Automatically modulating radiation dose based on ambiguous noise and behavior AI detection

Abstract

A system (30) for modulating a radiation dose during an X-ray guided process with an X-ray imaging system (50) includes a detector (32) and a controller (74). The detector (32) is configured to detect at least one noise generating event or user behavior event during an X-ray guided procedure with the X-ray imaging system (50), which makes the image quality for X-ray imaging not required to be above a threshold image quality. The controller (74) is configured to automatically modulate X-ray system parameters that affect image quality and radiation dose based on detection of at least one of a noise generation event or a user behavior event.

Description

Background Art

[0001] This embodiment generally relates to X-ray image-guided minimally invasive surgery, and more particularly to a system and method for automatically modulating radiation dose based on AI detection of blur noise and behavior.

[0002] During X-ray image-guided minimally invasive surgery, the use of fluoroscopy is manually controlled by the operating physician. Typically, this control is binary, i.e., the X-ray is turned on or off based on the user's interaction with a single pedal / button on the system, and the interventional physician does not have easy access to fine-grained control of the X-ray system properties, nor do they have the bandwidth to manipulate these fine-grained controls during the surgery. Therefore, it is desirable for the system to produce high-quality images at any time the pedal / button is pressed. Quality can be defined in several ways, including spatial resolution, temporal resolution, noise reduction, etc. However, high-quality images may not be required throughout the interventional procedure.

[0003] Consequently, the radiation dose delivered to both the patient and the interventional physician can be higher than is required at times during the surgery. In particular, the interventional physician can have a high radiation dose delivered to their hands over their career, and the patient can be exposed to extended use of fluoroscopy, especially during X-ray-guided minimally invasive surgery; for example, it has been reported that coronary angiography delivers a dose of approximately 5 - 15 millisieverts (mSv), which is equivalent to approximately 2 - 5 years of background radiation. High radiation doses are associated with an increased risk of excessive cancer, and thus it is desirable to reduce radiation exposure to patients and interventional personnel.

[0004] There is a trade-off between radiation dose and image quality. Generally, reducing radiation also degrades image quality in some way. For example, the X-ray system frame rate is proportional to the dose, i.e., reducing the frame rate by updating views less frequently, using fewer X-ray pulses / second degrades the temporal image quality while reducing the delivered radiation dose. In another example, only increasing the voltage of the X-ray source (kV or peak tube voltage kVp) increases the penetration of the X-ray beam, thereby increasing the intensity of the signal measured by the X-ray detector and the radiation dose delivered to the patient's tissue. However, in practice, when the peak tube voltage kVp is increased, the mAs (X-ray tube current * exposure time) is typically decreased to compensate for the increased number of photons reaching the X-ray detector, which results in an overall reduction in the delivered radiation dose.

[0005] In addition, when the peak tube voltage kVp is increased, the amount of scatter also increases due to the increased likelihood of photon interactions, resulting in a decrease in image sharpness. In practice, acquiring higher energy radiographs results in a trade-off between reduced radiation and increased penetration versus increased scatter. Figure 1An image view illustration providing an example of an energy dose tradeoff is shown. The left image 10 was acquired at a higher dose than the right image 12 (i.e., a lower peak tube voltage kVp but a higher mAs (X-ray tube current * exposure time)) and exhibits less scatter, resulting in an image with improved sharpness.

[0006] Other X-ray system properties can also have an impact on radiation dose, such as X-ray beam magnification. By combining X-ray beam magnification with collimation, a less focused beam can be used to image the same field of view, resulting in fewer photons reaching the X-ray detector. This effect can be interpreted as being similar to reducing the spatial resolution of a radiograph image as a tradeoff for reducing the delivered radiation dose. In summary, there are multiple factors that affect image quality, and these factors can be modulated on an X-ray system to vary the delivered radiation dose.

[0007] Currently, the existing techniques for modulating radiation dose in an X-ray system are a set of tools known as "automatic dose rate control" (ADRC) or "automatic exposure control" (AEC). Existing ADRC systems control system parameters such as tube voltage, current, and exposure time in order to reduce radiation to the patient. When a preset radiation level is received at the detector, a basic ADRC system terminates the exposure in order to maintain consistent intensity and signal-to-noise ratio on the radiograph image and ensure that too much radiation dose is not delivered to the patient. More advanced ADRC systems in modern X-ray systems automatically measure signal-to-noise ratio and patient thickness and control X-ray tube parameters as the beam moves. For example, in the Philips Allura TM system, the ADRC estimates the approximate or equivalent water thickness of the patient given the gantry geometry. This estimate is updated based on pixel intensity values and X-ray beam settings / geometry from previous runs. Using the patient thickness as a starting point, the radiation dose is also controlled to maintain consistent detector output and signal-to-noise ratio. Similar concepts also exist in computer tomography (CT) systems in the art; an example is the intelligent mA feature implemented on a CT system that automatically modulates tube current during a CT scan based on patient size and attenuation to maintain a specified noise level in the CT scan image.

[0008] Existing ADRC / AEC techniques disadvantageously focus on patient size and a predetermined noise model to inform the modulation of the dose control system. However, it is desirable to provide intelligent control of the dose based on real-time feedback from the X-ray image itself, especially when it comes to detecting noise caused by physical events such as patient movement, occlusion of the imaging field of view, or even actions from an interventional physician or user indicating that high image quality is not currently required.

[0009] Accordingly, there is a need for an improved method and apparatus for overcoming problems in the art. SUMMARY OF THE INVENTION

[0010] According to one aspect, minimizing the radiation delivered to a patient and an interventionalist during X-ray-guided surgery remains a critical task. An interventional physician (or interventionalist) can receive large doses of radiation over the course of their career, and patients typically receive large doses during fluoroscopy-guided procedures, resulting in an increased risk of malignancy. There is a trade-off between dose and image quality, and reducing one can affect the other. During an intervention, X-rays can be used for long periods of time; however, their use is manually controlled in a binary manner and may not be well optimized. High-quality images are not necessarily required at all times when the X-ray pedal is depressed (i.e., for activating (one or more) X-ray exposures). For example, if the patient is moving, or an occluding object is obscuring the field of view, a high-resolution (spatial, temporal, etc.) view may not be needed. Similarly, if the physician is simply observing or viewing the overall anatomy rather than an object that requires precise visualization, image quality may be less important than dose. According to one aspect, a method is disclosed herein for detecting an opportunity to automatically reduce the radiation output from an X-ray system in exchange for a reduced image quality without interrupting an X-ray image-guided procedure.

