Lung volume gated X-ray imaging system and method

Abstract

To provide a geometrically-resolved X-ray imaging system (GREX).SOLUTION: A method of imaging a patient's lung comprises the following steps of: positioning the patient at a first orientation relative to an X-ray imaging apparatus; obtaining a volumetric measurement of the patient's breathing; while the patient is positioned at the first orientation relative to the X-ray imaging apparatus, and while obtaining the volumetric measurement of the patient's breathing, determining, based on the volumetric measurement of the patient's breathing, a breathing phase of the patient; and, upon determination that the breathing phase of the patient matches a predefined breathing phase, gating the X-ray imaging apparatus to produce an X-ray projection image of the patient's lung.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Some embodiments of the present disclosure relate to medical imaging methods, and more specifically, systems and methods for performing geometrically resolved X-ray imaging methods. Some embodiments of the present disclosure relate to radiation therapy.

Background Art

[0002] By computational modeling of the human body, it has become possible to easily understand anatomical behaviors representing various physiological states. Physicians can visualize anatomical behaviors with state-of-the-art imaging techniques, but the latest techniques that can accurately image the complex movements of the heart and lungs are often too expensive to adopt. Furthermore, heartbeats, for example, the deformation of the heart that is complex in process and independent of the respiratory cycle, may appear as noise in respiratory motion measurements by CT or radiographs. As a result, the accuracy of the mathematical model representing respiratory motion decreases due to the seemingly random heart motion. As a method for dealing with this problem, a method of having the patient hold their breath so that the movement of the lungs does not appear in the image can be considered. Since the patient's respiratory motion is stopped, it may be difficult to identify important information regarding the health of the lungs from the image taken in the breath-holding state.

Summary of the Invention

[0003] Embodiments of the present disclosure aim to solve at least one of the problems of the prior art to some extent by providing a geometrically resolved X-ray imaging system (GREX).

[0004] According to some embodiments, the GREX imaging system acquires images of a patient's chest by calculating and targeting specific respiratory and cardiac phases. The GREX imaging system acquires time-independent anatomical shapes by taking snapshots of the chest during the target respiratory phase across a series of imaging planes. Based on the collected respiratory and cardiac signals, the time-independent anatomical shapes are aligned and interpolated to create an interactive volumetric video of the chest. Interactive videos offer significant advantages over conventional still imaging. Using the interactive video, GREX creates a biomechanical model that quantitatively describes the changes in chest shape that occur during the patient's respiration. Based on fundamental physical laws, the biomechanical model provides estimates of important indicators such as lung tissue elasticity, stress, strain, and respiratory compliance. The diagnostic capabilities based on these characteristics are not yet widely known among medical professionals such as physicians. Therefore, physicians can utilize the information provided by the GREX imaging system to understand the etiology of patients' lung diseases.

[0005] In some embodiments, the GREX imaging system includes proprietary hardware and software add-on packages for use with existing digital diagnostic X-ray units. The combination of hardware and software add-on packages enhances the diagnostic quality of conventional digital diagnostic X-ray units with new and improved imaging capabilities that provide custom imaging procedures using patient biosignals to offer more diagnostically useful information. The hardware and software add-on packages are general-purpose upgrades applicable to any commercially available digital diagnostic X-ray unit (e.g., a conventional X-ray unit).

[0006] According to some embodiments, a method for imaging a patient's lungs is provided. The method includes positioning the patient in a first orientation relative to an X-ray imaging device (e.g., an X-ray unit) and obtaining measurements of the patient's respiratory volume. The method includes determining the patient's respiratory phase based on the measurements of the patient's respiratory volume while the patient is positioned in the first orientation relative to the X-ray imaging device and while obtaining the measurements of the patient's respiratory volume. The method determines the patient's respiratory phase This includes, if it is determined that the patient's respiratory period matches a predetermined respiratory period, performing gate control on the X-ray imaging device to generate an X-ray projection image of the patient's lungs.

[0007] In some embodiments, the default respiratory period is a first default respiratory period among a plurality of default respiratory periods, and the method further includes, while acquiring patient respiratory volume measurements, performing gate control of the X-ray imaging device and generating an X-ray projection image of the patient's lungs if it is determined that the patient's respiratory period matches one of the plurality of default respiratory periods.

[0008] In some embodiments, X-ray measurements of the patient's lungs are obtained only if the patient's respiratory period, determined by the patient's respiratory volume measurement, matches one of several predetermined respiratory periods.

[0009] In some embodiments, a plurality of predetermined respiratory phases include an early expiratory phase, a late expiratory phase, a maximum expiratory phase, an early inspiratory phase, a late inspiratory phase, and a maximum inspiratory phase in the patient's complete respiratory cycle.

