置 Coeff value. Therefore, when the user turns up the persistence and improves the noise variance characteristics of the image, the noise floor characteristics, the signal-to-noise ratio, are successively improved. In a constructed embodiment, when the persistence control is turned up to increase the persistence, the noise in the image is seen to be reduced and noisy image areas become significantly darker and more noise-free. With the lower noise in the image, the user may decide to reduce the persistence to reduce the probability of image blur due to motion, or may decide to increase the gain of the image to better visualize fine structures that may otherwise be obscured by noise. It will be appreciated that the persistence control and the noise floor control may alternatively be controlled by separate controls rather than a single control as explained above.

[0035] Embodiments of the present invention can be used to improve ultrasound images in all imaging modes, including B-mode, color Doppler, color force, strain or shear wave elastic imaging, and contrast imaging. It will be appreciated that noise bias control may not be used in color Doppler and other parameter imaging modes to change pixel values, because that will not change the speed or other parameter values ​​indicated by the pixel. Instead, noise bias control is used to control color write priority in the color Doppler paradigm, which determines whether a pixel is to be displayed as speed or tissue or possibly mixed. For example, color values ​​near the noise base may not be displayed as the result of noise processing, and instead tissue (B-mode) pixel values ​​may be displayed.

[0036] It should be noted that the ultrasound system suitable for use in embodiments of the present invention and in particular Figure 2The component structures of the ultrasound system can be implemented in hardware, software, or a combination thereof. Various embodiments and / or components of the ultrasound system (e.g., noise model, N 偏置 Processor, persistent controller, and persistent processor, or the parts and controller therein) can also be implemented as part of one or more computers or microprocessors. A computer or processor can include a computing device, an input device, a display unit and an interface, for example, for accessing the Internet. A computer or processor can include a microprocessor. The microprocessor can be connected to a communication bus, for example, to access a PACS system or a data network for importing training images. A computer or processor can also include a memory. A memory device (such as an image memory 26 and a memory device for storing a noise model 32) can include a random access memory (RAM) and a read-only memory (ROM). A computer or processor can also include a storage device, which can be a hard disk drive or a removable storage drive, such as a floppy disk drive, an optical disk drive, a solid-state thumb drive, etc. A storage device can also be other similar modules for loading a computer program or other instructions into a computer or processor.

[0037] As used herein, the term "computer" or "module" or "processor" or "workstation" may include any processor-based or microprocessor-based system, including systems using microcontrollers, reduced instruction set computers (RISC), ASICs, logic circuits, and any other circuits or processors capable of performing the functions described herein. The above examples are exemplary only, and are therefore not intended to limit the definition and / or meaning of these terms in any way.

[0038] A computer or processor executes a set of instructions stored in one or more storage elements in order to process input data. Storage elements may also store data or other information as desired or required. Storage elements may be in the form of information sources or physical memory elements within a processing machine.

[0039] The instruction set of the ultrasound system, including those instructions for controlling the acquisition, processing and transmission of ultrasound images as described above, may include various commands that instruct a computer or processor as a processing machine to perform specific operations, such as the methods and processes of various embodiments of the present invention. The instruction set may be in the form of a software program. The software may take various forms, such as system software or application software, and may be embodied as a tangible and non-transient computer-readable medium. In addition, the software may take the form of a collection of separate programs or modules, a program module within a larger program, or a portion of a program module. Those skilled in the art will recognize that, for example, a persistent processor may be best implemented as a software module. The software may also include modular programming in the form of object-oriented programming. The processing of input data by the processing machine may be in response to an operator command, or in response to the results of a previous process, or in response to a request from another processing machine.

[0040] Furthermore, the following claim limitations are not written in a means-for-function format and are not to be interpreted based on 35 U.S.C. 112, sixth paragraph, unless and until such claim limitations expressly use the phrase "means for..." followed by a description of functionality without further structure.

Claims

1. An ultrasound imaging system for producing images with improved noise performance, comprising: An ultrasound probe adapted to collect ultrasound echo signals from an image field for imaging; an image processor coupled to the probe and adapted to generate an ultrasound image; a memory comprising a noise model representing noise levels at pixel locations of the ultrasound image, wherein the noise model is further adapted to be compared with pixel values ​​of the image to determine a likelihood that the pixel value is noise; A noise offset coefficient processor adapted to generate the noise offset coefficient by: comparing the pixel value of the ultrasound image to the noise model to determine the likelihood that the pixel is noise; and generating a noise bias coefficient as a function of the likelihood that the pixel value is noise; as well as A persistence processor is adapted to improve noise performance of an ultrasound image based on a persistence control coefficient and the noise bias coefficient.

2. The ultrasound imaging system of claim 1, comprising a user-operated continuity control coupled to the continuity processor and adapted to control the continuity control coefficient.

3. The ultrasound imaging system according to claim 1, wherein: The noise model is based at least in part on receive amplifier noise of the ultrasound imaging system, or from correlations between pixel values ​​in the ultrasound image.

4. The ultrasound imaging system according to claim 1, wherein: The noise model is based at least in part on ultrasound image data acquired without ultrasound signal transmission, or ultrasound image data from a noise spectrum.

5. The ultrasound imaging system according to claim 1, wherein: The noise offset coefficient processor is further adapted to generate noise offset coefficients over a range exhibiting desired conversion characteristics between the likelihood that a pixel value is noise and the noise offset coefficients.

6. The ultrasound imaging system according to claim 5, wherein: The desired transfer characteristic also includes a curve.

7. The ultrasound imaging system according to claim 1, wherein: The persistence processor also includes one of the following: a finite impulse response filter or an infinite impulse response filter.

8. A method for producing an image with improved noise performance, comprising: Acquiring ultrasonic echo signals from the image field for imaging; generating an ultrasonic image according to the ultrasonic echo signal; obtaining a noise level model for pixels of the ultrasound image, wherein the noise level model is adapted to be compared with pixel values ​​of the image to determine a likelihood that the pixel values ​​are noise; The noise offset coefficient is generated by: comparing the pixel value of the ultrasound image to the noise level model to determine the likelihood that the pixel is noise; generating a noise bias coefficient as a function of the likelihood that the pixel value is noise; The ultrasound image is persistence processed in response to both the noise bias coefficient and the persistence coefficient to improve noise performance of the image using the persistence processing.

9. The method according to claim 8, wherein: Generating the noise level model includes: using system amplifier noise properties; performing frame-to-frame image correlation; receiving an image without transmitting; receiving an image from a noise spectrum; or a combination thereof.

10. The method according to claim 8, wherein: The persistence processing includes processing the image using a finite impulse response filter or an infinite impulse response filter that operates in response to both the noise offset value and the user persistence control input.