[0011] According to one embodiment, a system for modulating radiation dose includes a detector configured to detect at least one noise generation event or user behavior event during an X-ray image-guided procedure using an X-ray imaging system, which indicates that the image quality for X-ray imaging does not need to be higher than a threshold image quality. The system further includes a controller configured to automatically modulate X-ray system parameters that affect image quality and radiation dose based on the detection of at least one of the noise generation event or user behavior event. In one embodiment, the detector includes a noise detection model designed to detect at least one noise generation event or user behavior event during an X-ray image-guided procedure. In another embodiment, the controller is further configured to automatically modulate X-ray system parameters according to a system settings model designed to control X-ray system settings based on the output from the detector.

[0012] According to one embodiment, a system for modulating radiation dose is trained by: receiving X-ray image data containing noise from a noise generating event, inputting the received X-ray image data into a noise detection model, using the noise detection model to generate a predicted noise level in the input image, inputting the predicted noise level into a system setting model, using the system setting model to generate predicted X-ray system parameters, and repeating the input, generation, and adjustment based on (i) a comparison between the predicted noise level and an expected noise level, or (ii) a comparison between a simulated image generated from the predicted X-ray system parameters and an expected X-ray image, or (iii) a combination thereof, until a stop criterion is met.

[0013] According to another embodiment, the noise detection model includes a noise classifier H(x) configured to detect noise levels from different noise sources, where the noise of the noise level refers to any blur source in the current X-ray image of X-ray image-guided surgery. The noise classifier H(x) is trained by generating a training dataset, where known motion, occlusion, user behavior, and other enhancements are introduced when using the X-ray imaging system.

[0014] In yet another embodiment, the controller is further configured to automatically modulate X-ray system parameters according to the system setting model, where the system setting model is designed to control X-ray system settings based on the output of the noise classifier H(x) of the noise detection model according to a set of predicted optimal X-ray system settings, where the image quality is reduced to a visually perceivable threshold level that is indistinguishable or almost indistinguishable from the original image with detected noise. In one embodiment, the system setting model includes at least one selected from the group consisting of: (i) a direct noise level to system setting mapping function that directly maps a set of detected noise levels to the set of predicted optimal X-ray system settings, (ii) a mapping of noise level to system setting based on semi-supervised images, which is trained without directly labeling the ideal X-ray system settings for each input training image or behavior, and (iii) a combination thereof, to provide the set of predicted optimal X-ray system settings.

[0015] In another embodiment, the detector is further configured to predict user behavior indicating the current surgical stage or event of an X-ray-guided surgery performed via the X-ray imaging system based on information inputs from (i) the X-ray imaging system and (ii) the operating room in which the X-ray imaging system is located, where the user behavior does not require an X-ray image with a quality greater than a threshold image quality, and the information inputs include at least one of (a) the current X-ray image, (b) X-ray imaging system user interaction information, and (c) operating room sensor / camera data.

[0016] In yet another embodiment, the noise detection model takes the series of X-ray images as input during a real-time fluoroscopy run when acquiring each image in the series of X-ray images, and considers the series of X-ray images in parallel with each newly acquired X-ray image to better predict changes in behavior, movement, or other temporal variables. In another embodiment, the output of the noise detection model additionally depends on the duration since the occurrence of either at least one noise-induced or user behavior event, wherein the modulation of the X-ray system parameters based on the output of the noise detection model is implemented in real time during the detected at least one noise-induced or user behavior event, and once the corresponding event is completed or no longer occurs and new X-ray images are acquired at a later time, the influence of the corresponding event on the X-ray system settings for the new X-ray image acquisition is lower and decays over time, and the controller restores the modulated X-ray system parameters to the parameters that existed prior to the corresponding modulation.

[0017] According to one embodiment, the system includes at least one of the following: (a) wherein at least one noise-induced or user behavior event includes (i) a blurring or motion noise generation event, or (ii) a user behavior event related to an image quality that requires an image quality less than or equal to a threshold image quality with respect to the current X-ray image of an X-ray-guided procedure, or (b) wherein the X-ray system parameters include at least one selected from the group consisting of X-ray tube voltage / peak tube voltage (kV / kVp), X-ray tube current (mA), exposure time, frame rate, magnification, and any combination thereof. Additionally, the confounding noise generation event can at least include an object that occludes the field of view of the X-ray tube of the X-ray imaging system, wherein the motion noise generation event at least includes movement of the patient or the X-ray tube or the X-ray detector component, and wherein the user behavior event at least includes a framing event.

[0018] According to another embodiment, the X-ray imaging system includes a detector configured to detect at least one noise generation event or user behavior event during an X-ray-guided procedure, which enables the image quality for X-ray imaging to not need to be higher than a threshold image quality. The X-ray imaging system further includes a controller configured to automatically modulate X-ray system parameters that affect image quality and radiation dose based on the detection of at least one of the noise generation event or the user behavior event. In one embodiment, the detector includes a noise detection model designed to detect at least one noise generation event or user behavior event during an X-ray-guided procedure. Additionally, the controller is further configured to automatically modulate the X-ray system parameters according to a system settings model designed to control the X-ray system settings based on the output from the detector.

[0019] In addition, the X-ray imaging system is trained by: receiving X-ray image data containing noise from a noise generation event, inputting the received X-ray image data into a noise detection model, using the noise detection model to generate a predicted noise level in the input image, inputting the predicted noise level into a system setting model, using the system setting model to generate predicted X-ray system parameters, adjusting the parameters of the noise detection model or the system setting model or both based on a comparison between: (i) the predicted noise level and the expected noise level, or (ii) a simulated image generated according to the predicted X-ray system parameters and the expected X-ray image, or (iii) a combination thereof, and repeating the input, generation, and adjustment until a stop criterion is met.

[0020] According to another embodiment, a method for modulating radiation dose includes: detecting at least one noise generation event or user behavior event during X-ray-guided surgery using an X-ray imaging system, the at least one noise generation event or user behavior event rendering the image quality for X-ray imaging not requiring to be higher than a threshold image quality; and automatically modulating X-ray system parameters affecting image quality and radiation dose based on detecting at least one of the noise generation event or the user behavior event. In one embodiment, the method further includes at least one of the following: (a) wherein the detection includes detecting via a noise detection model during X-ray-guided surgery, the noise detection model being designed to detect at least one noise generation event or user behavior event, or (b) wherein automatically modulating the X-ray system parameters includes modulating according to a system setting model, the system setting model being trained to control X-ray system settings based on the output from the detection step.