[0010] In some embodiments, the X-ray projection image is a first X-ray projection image, and the method further includes repositioning the patient to a second orientation relative to the X-ray imaging device. In some embodiments, the method further includes continuing to determine the patient's respiratory phase based on the patient's respiratory volume measurements while the patient is positioned to the second orientation relative to the X-ray imaging device and while continuing to acquire the patient's respiratory volume measurements, and, if it is determined that the patient's respiratory phase matches a predetermined respiratory phase, performing gate control of the X-ray imaging device to generate a second X-ray projection image of the patient's lungs. In some embodiments, the method includes using the first X-ray projection image and the second X-ray projection image to generate a still image cube corresponding to a predetermined respiratory phase.

[0011] In some embodiments, the still image cube corresponding to a predetermined respiratory period is generated from fewer than 10 X-ray projection images acquired from various angles during the predetermined respiratory period.

[0012] In some embodiments, the patient's respiratory volume measurement includes the patient's chest elevation measurement.

[0013] In some embodiments, the patient's respiratory volume is measured using one or more volumetric respiratory sensors consisting of a 3D scanner, a spirometer, and an abdominal belt.

[0014] In some embodiments, the method further includes creating a point cloud on the surface of the patient's chest, and the patient's respiratory volume measurement is determined from the point cloud on the surface of the patient's chest.

[0015] In some embodiments, a point cloud on the surface of the patient's chest is acquired using 3D imaging technology to measure one or more locations on the patient's chest.

[0016] In some embodiments, the method further includes identifying one or more anatomical landmarks on the surface of the patient's chest using a point cloud on the surface of the patient's chest, and inferring the location of one or more medial anatomical landmarks within the patient's chest from the point cloud on the surface of the patient's chest.

[0017] In some embodiments, the patient's respiratory period is a future respiratory period, and determining the patient's respiratory period based on the patient's respiratory volume measurements includes predicting a future respiratory period from one or more current and / or past respiratory periods.

[0018] In some embodiments, a method is provided. The method involves a first direction relative to a radiation source. The method includes positioning the patient, obtaining respiratory measurements from the patient, and obtaining cardiac function measurements from the patient. The method further includes determining the patient's respiratory phase from the patient's respiratory measurements and determining the patient's heart rate phase from the patient's cardiac function measurements while the patient is positioned in a first orientation relative to the radiation source and while obtaining the patient's respiratory measurements. The method further includes gate control of the radiation source based on the determination that the patient's respiratory phase coincides with a predetermined respiratory phase and that the patient's heart rate phase coincides with a predetermined frame of the cardiac cycle, thereby exposing the patient to radiation.

[0019] In some embodiments, the radiation source is an X-ray imaging device, and gate control of the radiation source to expose the patient to radiation includes gate control of the X-ray imaging device to generate an X-ray projection image of the patient's lungs.

[0020] In some embodiments, the radiation source is a radiotherapy source, and the gate control of the radiation source to expose the patient to radiation includes the gate control of the radiotherapy source to deliver a therapeutic dose to the patient's lung region.

[0021] In some embodiments, the method further includes obtaining patient cardiac function measurements from multiple cardiac cycles of the patient before gate control of the radiation source to expose the patient to radiation, and using the patient cardiac function measurements obtained from multiple cardiac cycles to determine the average interval between a predetermined heart rate phase and the start of a predetermined frame of cardiac cycles. In some embodiments, determining that the patient's heart rate phase coincides with a predetermined frame of cardiac cycles includes predicting the predetermined frame of cardiac cycles by detecting a predetermined heart rate phase in real time and waiting for a length of time corresponding to the average interval between the predetermined heart rate phase and the start of the predetermined frame of cardiac cycles.

[0022] In some embodiments, the measurements obtained from multiple cardiac cycles of a patient are waveform measurements of multiple cardiac cycles, and the method further includes verifying the waveform measurements of multiple cardiac cycles as being statistically stable.

[0023] In some embodiments, the default frame of the cardiac cycle is the sedation frame of the cardiac cycle.

[0024] In some embodiments, a method is provided for determining a biophysical model of a patient's lung from multiple X-ray measurements corresponding to different respiratory phases of the lung. The method includes extracting multiple displacement fields of lung tissue from multiple X-ray measurements corresponding to different respiratory phases of the lung. Each displacement field represents the movement of lung tissue from a first respiratory phase to a second respiratory phase, and each respiratory phase has a corresponding set of bioparameters. The method further includes calculating one or more biophysical parameters of the lung biophysical model using the multiple displacement fields of lung tissue between different respiratory phases of the lung and the corresponding set of bioparameters.

[0025] In some embodiments, one or more biophysical parameters define the physical relationship between a biological parameter associated with different respiratory phases of the lung and multiple displacement fields of the lung tissue.

