WO2023061910A1

WO2023061910A1 - Improving image quality of medical images

Info

Publication number: WO2023061910A1
Application number: PCT/EP2022/078045
Authority: WO
Inventors: Nitesh KAUSHAL; Aliarshad Aameer KOTHAWALA; Giridhar Narasapura RAJAGOPALAIAH; Soumabha BHOWMICK; Subhendu Seth; Pallavi Vajinepalli; Anne Woodside HOLMES; Earl Monroe CANFIELD II; David Nigel ROUNDHILL; Matthew Robert RIELLY
Original assignee: Koninklijke Philips N.V.
Priority date: 2021-10-14
Filing date: 2022-10-10
Publication date: 2023-04-20

Abstract

A method for modifying imaging parameters of a medical imaging system to improve the image quality of medical images. The method comprises obtaining a plurality of medical images and a corresponding set of imaging parameters for each medical image, wherein each set of imaging parameters is different. For each medical image, processing, using one or more image analysis algorithms, the medical image to thereby obtain one or more confidence scores and further processing, using an optimization algorithm, the one or more confidence scores and the corresponding set of imaging parameters for each medical image to thereby obtain an improved set of parameters.

Description

IMPROVING IMAGE QUALITY OF MEDICAL IMAGES

FIELD OF THE INVENTION

The invention relates to improving the image quality of medical images. In particular, the invention relates to modifying imaging parameters of medical images in order to improve the image quality.

BACKGROUND OF THE INVENTION

Modem Ultrasound scanners are equipped with several features and parameters to optimize ultrasound B-mode images. Due to the high operator dependency involved in the adjustment of such features and parameters (with an aim to improve image quality), experienced sonographers tend to achieve high quality ultrasound images which depict the underlying anatomical structures of interest while novice ultrasound operators struggle to achieve the same results.

To bridge this gap, as well as to ease the present workflow, artificial intelligence (Al) based ultrasound workflows are gaining importance and targeting routine clinical applications. Although the assessment of the clinical image quality of ultrasound B-mode images are generally performed subjectively, bringing objectivity in this assessment is an essential pre-requisite to scale-up and standardize ultrasound workflows. Based on the availability of reference images, these objective image quality metrics may be classified into full reference (i.e. with reference images) and no-reference (i.e. without reference images) metrics.

Although full reference image quality metrics demonstrate high correlation with a subjective judgement, absence of the reference images limits its adaptability. In such context, noreference image quality metrics may be the next choice. However, no-reference metrics do not always correlate well with the perceived ultrasound B-mode image quality. Hence, it is important to have a robust solution to formulate the no-reference image quality metric, particularly to accomplish objective assessment of the ultrasound image quality.

Several articles in the literature have demonstrated the adversarial effects of deep learning models in the domain of natural scenes (i.e. small perturbations of the input vector causing large changes in output). Such adversarial effects are inevitable in ultrasound imaging. For example, significant changes to the output of deep learning models can be made by perturbing the ultrasound end-user parameters (e.g. gain, TGC, focus, zoom etc.).

Thus, while designing the deep learning models, actions are usually taken to address such scenarios (e.g. adding representative dataset of such adversarial scenarios in training set). However, that usually complicates the overall model behavior. Hence, there is a need to have a mechanism of limiting the possibilities of such adversarial effects in ultrasound imaging and medical imaging more generally, ultimately boosting the throughput of Al models.

SUMMARY OF THE INVENTION

The invention is defined by the claims.

According to examples in accordance with an aspect of the invention, there is provided a method for modifying imaging parameters of a medical imaging system to improve the image quality of medical images, the method comprises: obtaining a plurality of medical images and a corresponding set of imaging parameters for each medical image, wherein each set of imaging parameters is different; for each medical image, processing, using one or more image analysis algorithms, the medical image to thereby obtain one or more confidence scores; and processing, using an optimization algorithm, the one or more confidence scores and the corresponding set of imaging parameters for each medical image to thereby obtain an improved set of parameters.

The quality of a medical image often depends on the imaging parameters used. For example, the imaging parameters may comprise transmit receive parameters (e.g. radiofrequency (RF), in-phase quadrature, signal image path (SIP) etc.), post-processing parameters (e.g. quadrature bandpass (QBP) filters, post-processing modules (e.g. COSMIX) etc.) and/or end-user exposed parameters (e.g. focus, gain, dynamic range, zoom, depth etc.). Image analysis algorithms exist which can provide information on the medical image. For example, some image analysis algorithms may identify or classify particular objects in the medical image. Conventionally, image analysis algorithms provide a confidence score (i.e. how confident the algorithm is that is how identified/classified an object correctly).