[0021] According to yet another embodiment, a non-transitory computer-readable medium is encoded with computer program code, the computer program code including a set of instructions executable by a computer to enable the computer to perform the method of modulating radiation dose based on detecting at least one noise generation event or user behavior event during X-ray-guided surgery using an X-ray imaging system.

[0022] After reading and understanding the following detailed description, the advantages and benefits will become apparent to those of ordinary skill in the art. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Embodiments of the present disclosure may take various forms of components and component arrangements as well as various forms of steps and step arrangements. Accordingly, the drawings are for the purpose of illustrating various embodiments and should not be construed as limiting the embodiments. In the drawings, like reference numerals refer to like elements. Additionally, it should be noted that the drawings may not be drawn to scale.

[0024] Figure 1An image view illustration of an example of energy dose compromise;

[0025] Figure 2 An image view illustration of an example of motion blur affecting image sharpness according to an embodiment of the present disclosure;

[0026] Figure 3 An image view illustration of an example of the effect of changing X-ray tube parameters to reduce radiation dose in the presence of motion blur according to an embodiment of the present disclosure;

[0027] Figure 4 A block diagram view of a noise detection model including a noise classifier H(x) according to an embodiment of the present disclosure;

[0028] Figure 5 A block diagram view of a model for generating a new image in the case of a change in a given X-ray imaging system setup according to an embodiment of the present disclosure;

[0029] Figure 6 A block diagram view of a generative adversarial semi-supervised training scheme for unlabeled data according to an embodiment of the present disclosure;

[0030] Figure 7 A block diagram view of an X-ray imaging system configured to modulate radiation dose according to an embodiment of the present disclosure; and

[0031] Figure 8 A flowchart of a method for modulating radiation dose according to an embodiment of the present disclosure. Detailed Description

[0032] The embodiments of the present disclosure and their various features and advantageous details are more fully explained with reference to the non-limiting examples described and / or illustrated in the accompanying drawings and detailed in the following description. It should be noted that the features shown in the drawings are not necessarily drawn to scale, and the features of one embodiment may be used with other embodiments that will be recognized by those skilled in the art, even if not explicitly stated herein. Descriptions of well-known components and processing techniques may be omitted so as not to unnecessarily obscure the embodiments of the present disclosure. The examples used herein are only intended to facilitate an understanding of the manner in which the embodiments of the present invention may be practiced and to further enable those skilled in the art to practice the present invention. Therefore, the examples herein should not be construed as limiting the scope of the embodiments of the present disclosure, which is defined only by the appended claims and applicable law.

[0033] It should be understood that the embodiments of the present disclosure are not limited to the specific methods, protocols, devices, apparatuses, materials, applications, etc. described herein, as these may vary. It should also be understood that the terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of the claimed embodiments. It must be noted that, as used herein and in the appended claims, the singular forms "a", "an", and "the" include plural references unless the context clearly dictates otherwise.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of the present disclosure belong. Preferred methods, apparatuses, and materials are described, but any methods and materials similar or equivalent to those described herein may be used in the practice or testing of the embodiments.

[0035] According to one embodiment, a system is provided for detecting torque during interventional X-ray guided surgery when high image quality is not required (such as when noise or other movements occlude the field of view of an X-ray detector), e.g., detecting noise generated by physical events such as patient movement, occlusion of the imaging field of view, or even user actions indicating that high image quality is not currently required. For example, if a sudden or rapid movement occurs (i.e., if the patient moves or the C-arm moves), the human eye cannot resolve small details in the radiograph or X-ray image, regardless of the image quality. In a non-medical example, when a dog jumps, individual hairs on the dog's body cannot be seen due to the movement, and thus a gigapixel image is not required. There are similar situations in medical imaging where, given a chaotic scene in terms of time or space, the user's eyes cannot resolve small details of the corresponding medical image on the screen of a medical imaging system anyway.

[0036] In an X-ray imaging system according to an embodiment of the present disclosure, a neural network is trained to detect blur or motion in a radiograph or X-ray image, and to determine whether the current image quality of the corresponding radiograph or X-ray image is appropriate based on detecting one or more blur / noise generating events that may occur during X-ray guided fluoroscopic interventional surgery. The complex mapping between the X-ray system parameters, the resulting image quality, and the blur level of the last X-ray image currently detected can be used to identify the optimal X-ray system settings in an event where the radiation dose can be reduced without significantly affecting the ability to see objects in the updated X-ray image with the new X-ray system settings.

[0037] Now turning to Figure 2 , an image view illustration of an example of motion blur affecting image sharpness according to an embodiment of the present disclosure is shown. Figure 2 The examples in show an image 14 without motion artifacts and an image 16 with motion artifacts.Figure 2 Further illustrate what the human eye will see in the presence of sudden or rapid motion. In fact, the image viewing screen of an X-ray imaging system may still display high-resolution and high-quality images, but the human eye may not be able to fully benefit from this image sharpness.

[0038] Now refer to Figure 3 , an image view illustration of several examples showing the effect of changing X-ray tube parameters to reduce radiation dose in the presence of motion blur according to an embodiment of the present disclosure is shown. Comparing the left column 18 and the center column 20 respectively, the effect of simulating the same amount of scatter in the normal image and the motion blur image (center column 20, the blur simulated due to scatter) shows the difference in image sharpness loss in the presence of noise. The difference between the motion blur image with simulated scatter 24 and the motion blur image without simulated scatter 26 is hardly noticeable, while the difference in the normal image 28 is clearly noticeable. This difference between the normal image 28 and the motion blur image 26 indicates that there may be some excess in the radiation dose / image quality that can be exploited in the presence of blurred noise. In Figure 3 In the third column 22 of

[0039] , the image represents a "simulated magnification" image, which refers to the simulated magnification of the image in the first column 18. The effect in the images of column 22 is simply a reduction in image resolution. This occurs when using digital zoom, although the field of view (FOV) will not remain the same as shown in the image. The images in column 22 are shown to provide a comparison between the same images in the first column 18. Alternatively, the third column can be referred to as "resolution reduction" rather than "simulated magnification" for more clarity.

[0040] According to a first aspect, the method and system include a model or algorithm trained to detect blur or user behavior (e.g., framing) in an X-ray image that does not require high image quality. According to a second aspect, the system includes a controller that operates in response to the detection of the model or algorithm by automatically modulating X-ray system parameters so as to reduce image quality to the highest level that can be reasonably resolved by the human eye given the detection, which will reduce radiation dose as a side effect. The X-ray system parameters include one or more of the following: X-ray tube voltage (kV), peak X-ray tube voltage (kVp), X-ray tube current mA, exposure time, frame rate, magnification, etc.

[0041] Reference now Figure 4 , shows a block diagram view of a noise detection model 30 including a noise classifier H(x) 32 according to an embodiment of the present disclosure. In one embodiment, to detect blurry noise or behavior, the noise detection model 30 includes a neural network that is trained to receive input 34 including an X-ray image, system parameters, user input, and information from an operating room where the X-ray imaging system is located. Note that here, "noise" includes any blur source in the X-ray image as well as signal or measurement noise.

[0042] In one example, from a stream of acquired X-ray images, a neural network can extract image-based features associated with relevant sources of image blur, such as detecting blur through motion blur. For example, a neural network can be trained to detect a lack of sharpness in the edges of objects in a given X-ray image (potentially indicating motion blur), or to refine a series of X-ray images in a recurrent neural network or transformer-like architecture to detect significant changes from frame to frame that would indicate rapid motion. Computer vision methods such as optical flow can also be substituted to detect motion.

[0043] Additionally or alternatively, live X-ray system settings / parameters may also be refined to detect blur in the image. For example, as previously described, a combination of blur / motion detected in the image (such as rapid movement of the patient table or C-arm detected according to system parameters or longer exposure times detected according to system settings) may indicate a high confidence that blur is specifically caused by motion. The speed of motion may also be inferred from system parameters (such as patient table movement or C-arm angle), and thus the amount of noise estimated due to motion may increase as the speed of motion increases. The output of the model or neural network may be a score (e.g., a scalar score between 0 and 1) indicating the magnitude of motion corresponding to a single source of motion, such as C-arm rotation, or the joint sum of all sources of motion.

[0044] In another example, a neural network can be trained to detect blurring by the presence of unexpected objects in a frame, such as a table in the X-ray beam path, an intravenous line (IV), an ECG lead, or a physician's hand. The object detection neural network can be trained to identify unexpected objects in the received X-ray image and produce an estimate of how much of the anatomical structure of interest has been occluded. Similarly, information extracted from the operating room can be used to detect physical objects blocking the X-ray beam. For example, an algorithm can use images from cameras in the operating room to detect the X-ray source and detector positions and angles, and estimate whether any objects are in the path of the X-ray beam. In another example, the C-arm configuration can be extracted from the system settings and the known geometry of the room and compared to objects detected from a spatially calibrated camera system to detect an intersection of the X-ray beam with an unexpected object.

[0045] In yet another example, a neural network can be configured to distill a combination of system settings / parameters, user input to the system, and information extracted from the operating room to detect user behavior indicating that high image fidelity is not required. For example, one instance when user behavior can indicate that high image quality is not needed is when the user is attempting to adjust the X-ray imaging system to find the best view / position for a particular step in a surgical procedure. Typically during a framing activity, high-resolution image details are not required, and only a macroscopic view of the imaging target is needed to select the desired view. The neural network can be trained to detect specific interactions with the X-ray imaging system that indicate a high likelihood of a framing behavior. For example, rapid translation of the C-arm / patient table, back-and-forth changes in the C-arm angle, or rapid changes in zooming in or out can indicate that the user is busy with a framing behavior. Similarly, information extracted from the operating room can enhance the input to the neural network. For example, tracking the user's eye gaze can indicate whether the user is focused on a specific part of the anatomical structure, in which case high image quality may be required, or if they are rapidly scanning multiple different parts of the image while changing system parameters, in which case they may be busy with framing.

[0046] In yet another example, a neural network can be configured to distill user interactions with the X-ray imaging system and information extracted from the operating room to detect user inattentiveness. For example, a user may be pressing the X-ray pedal but interacting with menu options on the system's user interface, or their eye gaze may be completely away from the screen, indicating that the X-ray image being generated is not very important. The neural network can be trained to interpret a combination of inattentive eye gaze and active user interaction with non-imaging aspects of the system (such as scrolling through the system menu) as an indication of a high likelihood of inattentiveness to the display screen.

[0047] Other noise sources that can be detected from the above input combinations, in addition to motion or physical occlusion or user actions, can include poor detector signal-to-noise ratio, overexposure in regions of the X-ray image, a high degree of scatter noise in the X-ray image, excessive blurring from non-motion sources (e.g., like focus blurring), overuse of digital zoom, an imaging target not centered in the field of view, etc. Data inputs from different sources (such as the image stream, X-ray imaging system settings, etc.) can have an interaction that provides more robust noise detection.

[0048] As will be further discussed herein, the noise detection model or neural network 30 output will appropriately change a new set of system parameters for image quality and radiation dose. The neural network can be trained in multiple parts. Figure 4 The illustration of the noise classifier H(x) in includes several boxes representing different downstream layers in the convolutional encoder, which are further part of the neural network. The number / size of the boxes can vary according to the network architecture of a given neural network. Neural networks are known in the art and thus will not be described in further detail herein.

[0049] First, the classification network 32 detects the noise levels from different noise sources, as Figure 4 shown. The neural network can predict the probability of the presence of noise for each of a set of known noise types. For example, these noise types can include physical occlusion of the imaging target, motion, user inattentiveness, user framing behavior, image overexposure, blurring from other sources, etc. The noise classification network H(x) 32 can be trained by generating a training data set, where known motion, occlusion, user actions, and other augmentations are introduced when using the X-ray imaging system. In one example, to generate the training data set, X-ray images can be recorded while the C-arm rotates at different speeds to indicate different simple motion noise levels, where no C-arm rotation indicates a 0% motion noise level and the maximum C-arm rotation speed indicates a 100% motion noise level. Of course, the network can be jointly trained on multiple different motion sources (including patient motion, patient table motion, etc.), in which case a similar score can be generated based on the joint combination of all inputs as a marker indicating the amplitude of motion. At inference time, the neural network will receive a series of X-ray images and estimate the probability of motion noise in the new input based on the trained model. Although the training data set will include various inputs such as X-ray images, system settings, user interactions, and information from the room, in the example of noise caused by excessive motion, the neural network trained to detect this type of noise will reveal the relationship between the fast movement in subsequent X-ray images and the system parameters indicating the movement of the C-arm in order to detect motion.