The inventors realized that the higher the image quality of a medical image, the higher the confidence score from the image analysis algorithms. Thus, the confidence scores output by image analysis algorithms can be used as the evaluation criteria in the objective function of an optimization algorithm used to optimize the set of imaging parameters. In other words, the set of imaging parameters can be improved for image quality by maximizing the confidence scores from the image analysis algorithms.

In a first embodiment, the plurality of medical images all correspond to a first image with a different set of imaging parameters. The set of imaging parameters may be iteratively altered (thus generating further medical images) by the optimization algorithm.

In a second embodiment, the plurality of images are obtained in real time and the imaging parameters contain at least transmit receive imaging parameters which are altered over time by, for example, the optimization algorithm.

The medical images may be of an anatomical region. The imaging parameters of a medical image may define visual aspects of the medical image. For example, the imaging parameters of a medical image may define the relationship between the pixels of the medical image, the relationship between an initial value of a pixel and a final value of the pixel and/or which pixels are shown.

The image analysis algorithms may be configured to detect and/or identify regions of interest, objects of interest and/or anatomies of interest in the medical image.

The optimization algorithm may be a heuristic optimization algorithm.

Heuristic optimization algorithms are not concerned with finding the global maximum/maxima of the objective function and, instead, may find a/the local maximum/maxima which approximates the global maximum/maxima. This significantly reduces the processing resources needed for the optimization algorithm and enables the optimization algorithm to output an optimal set of parameters quicker. Additionally, the quicker processing times may further enable the optimization algorithm to be used in real time.

The heuristic optimization algorithm may be one of an iterative optimization algorithm and a gradient descent optimization algorithm.

The one or more image analysis algorithms may comprise one or more of an artificial intelligence based object detection algorithm and an artificial intelligence based classification algorithm.

At least one of the one or more image analysis algorithms may be a you-only-look-once, YOLO, object detection algorithm.

The method may further comprise transmitting one or more imaging parameters in a set of imaging parameters to a server, wherein processing the one or more confidence scores and the corresponding set of imaging parameters via an optimization algorithm (110) comprises processing, at the server, the transmitted imaging parameters and the corresponding sets of imaging parameters with first optimization algorithm, thereby to obtain a first improved subset of imaging parameters and processing the non-transmitted imaging parameters and the corresponding sets of imaging parameters with a second optimization algorithm, thereby to obtain a second improved subset of imaging parameters.

In some cases, in order to enable the optimization process to be done in real time, some of the parameters may be improved online at a server and the remaining imaging parameters can be improved on-site. For example, the transmit receive parameters could be improved online whilst the enduser exposed parameters could be improved offline (i.e. on-site). Imaging parameters corresponding to default presets may be improved online such that the default presets can be improved.

Additionally, in some cases, the medical images can be transmitted to a server such that the confidence scores (by inputting the medical images to image analysis algorithm(s)) can be obtained online. Thus, offline (on-site) processing is reduced.

The medical images may be ultrasound B-mode images. The invention further provides a computer program product comprising computer program code means which, when executed on a computing device having a processing system, cause the processing system to perform all of the steps of the afore-mentioned method.

The invention further provides a device for modifying imaging parameters of a medical imaging system to improve the image quality of medical images, the device comprising one or more processors configured to: obtain a plurality of medical images and a corresponding set of imaging parameters (102) for each medical image, wherein each set of imaging parameters is different; for each medical image, process, using one or more image analysis algorithms, the medical image to thereby obtain one or more confidence scores; and process, using an optimization algorithm, the one or more confidence scores and the corresponding set of imaging parameters for each medical image to thereby obtain an improved set of parameters.

The optimization algorithm may be a heuristic optimization algorithm.

The one or more processors may be further configured to transmit one or more imaging parameters in a set of imaging parameters to a server, wherein processing the one or more confidence scores and the corresponding set of imaging parameters via an optimization algorithm comprises processing, at the server, the transmitted imaging parameters and the corresponding sets of imaging parameters with first optimization algorithm, thereby to obtain a first improved subset of imaging parameters and processing the non-transmitted imaging parameters and the corresponding sets of imaging parameters with a second optimization algorithm, thereby to obtain a second improved subset of imaging parameters.

The medical images may be ultrasound B-mode images.

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.

BRIEF DESCRIPTION OF THE DRAWINGS

For a better understanding of the invention, and to show more clearly how it may be carried into effect, reference will now be made, by way of example only, to the accompanying drawings, in which: Figure 1 shows a method for modifying the imaging parameters of a medical image according to an embodiment of the invention;

Figure 2 shows a first example of the changes to confidence scores when changing imaging parameters;

Figure 3 shows a second example of the changes to confidence scores when changing imaging parameters; and

Figure 4 shows experimental results for the optimization method according to embodiments of the invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

The invention will be described with reference to the Figures.

It should be understood that the detailed description and specific examples, while indicating exemplary embodiments of the apparatus, systems and methods, are intended for purposes of illustration only and are not intended to limit the scope of the invention. These and other features, aspects, and advantages of the apparatus, systems and methods of the present invention will become better understood from the following description, appended claims, and accompanying drawings. It should be understood that the Figures are merely schematic and are not drawn to scale. It should also be understood that the same reference numerals are used throughout the Figures to indicate the same or similar parts.

The invention provides a method for modifying imaging parameters of a medical imaging system to improve the image quality of medical images. The method comprises obtaining a plurality of medical images and a corresponding set of imaging parameters for each medical image, wherein each set of imaging parameters is different. For each medical image, processing, using one or more image analysis algorithms, the medical image to thereby obtain one or more confidence scores and further processing, using an optimization algorithm, the one or more confidence scores and the corresponding set of imaging parameters for each medical image to thereby obtain an improved set of parameters.

Embodiments are based on the realization that confidence scores of one or more image analysis algorithms can serve as an evaluation metric for an optimization algorithm (to evaluate different imaging parameters).

Figure 1 shows a method 100 for modifying the imaging parameters 102 of medical imaging system used to generate a medical image 104 according to an embodiment of the invention. The characterization of a medical image (e.g. ultrasound B-mode image) is based on the combination of an appropriate probe placement with respect to the underlying anatomy being imaged as well as appropriate adjustment of the imaging parameters 102 in the imaging device (which enhances the overall image quality).

Medical imaging, and in particular ultrasound imaging, is a highly operator-dependent imaging modality. Bringing objectivity into the probe placement and adjustment of the imaging parameters 102 possess a challenge. Embodiments of the present invention solve the problem of imaging parameter 102 adjustment.

The following embodiments and examples refer to ultrasound B-mode images as the medical images 102. However, it should be understood that other types of medical images 102 may also be used (e.g. CT images, MRI images, Doppler ultrasound etc.)

Conventional ultrasound systems come with pre-programmed default presets (i.e. TSP) as the set of imaging parameters 102. It should be understood that the set of imaging parameters 102 may contain all of the imaging parameters of a TSP or only a subset of imaging parameters of the TSP. Different TSPs 102 are improved for specific types of ultrasound scans. The user may be given access to some imaging parameters of a TSP 102 in the form of Knobology parameters (e.g. focus, gain, dynamic range, zoom, depth, etc.) for further fine tuning of these imaging parameters to achieve an improved image quality. These additional controls make the ultrasound imaging modality highly skill-oriented (along with the anatomical sense of appropriate probe placement).

Imaging parameters affect how a medical image is visualized. Imaging parameters change visual aspects of the medical image without changing the content of the medical image.

Embodiments of the invention propose ultrasound B-mode image quality improvement as an objective optimization scheme where the set of imaging parameters 102 are modified to accomplish an improved image quality (e.g. for subsequently obtained images and/or reprocessed images).

The method 100 comprises a step 104 of obtaining medical images. Each medical image is associated with a different set of imaging parameters. In some examples, each medical image is of a same scene (e.g. imaging a same anatomical feature/area), but obtained produced using a different set of imaging parameters.

The present disclosure proposes to use an optimization algorithm to select or define the most appropriate imaging parameters to improve the quality of (future) imaging.

Defining a robust evaluation metric for an objective function is a pre-requisite of any optimization process. The present disclosure proposes the use of the confidence scores 108 output from one or more image analysis algorithms 106 as the evaluation metric in the objective function of an optimization algorithm 110.