[0050] In another example, to detect physical occlusions, a training dataset can be generated by recording an X-ray image and an operating room camera feed image of an object that blocks the X-ray beam path from the source to the detector. In cases where physical occlusions in the image are scored, similar labels can be generated by manually scoring the degradation of the image in the target region or quantifying the physical area of the target region blocked by the intruding object (e.g., 0 to 100%, where 100% indicates that the region of interest is completely blocked by an unexpected object). In Figure 4 one example, an output 36 of a noise level detected by a noise classifier H(x) can include motion (e.g., motion = 0.3), occlusion (e.g., occlusion = 0.1), framing (e.g., framing = 0.6), etc.

[0051] Regarding the conversion of system settings detected with respect to noise, embodiments of the present disclosure provide several methods to train an algorithm or model to control the system settings of an X-ray imaging system based on the output from a noise classification network or a noise classifier H(x) 32. The methods can include one or more of the following: a first method of direct mapping of noise level to system settings; and, a second method of mapping of noise level to system settings based on semi-supervised images. In another embodiment, a fully supervised training dataset can be generated to train the noise classifier.

[0052] For the method of direct mapping of noise level to system settings according to one embodiment, a function can be provided that directly maps a set of detected noise levels to a set of optimal system settings. For example, a set of probabilities / levels corresponding to a set of known noise types is generated by the above-described noise classifier H(x). The set of noise levels is fed as input into a separate model that functions to convert the noise levels into changes in the X-ray imaging system settings that produce an image with the minimum possible radiation while imperceptibly degrading the image quality in the presence of the detected noise. As an example, the separate model can be a simple pre-determined mapping of one noise type to one system setting. In particular, the model can be designed such that detecting motion blur will be directly mapped to a reduction in the peak tube voltage (kVp).

[0053] In another, more complex example, a multi-variable function maps multiple noise levels to multiple X-ray imaging system outputs or settings. Fitting this multi-variable function will require a training dataset where the best system settings corresponding to a particular noise level are annotated. The term "annotated" as used here refers to some manual tagging of the training dataset by an expert. In other words, to combine multiple noise factors, the expert can be shown examples of noise-corrupted images and can manually vary the X-ray system parameters in the direction of reducing radiation until the image begins to perceptibly degrade. The maximum amount of change in the parameters before the image begins to degrade significantly will be the ground truth label or annotation. Depending on how the problem is formulated, either a set of X-ray imaging system settings or the amount by which the X-ray system settings are changed is valid as the annotation.

[0054] In yet another example, the least squares method can be used to fit a polynomial function to the training data. To generate a training dataset to fit such a model in a supervised manner, first a set of "output" imaging system settings are identified that have an impact on radiation dose and image quality. Then, in one example, a dataset can be generated where a known noise level is applied to the image (i.e., by manually introducing a known amount of motion noise or a known amount of physical occlusion, etc.), and a set of "output" imaging system settings are modified to reduce radiation until the image quality begins to perceptibly degrade in the presence of the noise. Then, the system settings with the lowest radiation dose before the image begins to perceptibly degrade will serve as the dataset output, while the noise level serves as the dataset input. These inputs and outputs can be used to fit a simple model such as a polynomial function or a more complex model such as a neural network.

[0055] For example, in the simplest embodiment where patient table motion is the only noise source and kVp / mAs are the only system settings to be modulated, a dataset can be generated where the patient table is translated in varying directions and speeds. For a particular resulting series of images, the kVp and mAs settings can be modulated so as to reduce the radiation dose at the expense of increasing the scatter (and thus blurring) of the image. Due to the motion inherent in the dataset, the increased blurring due to the change in system settings may be imperceptible at a certain level. This level can be determined experimentally and used as the ground truth label to train the model to produce the optimal system settings (or the change in system settings).

[0056] Similarly, to generate a dataset for training a model to translate a noise level associated with user behavior into an "output" system setting, a user can be instructed to perform tasks such as framing while operating an X-ray imaging system. The framing operation can be performed multiple times under different "output" system settings until the user identifies the system setting with the lowest radiation dose without a significant degradation in image quality. Similar data collection can be performed for behaviors such as inattentiveness to the screen. The framing and inattentiveness can be predicted as binary labels or non-binary levels by a noise classification network. As an example, to generate a dataset of non-binary user behaviors, the user can be asked or influenced to give different levels of attention to the screen.

[0057] For a method of semi-supervised image-based mapping of noise levels to system settings according to another embodiment, training an algorithm or model in a semi-supervised manner is done without directly labeling the ideal system setting for each input training image or behavior. For example, a second algorithm can be trained to predict the optimal system setting that reduces the radiation dose while not significantly degrading the X-ray image (to be acquired) beyond the degradation caused by the detected noise. To train this algorithm, an intermediate simulation model is employed that describes how changes in system settings will affect the X-ray image to be acquired, referred to herein as the "generator".

[0058] The generator model takes as input the current system setting, the current X-ray image, and a new set of system settings, and produces as output a new image that approximates the expected image given the new system settings. Note that the generator model has been trained or fit, so its parameters are not modified during this process. The generator model can take the form of a physics-based model or a pre-trained neural network. For example, the generator model can take as input the current X-ray image and settings and any new set of system settings where the kVp is increased and the mAs is decreased (resulting in lower dose but higher scatter), and the generator model will output a new version of the X-ray image where the scatter is artificially increased due to modeling the effects of changing the X-ray system settings. The resulting image simulated by the generator is then compared to the original input image (or series of images) by a discriminator model whose role is to score the perceptibility of the change in image quality. When the change in the currently selected system setting results in a significantly and perceptibly different image, the discriminator will give a poor score to the output (indicating that the user will be able to perceive a loss in image quality), and vice versa. Iterative training is performed until the system setting model is trained to produce an ideal set of system settings that rewards a reduction in radiation but penalizes a perceptible loss in image quality. The purpose of this intermediate model is to enable training a model to identify the optimal changes in system settings without a large manually labeled dataset of noise level and corresponding ideal system setting pairs.