Thus, the method 100 processes the medical images 104 obtained in step 105 using image analysis algorithm(s) 106 (e.g. classification algorithm, object detection algorithm, segmentation algorithm etc.). Each image analysis algorithm 106 outputs a confidence score 108 which corresponds to how “confident” the image analysis algorithm is of its main output (e.g. the object detection box or classification of anatomy). It will be appreciated that each confidence score is associated with a particular image and, therefore, a particular set of imaging parameters.

The method 100 then processes the confidence score(s) 108, relative to the set of imaging parameters 102, using an optimization algorithm. In particular, the confidence scores are used as the evaluation metric in the objective function of the optimization algorithm 110. The optimization algorithm 110 finds a maximum in the confidence score(s) 108 and the corresponding set of imaging parameters 102 for the maximum in the confidence scores 108 can be used to obtain or define the improved imaging parameters 112. In some embodiments, the set of imaging parameters 102 corresponding to the maximum in the confidence scores 108 is selected as the improved imaging parameters 112.

Preferably, the optimization algorithm 110 is a heuristic optimization algorithm (e.g. iterative optimization, gradient descent etc.). A heuristic optimization algorithm will search for a maximum of the objective function which may be a local maximum or a global maximum instead of explicitly searching for the global maximum. This allows the optimization algorithm 110 to find improved imaging parameters 112 with less processing resources and faster than non-heuristic optimization algorithms.

One example of a suitable heuristic optimization algorithm is the Particle Swarm Optimization (PSO) algorithm, such as those disclosed by Trelea, Ioan Cristian. "The particle swarm optimization algorithm: convergence analysis and parameter selection." Information processing letters 85.6 (2003): 317-325.

Based on the complexity and the pre-requisites for real-time operations, strategically selected imaging parameters (from transmit-scheme parameters to post-processing parameters, including the end-user exposed imaging parameters) can be improved online, while some of the remaining imaging parameters can be improved offline by obtaining the data in interim stages of the ultrasound signal processing pipeline.

Testing:

Figure 2 and Figure 3 (further detailed below) show a combinatorial test of how changing imaging parameters affects confidence scores.

Figure 2 shows a first example of the changes to confidence scores when changing imaging parameters. The data points 202 show how changes to the imaging parameters of an ultrasound B-mode image affect the confidence scores of an image analysis algorithm. The y-axis shows the confidence scores output from an image analysis algorithm and the x-axis shows the number of changes to the imaging parameters.

The imaging parameters changed in Figure 2 are internal clinical optimizers not exposed to the end-users related to post-processing parameters in the ultrasound signal image path (SIP). The interdependencies between imaging parameters can be analyzed based on the patterns and confidence values shown by the data points 202.

In this case, the image analysis algorithm is an artificial intelligence (Al) based you only look once (Y OLO) object detection algorithm. An example YOLO object detection algorithm can be found in the paper by Redmon, Joseph, et al. "You only look once: Unified, real-time object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016. Another example YOLO objection detection algorithm is suggested by Shafiee, Mohammad Javad, et al. "Fast YOLO: A fast you only look once system for real-time embedded object detection in video." arXiv preprint arXiv: 1709.05943 (2017).

The data points 202 show that changing the post-processing parameters post the signalimage path (SIP) affects the confidence score of the Al algorithm running on reconstructed B-mode ultrasound image. Thus, if the model is trained on good quality ultrasound B-mode images (majorly including the clinical variability), then the confidence values output by the image analysis algorithm can work as an image quality indicator of the optimization process.

If the training dataset of the image analysis algorithms are good quality, then they can be leveraged as an objective function for achieving better image quality via the optimization algorithm.

Moreover, to achieve improved robustness, some current image analysis algorithms are forcefully trained with bad quality images, making the image analysis algorithms more complex. It would therefore be preferable to ensure that only good quality images are used for image analysis training and inference, to simplify the image analysis algorithms and reduce processing resource required. The present disclosure provides a mechanism by which images of improved quality can be selected.

Ultrasound B-mode images 204 and 206 show the difference in image quality between images associated with a relatively high and low confidence score. In the illustrated example, image 204 has a relatively high confidence score and has a relatively high image quality compared to image 206.

The changes in the post-processing imaging parameters results in the sharpening effect of image 204, whereas blurred image 206 (which is therefore of lower quality) corresponds to the extreme (right) end of the data points 202. This clearly indicates that images that are associated with a lower confidence score have a relatively low image quality.