[0059] Referring now to Figure 5 , a block diagram view of a model 38 that generates a new image 42 in the presence of changes in a given system setup in accordance with an embodiment of the present disclosure is shown. In essence, the model G(x) 38 facilitates quantification of the resulting image quality loss / increase based on changes in given system parameters by simulating the resulting images, in order to assist in selecting the magnitude of change for each parameter. The model 38 receives an input image 40 and outputs an updated or expected image 42 in the new system setup. Figure 5 The illustration of the model G(x) in

[0060] Referring now to Figure 6 , a block diagram view of a generative adversarial semi-supervised training scheme for unlabeled data in accordance with an embodiment of the present disclosure is shown. Given the model G(x) 38 with fixed parameters after training, an additional module that includes a model 44 (i.e., the system setup model) for mapping a detected noise level from H(x) 32 to a system setup is trained in an adversarial manner without including labeled training data that contains the annotated optimal system setup. The system setup model parameters are optimized by feeding or inputting the predicted optimal system setup 46 into the image simulator G(x) 38, and the expected image output 42 is compared with the original image 34 by the discriminator D 48.

[0061] In an ideal scenario, the system setup 46 can be changed to reduce the radiation dose such that the resulting image 42 is indistinguishable from the original image 34. Adversarial training is an example of an ideal training for this optimization. Here, the loss function to be optimized will be a combination of the discriminator's ability to distinguish between the two images, the reduction (or increase) in the resulting radiation dose, and the change in the resulting noise level (which can be calculated using the noise classifier H(x) 32). This workflow is shown in Figure 6 .

[0062] The loss function to be minimized is:

[0063] loss 生成器 = E 生成 [log(1 - D(G(x)))] + δ 噪声 + δ 剂量 ,

[0064] loss 鉴别器 = -E I [logD(I)] - E 生成 [log1 - D(G(x))],

[0065] where I is a set of original images, x is a set of system settings predicted from the output of the noise classifier H(x), G(x) is a set of modified images (i.e., expected images), D is the discriminator output, and δ 噪声 and δ 剂量 are changes in the noise level and radiation dose.

[0066] Regarding semi-supervised training, in one embodiment, automatically modulating X-ray system parameters further includes predicting an optimal X-ray system setting that reduces the radiation dose while not significantly degrading subsequent acquired X-ray images beyond the degradation caused by the detected noise, where predicting the optimal X-ray system setting includes using (i) the image simulator model G(x) 38 and (ii) the discriminator D 48. In one embodiment, the image simulator model G(x) 38 is trained to quantify the loss or increase in the resulting image quality based on a given change in X-ray system parameters to assist in selecting the magnitude of each parameter change. In another embodiment, the image simulator model G(x) 38 describes how changes in X-ray system settings will affect the current X-ray image, where the image simulator model G(x) receives (i) the current X-ray system settings, (ii) the current X-ray image 34, and (iii) a new set of system settings 46 as inputs and produces a new image approximating the expected image as output 42 given a change in X-ray system settings. Additionally, in another embodiment, the image simulator model G(x) includes a physics-based model or a neural network.

[0067] According to another embodiment, the system and method further optimize the predicted optimal X-ray system setting output from the system setting model by: (i) simulating the expected X-ray image output 42 via the image simulator model G(x) 38 based on the predicted optimal X-ray system setting 46, (ii) comparing (ii)(a) the expected X-ray image output 42 with (ii)(b) the original X-ray image 34 corresponding to the current X-ray image via the discriminator D 48, and (iii) changing the predicted optimal X-ray system setting to a new X-ray system setting that reduces the radiation dose such that the resulting X-ray image is substantially indistinguishable from the original X-ray image.

[0068] In another embodiment, information from the X-ray imaging system and from the operating room (i.e., where the X-ray imaging system is located), such as room sensor / camera data and system interaction information, is used independently or in combination with the first embodiment to predict user behavior. Certain types of user behavior can indicate that the current surgical stage or event does not require high-quality images. For example, if the physician exhibits behavior indicating that he or she is only looking for the correct view of the gross anatomy without viewing precise details such as small vasculature or devices, then this behavior can indicate to the algorithm that a system setting with a lower radiation output would be appropriate. Depending on the situation, these additional inputs can be passed along with the modified images in the network architecture to Figure 4 the noise classification network H(x)32 in

[0069] In another embodiment, Figure 4 the noise classification network H(x)32 is a recurrent neural network that takes as input a series of images during a live fluoroscopy run (when it (i.e., the series of individual images) occurs), and updates an internal hidden state with each new image in order to better predict changes in behavior, motion, or other temporal variables.

[0070] In yet another embodiment, the output of the noise classification network can depend on the time since a blurring event occurred. For example, when the user performs isocenter positioning using live fluoroscopy (i.e., moving the C-arm and / or the table to center the region of interest), rapid movement in the acquired fluoroscopy images should indicate lower image quality, and thus a lower dose is acceptable. Therefore, the proposed network changes can be made in real time during isocenter. However, once isocenter is complete and new fluoroscopy images are acquired at a later time, the impact of the blurring event on the X-ray imaging system settings for the new image acquisition is lower. That is, the impact of the blurring event on the X-ray imaging system settings for the images can decay over time. In other words, the output of the noise detection neural network can additionally depend on the duration since any one of one or more noise generation events and / or user behavior events occurred. Modulation of the X-ray system parameters is implemented in real time based on the output of the noise detection neural network during one or more detected noise generation events and / or user behavior events. Once the corresponding event is complete or no longer occurs and new X-ray images are acquired at a later time, the impact of the corresponding event on the X-ray system settings for the new X-ray image acquisition is lower and decays over time, where the controller 74 (as will be discussed herein with respect to Figure 7 restores the modulated X-ray system parameters to the parameters that existed prior to the corresponding modulation.

[0071] Now referring to Figure 7, showing a block diagram view of an X-ray imaging system 50 configured to modulate radiation dose according to an embodiment of the present disclosure. Although Figure 7 an imager 50 of the C-arm type is shown, it should be understood that other imager configurations may also be used. The imager 50 includes a rigid C-arm 52 having a detector 54 fixed at one end thereof and an X-ray source (e.g., an X-ray tube) 56 and a collimator 58 (collectively referred to hereinafter as the CX assembly) fixed at the other end. The X-ray source 56 operates to generate and emit a primary radiation X-ray beam p, the main direction of which is schematically indicated by the vector p. The collimator 58 operates to collimate the X-ray beam with respect to a region of interest ROI within a patient 60 located on an examination table 62.