Figure 3 shows a second example of the changes to confidence scores when changing imaging parameters. The data points 302 show how changes to the imaging parameters of an ultrasound B-mode image affect the confidence scores of an image analysis algorithm. The y-axis shows the confidence scores output from an image analysis algorithm and the x-axis shows the number of changes to the imaging parameters.

In this case, the image analysis algorithm is an Al based classification algorithm. The data points 302 show the effect on the confidence scores of Al classification algorithms by changing some of the post processing parameters post signal image path (SIP).

Ultrasound B-mode images 304 and 306 show the difference in image quality between a relatively high and low confidence score. Image 304 has a relatively high confidence score and has a relatively high image quality when compared to image 306.

Figures 2 and 3 showcase the confidence score plot of a YOUO algorithm and Classification algorithm (both for fetal heart), respectively, that process different ultrasound B-mode images, in which different images are obtained by changing the post-processing imaging parameters of the same raw imaging data. The characteristic saw-tooth pattern depicts interesting observations. Similar sorts of patterns can be found with traditional full reference image quality metrics in the field of ultrasound, like point spread functions, signal to noise ratio or contrast to noise ratio upon changing the imaging parameters. The similar pattern strengthens the application of confidence scores as a noreference image quality metric in this context.

The YOLO algorithm used to obtain the confidence scores in Figure 2 was trained with sharper object structures whereas the Classification algorithm used to obtain the confidence scores in Figure 3 was built with moderate (i.e. neither sharp nor blurred) images as input.

The YOLO algorithm used for Figure 2 was biased to sharp representations of the object of interest, hence the confidence score is higher at the sharper side (i.e. left side of x-axis) whereas it drops more towards the end (i.e. right side of x-axis) where the image gets blurred. The confidence scores output by the YOLO algorithm highlight and capture the local property of the ultrasound B-mode images, making the image quality tuned locally (i.e. for local structures in the ultrasound images).

In the case of the Classification algorithm used for Figure 3, it usually takes the overall appearance of the image at the training stage. Hence, a boosted confidence score can be found somewhere in the middle of the data points 302. The confidence scores output by the Classification algorithm stress the global property of the ultrasound B-mode images, making image quality tuned globally (i.e. for the entire ultrasound image).

Other statistical metrics (e.g. BRISQUE, NIQE etc.) or even opinion aware Al algorithms can be used as the image analysis algorithms. Statistical based Al algorithms may help to optimize confidence scores when the regions of interest are not visible in the medical image.

Embodiments of the invention effectively bridge the fields of Al models (for certain anatomical findings) with the desired ultrasound B-mode image quality, thus boosting both the entities together. In other words, the optimization algorithm will output an improved set of imaging parameters which will improve image quality and will also boost the confidence score of the Al algorithm(s).

The data points 202 and 302 shown in Figure 2 and Figure 3 respectively show the combinatorial testing phase according to embodiments of the invention. The testing phase was performed to analyze the change in the value of a confidence score by changing a limited number of imaging parameters.

For combinatorial testing, ordering of the changes in imaging parameters is irrelevant. Ultimately, for any combination of imaging parameters, it yields a definite confidence score. However, for algorithmic optimization, ordering might help in reducing the search space for fine-tuning. In present clinical practice, experts usually follow a particular order while manually changing the imaging parameters, thereby achieving an improved image quality. Using the existing knowledge and understanding the inter-dependencies of imaging parameters, a certain or predetermined order could be imposed while executing an algorithmic optimization process. Results:

Figure 4 shows experimental results for the method according to embodiments of the invention. The Particle Swarm Optimization (PSO) algorithm was used as the optimization algorithm, and provides one suitable example of an optimization algorithm. However, any other suitable form of optimization algorithm could be used.

The post-processing imaging parameters after signal image path (SIP) and the end-user imaging parameters (e.g. gain, dynamic range, focus etc.) were changed in the following experimental results.