[0072] The position of the C-arm 52 is adjustable such that projection images can be acquired along different projection directions p. The C-arm 52 is rotatably mounted around the examination table 62. The C-arm 52 and the CX assembly therewith are driven by a stepper motor or other suitable actuator (not shown).

[0073] The overall operation of the imager 50 is controlled by an operator from a computer console 64. The console 64 is coupled to a screen 66. The operator can control the acquisition of any image by releasing an individual X-ray exposure via the console 64, for example by actuating a joystick or pedal or other suitable input device (not shown) coupled to the console 64. During intervention and imaging, the examination table 62 (and the patient 60) is positioned between the detector 54 and the X-ray source 56 such that the region of interest ROI is irradiated by the radiation beam.

[0074] Broadly, during image acquisition, the collimated X-ray beam is emitted from the X-ray source 56, passes through the patient 60 at the region ROI, undergoes attenuation by interaction with the substances therein, and the thus attenuated beam then impinges on the surface of the detector 54 at a plurality of detector units. Each unit struck by an individual ray (of the primary beam) responds by emitting a corresponding electrical signal. The set of such signals is then converted by a data acquisition system (“DAS”-not shown) into corresponding digital values representative of the attenuation. The density of the organic material constituting the ROI determines the level of attenuation. High-density materials (such as bone) result in higher attenuation than less dense materials (such as vascular tissue). Then the set of such registered digital values for each X-ray is combined into an array of digital values, thereby forming an X-ray projection image for a given acquisition time and projection direction.

[0075] During a given image-guided interventional procedure, the imager 50 is operated to acquire a sequence of "live" fluoroscopic X-ray projection images 68 ("fluoroscopy") or angiograms ("angiography") during the given interventional procedure. Additionally, during the interventional procedure, the user can initiate or specify various procedure steps via keystroke input or via a graphical user interface (GUI) widget 70 such as a drop-down menu or other graphical input arrangement or otherwise. For example, the GUI widget 70 can appear as an overlay on the currently acquired X-ray image 72 of the image sequence 68.

[0076] Still referring to Figure 7 , the system 50 also includes a controller 74 located within the console 64. In one embodiment, the controller 74 includes one or more of the following: a microprocessor, a microcontroller, a field-programmable gate array (FPGA), an integrated circuit, discrete analog or digital circuit components, hardware, software, firmware, or any combination thereof, for further performing the various functions discussed herein in accordance with the requirements of a given interventional X-ray imaging system implementation and / or application for modulating radiation dose as discussed in this disclosure.

[0077] According to one or more embodiments of the present disclosure, the controller 74 can also include one or more of various modules, which are configured to implement one or more corresponding steps of the method for modulating radiation dose discussed herein. For example, one module can include a room sensor / camera module, which is configured to capture inputs from room sensors, cameras, or both (collectively indicated by reference numeral 76 in Figure 7 ). Such room sensor or camera inputs will be captured for use in one or more embodiments of the method for modulating radiation dose discussed herein.

[0078] It should be understood that one or more of the various modules can be computer program modules presented in a non-transitory computer-readable medium. In one embodiment, a non-transitory computer-readable medium is encoded with computer program code, the computer program code including a set of instructions that are executable by a computer to enable the computer to perform the method for modulating radiation dose based on detecting at least one noise generation event or user behavior event during an X-ray-guided procedure using an X-ray imaging system, as discussed herein with respect to the various embodiments of the present disclosure.

[0079] Now referring to Figure 8, which shows a flowchart view of a method 80 for modulating radiation dose according to an embodiment of the present disclosure. The method for modulating radiation dose includes: detecting (step 82) at least one noise generation event or user behavior event during X-ray guided surgery using an X-ray imaging system, which enables the image quality for X-ray imaging not to need to be higher than a threshold image quality; and automatically modulating (step 84) X-ray imaging system parameters that affect image quality and radiation dose based on the detection of at least one of the noise generation event or user behavior event. In one embodiment, the claimed method further includes at least one of the following: (a) wherein the detection includes detecting during X-ray guided surgery via a noise detection model designed to detect at least one noise generation event or user behavior event, or (b) wherein automatically modulating the X-ray system parameters includes modulating according to a system settings model, the system settings model being trained to control X-ray system settings based on the output from the detection step. In an exemplary embodiment, the noise detection model and the system settings model are combined into a single model.

[0080] Although only a few exemplary embodiments have been described in detail above, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without substantially departing from the novel teachings and advantages of the embodiments of the present disclosure. Accordingly, all such modifications are intended to be included within the scope of the embodiments of the present disclosure as defined in the appended claims. After reading the specification, combinations of the above embodiments and other embodiments not specifically described herein will be apparent to those skilled in the art. In the claims, the apparatus-plus-function clauses are intended to cover the structures described herein as performing the recited function and not only cover structural equivalents but also equivalent structures.

[0081] In addition, any reference numerals placed in parentheses in one or more claims should not be construed as limiting the claims. Words such as "comprising" and "including" etc. do not exclude the presence of elements or steps other than those listed as a whole in any claim or specification. The singular reference to an element does not exclude the plural reference to such elements and vice versa. One or more embodiments can be implemented by hardware including several different elements and / or by a suitably programmed computer. In a device claim enumerating several devices, several of these devices can be embodied by one and the same item of hardware. 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 advantageously.

Claims

1. A system (50) for modulating radiation dose, the system comprising: A detector (32) configured to detect at least one noise generation event or user behavior event during X-ray guided surgery using an X-ray imaging system (50), the at least one noise generation event or user behavior event causing the image quality for X-ray imaging not to need to be higher than a threshold image quality; And A controller (74) configured to automatically modulate X-ray system parameters that affect image quality and radiation dose based on the detection of the at least one of the noise generation event or user behavior event.

2. The system (50) according to claim 1, wherein, The detector (32) includes a noise detection model (30) designed to detect the at least one noise generation event or user behavior event during the X-ray guided surgery.

3. The system (50) according to claim 1, wherein, The controller (74) is further configured to automatically modulate the X-ray system parameters according to a system setting model (44), the system setting model (44) being designed to control X-ray system settings based on an output from the detector (32).

4. The system (50) according to claim 3, wherein, The detector (32) includes a noise detection model designed to detect the at least one noise generation event or user behavior event during the X-ray guided surgery.

5. The system (50) according to claim 4, wherein, The system for modulating radiation dose is trained by: Receiving X-ray image data containing noise from a noise generation event, Inputting the received X-ray image data into the noise detection model, Using the noise detection model to generate a predicted noise level in the input image, Inputting the predicted noise level into the system setting model, Using the system setting model to generate predicted X-ray system parameters, Adjusting the parameters of the noise detection model or the system setting model or both based on a comparison between: (i) The predicted noise level and an expected noise level, or (ii) A simulated image generated according to the predicted X-ray system parameters and an expected X-ray image, or (iii) A combination thereof, and Repeating the input, the generation, and the adjustment until a stop criterion is met.

6. The system (50) according to claim 2, wherein, The noise detection model (32) includes a noise classifier H(x) configured to detect noise levels from different noise sources, wherein the noise of the noise level refers to any blurring source in the current X-ray image of the X-ray guided surgery.

7. The system (50) according to claim 6, wherein, The noise classifier H(x) is trained by generating a training data set, and known motions, occlusions, user behaviors, and other enhancements are introduced into the training data set when using the X-ray imaging system.

8. The system (50) according to claim 6, wherein, The controller (74) is further configured to automatically modulate the X-ray system parameters according to a system settings model (44), wherein the system settings model is designed to control the X-ray system settings based on a set of predicted optimal X-ray system settings according to the output of the noise classifier H(x) from the noise detection model (32), wherein the image quality is reduced to a visually perceptible threshold level indistinguishable or almost indistinguishable from the original image with detected noise.

9. The system (50) according to claim 8, wherein, The system settings model (44) includes at least one selected from the group consisting of: (i) a direct noise level to system settings mapping function that directly maps a set of detected noise levels to the set of predicted optimal X-ray system settings, (ii) a semi-supervised image-based mapping of noise level to system settings that is trained without directly labeling the ideal X-ray system settings for each input training image or behavior, and (iii) a combination thereof to provide the set of predicted optimal X-ray system settings.

10. The system (50) according to claim 1, wherein, The detector (32) is further configured to predict user behavior for X-ray images that do not need to have a quality greater than a threshold image quality based on information inputs from (i) the X-ray imaging system and (ii) the operating room in which the X-ray imaging system is located, the user behavior indicating the current surgical stage or event of the X-ray-guided surgery performed via the X-ray imaging system, the information inputs including at least one of the following: (a) the current X-ray image; (b) X-ray imaging system user interaction information; and, (c) operating room sensor / camera data.

11. The system (50) according to claim 2, wherein, The noise detection model (32) takes the series of X-ray images as input during a live fluoroscopy run when acquiring individual images in the series of X-ray images, and considers the series of X-ray images in parallel with each newly acquired X-ray image to better predict changes in behavior, movement, or other temporal variables.

12. The system (50) according to claim 2, wherein, The output of the noise detection model (32) additionally depends on the duration since any one of the at least one noise generation event or user behavior event occurred, wherein the modulation of the X-ray system parameters based on the output of the noise detection model is implemented in real time during the detected at least one noise generation event or user behavior event, and once the corresponding event is completed or no longer occurs and new X-ray images are acquired at a later time, the influence of the corresponding event on the X-ray system settings for the new X-ray image acquisition is lower and decays over time, wherein the controller restores the modulated X-ray system parameters back to the parameters that existed prior to the corresponding modulation.

13. The system (50) according to claim 1, further comprising at least one of the following: (a) wherein, The at least one noise generation event or user behavior event includes (i) a blur or motion noise generation event, or (ii) a user behavior event that requires an image quality less than or equal to the threshold image quality with respect to the current X-ray image of the X-ray guided surgery, or (b) wherein the X-ray system parameters include at least one selected from the group consisting of: kV / kVp, mA, exposure time, frame rate, magnification, and any combination thereof.

14. The system (50) according to claim 13, wherein, The blur noise generation event at least includes an object that occludes the field of view of the X-ray tube of the X-ray imaging system, wherein the motion noise generation event at least includes movement of the patient or the X-ray tube or the X-ray detector component, and wherein the user behavior event at least includes a framing event.

15. An X-ray imaging system (50) comprising: A detector (32) configured to detect at least one noise generation event or user behavior event during an X-ray guided surgery, the at least one noise generation event or user behavior event causing the image quality for X-ray imaging not to need to be higher than a threshold image quality; And A controller (74) configured to automatically modulate X-ray system parameters that affect image quality and radiation dose based on the detection of the at least one of the noise generation event or the user behavior event.

16. The X-ray imaging system (50) according to claim 15, wherein, The detector (32) includes a noise detection model designed to detect the at least one noise generation event or user behavior event during the X-ray guided surgery, and wherein the controller (74) is further configured to automatically modulate the X-ray system parameters according to a system setting model designed to control X-ray system settings based on the output from the detector.

17. The X-ray imaging system (50) according to claim 16, wherein, The X-ray imaging system is trained by: Receiving X-ray image data containing noise from a noise generation event, Inputting the received X-ray image data into the noise detection model, Using the noise detection model to generate a predicted noise level in the input image, Inputting the predicted noise level into the system setting model, Using the system setting model to generate predicted X-ray system parameters, Adjusting the parameters of the noise detection model or the system setting model or both based on a comparison between: (i) the predicted noise level and the expected noise level, or (ii) a simulated image generated according to the predicted X-ray system parameters and the expected X-ray image, or (iii) a combination thereof, and Repeating the input, the generation, and the adjustment until a stop criterion is met.

18. A method for modulating radiation dose, the method comprising: Detecting at least one noise generation event or user behavior event during an X-ray guided surgery using an X-ray imaging system, the at least one noise generation event or user behavior event causing the image quality for X-ray imaging not to need to be higher than a threshold image quality; And Automatically modulate X-ray system parameters that affect image quality and radiation dose based on the detection of the at least one of the noise generation event or user behavior event.

19. The method according to claim 18, further comprising at least one of the following: (a) wherein the detection comprises detection via a noise detection model during the X-ray guided surgery, the noise detection model being designed to detect the at least one noise generation event or user behavior event, or (b) wherein, Automatically modulating the X-ray system parameters includes modulating according to a system setting model, which is trained to control X-ray system settings based on the output from the detection step.

20. A non-transitory computer-readable medium encoded with computer program code, the computer program code including a set of instructions that can be executed by a computer to enable the computer to perform the method according to claim 18 to modulate radiation dose based on detecting at least one noise generation event or user behavior event during an X-ray guidance procedure performed with an X-ray imaging system.

Images