Figure 4 (a) shows a first image 402a with default imaging parameters (i.e. default TSP) and a second image 402b with improved imaging parameters (improved TSP) produced by performing a method according to an embodiment. In the present scenario, the imaging parameters changed in Figure 4 (a) correspond to internal clinical optimizers that are not exposed to the end-users related to postprocessing parameters in the ultrasound signal path (SIP). By way of example, the default TSP is [1.87, 0.45, -2, 0] and the improved TSP, found following application of the PSO algorithm in accordance with an embodiment, is [2, 0.75, 0, 2], The image analysis algorithms used (to produce the confidence scores used by the optimization algorithm) were a heart detection algorithm trained on phantom data and a heart detection algorithm trained on clinical data. The confidence scores are out of 100:

Confidence score (phantom data) for image 402a with default TSP - 96.15

Confidence score (phantom data) for image 402b with improved TSP - 96.19

Confidence score (clinical data) for image 402a with default TSP - 98.99

Confidence score (clinical data) for image 402b with improved TSP - 99.84

Figure 4 (b) shows a third image 404a with a default TSP and a fourth image 404b with an improved TSP. The imaging parameters correspond to [chronologically focus, dynamic range, gain]. The default TSP is [3, 55, 63] and the improved TSP, found by the PSO algorithm, is [3, 74, 67], The image analysis algorithm used was a heart detection algorithm trained on phantom data. The confidence scores are out of 100:

Confidence score (phantom data) for image 404a with default TSP - 89.08

Confidence score (phantom data) for image 404b with improved TSP - 92.14

Figure 4 (c) shows a fifth image 406a with a default TSP and a sixth image 406b with an improved TSP. The imaging parameters correspond to [chronologically focus, dynamic range, gain, zoom]. The default TSP is [3, 55, 63, 1] and the improved TSP, found by the PSO algorithm, is [3, 70, 65, 1.3]. The image analysis algorithm used was a heart detection algorithm trained on phantom data. The confidence scores are out of 100:

Confidence score (phantom data) for image 406a with default TSP - 89.08

Confidence score (phantom data) for image 406b with improved TSP - 99.80

Despite the PSO algorithm being used the optimization algorithm, other iterative optimization algorithms may be used instead, such as a genetic algorithm, simulated annealing, Antcolony optimization etc. Alternatively, other optimization techniques like gradient descent could also be used.

In summary, embodiments of the invention propose a modification to the set of imaging parameters of a medical image in order to improve the image quality of a medical image. Imaging parameters can be improved in an offline/online mode (i.e. on-site and/or at a server) to achieve a perceivable and diagnostically-relevant improvement in image quality.

Embodiments of the invention leverage the confidence scores of one or more Al models (e.g. classification algorithms, object detection algorithms, segmentation algorithms etc.) as the evaluation criteria for an optimization algorithm. Classification algorithms usually evaluate the global appearance of a medical image whereas object detection and segmentation algorithms emphasize the property of local structures in the image. A combination of these Al models can be utilized to perform global and localized image quality optimization.

Modifying the set of image parameters for a medical image in this way also boosts (i.e. improves) the confidence score of the Al algorithms running on the medical image for any specific clinical findings.

The skilled person would be readily capable of developing a processor or device for carrying out any herein described method. Thus, each step of the flow chart may represent a different action performed by a processor or device, and may be performed by a respective module of the processor or device.

Embodiments may therefore make use of a processor or device. The processor or device can be implemented in numerous ways, with software and/or hardware, to perform the various functions required. A processor or device is one example of a processor or device which employs one or more microprocessor or devices that may be programmed using software (e.g., microcode) to perform the required functions. A processor or device may however be implemented with or without employing a processor or device, and also may be implemented as a combination of dedicated hardware to perform some functions and a processor or device (e.g., one or more programmed microprocessor or devices and associated circuitry) to perform other functions. Examples of processor or device components that may be employed in various embodiments of the present disclosure include, but are not limited to, conventional microprocessor or devices, application specific integrated circuits (ASICs), and field-programmable gate arrays (FPGAs).

In various implementations, a processor or device or processor or device may be associated with one or more storage media such as volatile and non-volatile computer memory such as RAM, PROM, EPROM, and EEPROM. The storage media may be encoded with one or more programs that, when executed on one or more processor or devices and/or processor or devices, perform the required functions. Various storage media may be fixed within a processor or device or processor or device or may be transportable, such that the one or more programs stored thereon can be loaded into a processor or device or processor or device.

It will be understood that disclosed methods are preferably computer-implemented methods. As such, there is also proposed the concept of a computer program comprising code means for implementing any described method when said program is run on a processor or device, such as a computer. Thus, different portions, lines or blocks of code of a computer program according to an embodiment may be executed by a processor or device or computer to perform any herein described method. In some alternative implementations, the functions noted in the block diagram(s) or flow chart(s) may occur out of the order noted in the Figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.

Variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or device or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. If a computer program is discussed above, it may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. If the term "adapted to" is used in the claims or description, it is noted the term "adapted to" is intended to be equivalent to the term "configured to". Any reference signs in the claims should not be construed as limiting the scope.

Claims

CLAIMS:

1. A method for modifying imaging parameters of a medical imaging system to improve the image quality of medical images (104), the method comprises: obtaining a plurality of medical images (104) and a corresponding set of imaging parameters (102) for each medical image (104), wherein each set of imaging parameters (102) is different; for each medical image (104), processing, using one or more image analysis algorithms (106), the medical image (104) to thereby obtain one or more confidence scores (108); and processing, using an optimization algorithm (110), the one or more confidence scores (108) and the corresponding set of imaging parameters for each medical image (104) to thereby obtain an improved set of parameters (112).

2. The method of claim 1, wherein the optimization algorithm (110) is a heuristic optimization algorithm.

3. The method of claim 2, wherein the heuristic optimization algorithm is one of: an iterative optimization algorithm; and a gradient descent optimization algorithm.

4. The method of any one of claims 1 to 3, wherein the one or more image analysis algorithms (106) comprise one or more of: an artificial intelligence based object detection algorithm; and an artificial intelligence based classification algorithm.

5. The method of claim 4, wherein at least one of the one or more image analysis algorithms (106) is a you-only-look-once, YOLO, object detection algorithm.

6. The method of any one of claims 1 to 5, further comprising transmitting one or more imaging parameters in a set of imaging parameters (102) to a server, wherein processing the one or more confidence scores (108) and the corresponding set of imaging parameters (102) via an optimization algorithm (110) comprises: processing, at the server, the transmitted imaging parameters and the corresponding sets of imaging parameters with first optimization algorithm, thereby to obtain a first improved subset of imaging parameters; and processing the non-transmitted imaging parameters and the corresponding sets of imaging parameters with a second optimization algorithm, thereby to obtain a second improved subset of imaging parameters.

7. The method of any one of claims 1 to 6, wherein the medical images (104) are ultrasound B-mode images.

8. A computer program product comprising computer program code means which, when executed on a computing device having a processing system, cause the processing system to perform all of the steps of the method according to any of claims 1 to 7.

9. A device for modifying imaging parameters of a medical imaging system to improve the image quality of medical images (104), the device comprising one or more processors configured to: obtain a plurality of medical images (104) and a corresponding set of imaging parameters (102) for each medical image (104), wherein each set of imaging parameters (102) is different; for each medical image (104), process, using one or more image analysis algorithms (106), the medical image (104) to thereby obtain one or more confidence scores (108); and process, using an optimization algorithm (110), the one or more confidence scores (108) and the corresponding set of imaging parameters for each medical image (104) to thereby obtain an improved set of parameters (112).

10. The device of claim 9, wherein the optimization algorithm (110) is a heuristic optimization algorithm.

11. The device of claim 10, wherein the heuristic optimization algorithm is one of: an iterative optimization algorithm; and a gradient descent optimization algorithm.

12. The device of any one of claims 9 to 11, wherein the one or more image analysis algorithms (106) comprise one or more of: an artificial intelligence based object detection algorithm; and an artificial intelligence based classification algorithm. 15

13. The device of claim 12, wherein at least one of the one or more image analysis algorithms (106) is a you-only-look-once, YOLO, object detection algorithm.

14. The device of any one of claims 9 to 13, wherein the one or more processors are further configured to transmit one or more imaging parameters in a set of imaging parameters (102) to a server, wherein processing the one or more confidence scores (108) and the corresponding set of imaging parameters (102) via an optimization algorithm (110) comprises: processing, at the server, the transmitted imaging parameters and the corresponding sets of imaging parameters with first optimization algorithm, thereby to obtain a first improved subset of imaging parameters; and processing the non-transmitted imaging parameters and the corresponding sets of imaging parameters with a second optimization algorithm, thereby to obtain a second improved subset of imaging parameters.

15. The device of any one of claims 9 to 14, wherein the medical images (104) are ultrasound

B-mode images.