[0026] In some embodiments, the physical relationship between the bioparameters associated with different respiratory phases of the lung and the multiple displacement fields of the lung tissue is defined by the following equation.

number

[0027]

number

number

number

number

number

[0028] In some embodiments, the method further includes generating multiple medical image cubes corresponding to different respiratory phases of the lung from multiple X-ray measurements corresponding to different respiratory phases of the lung. Multiple displacement fields of lung tissue are extracted from the multiple medical image cubes corresponding to different respiratory phases of the lung by further depicting the lung tissue from the remainder of the first medical image cube by image segmentation. The method includes determining the displacement vector between the voxel in the first medical image cube and the second medical image cube for each voxel in the first medical image cube using intensity-based structural mapping between the first medical image cube and the second medical image cube, and iteratively adjusting the displacement vectors of different voxels in the first medical image cube and corresponding voxels in the second medical image cube.

[0029] In some embodiments, the set of bioparameters associated with each respiratory phase includes the tidal volume and airflow of the lungs during each respiratory phase, as well as the heart rate phase corresponding to each respiratory phase of the lungs.

[0030] In some embodiments, the method further includes generating multiple medical image cubes corresponding to different respiratory phases of the lung from multiple X-ray measurements corresponding to different respiratory phases of the lung. In some embodiments, the method includes selecting one or more of the multiple medical image cubes as reference medical image cubes, determining a set of bioparameters associated with each reference medical image cube, and selecting one set of bioparameters based on lung biometric measurements between two sets of bioparameters associated with two reference medical image cubes. In some embodiments, the method further includes simulating medical image cubes between the two reference medical image cubes by applying the set of bioparameters based on lung biometric measurements to a biophysical model.

[0031] In some embodiments, the different respiratory phases of the lungs include the early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase in the patient's complete respiratory cycle.

[0032] In some embodiments, one or more sensors are used to measure a patient's biological signals as one or more time-series sequences, and one or more sensors include one or more 3D spatial position localizers, respiratory phase sensors, and heart rate phase sensors.

[0033] In some embodiments, the 3D spatial position localizer is configured to measure the patient's body movements caused by respiration and heartbeat in real time and output them as a time series.

[0034] In some embodiments, the respiratory sensor is configured to measure one or more physiological indicators related to a patient's respiration, including tidal volume and its first-order time derivative.

[0035] In some embodiments, the heart rate phase sensor is configured to measure periodic and steady-state electrical signals generated by the patient's heart.

[0036] In some embodiments, an X-ray unit is triggered using the patient's biosignals measured by one or more sensors to acquire X-ray images of the patient during specific respiratory and cardiac phases.

[0037] In some embodiments, the X-ray unit includes a clock, and the patient's biosignals, measured by one or more sensors, are synchronized with the X-ray unit's clock. In some embodiments, each value of the biosignal is recorded in association with the acquired X-ray image.

[0038] In some embodiments, patient biosignals measured during training sessions are used to construct an optimized respiratory prediction model to predict desired respiratory periods for triggering an X-ray unit to take X-ray images of the patient.

[0039] In some embodiments, a method is provided for generating a 3D X-ray image cube video from 2D X-ray images of a patient. The method includes converting a first set of X-ray images of the lung taken at different projection angles into a second set of X-ray images of the lung corresponding to different respiratory phases. The method further includes generating still image cubes from each set of the second set of X-ray images for each respiratory phase using a back projection method, and combining the still image cubes corresponding to different respiratory phases of the lung into a 3D X-ray image cube video using a time interpolation method.

[0040] In some embodiments, converting a first set of X-ray images of the lung taken at different projection angles into a second set of X-ray images of the lung corresponding to different respiratory phases further includes taking the first set of X-ray images of the lung at different projection angles. Each set in the first set of X-ray images of the lung corresponds to a different respiratory phase of the lung at a particular projection angle. The conversion further includes reorganizing the first set of X-ray images of the lung into a second set of X-ray images of the lung according to the associated respiratory phase. Each set in the second set of X-ray images corresponds to a different respiratory phase of the lung.

[0041] In some embodiments, X-ray images within any particular set are geometrically resolved and time-independent.

[0042] In some embodiments, different respiratory phases of the lungs correspond to different tidal volume percentiles of lung movement.

[0043] In some embodiments, the different respiratory phases of the lungs include the early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase in the patient's complete respiratory cycle.

[0044] In some embodiments, multiple X-ray images of the lung taken at different projection angles all correspond to the same respiratory phase.

[0045] In some embodiments, one or more sensors are used to measure a patient's biological signals as one or more time-series sequences, and one or more sensors include one or more 3D spatial position localizers, respiratory phase sensors, and heart rate phase sensors.

[0046] In some embodiments, the method further includes identifying a heart rate phase gate frame using measurements from one or more heart rate phase sensors, predicting a respiratory phase using measurements from one or more respiratory phase sensors, identifying a point of agreement between the heart rate phase gate frame and the predicted respiratory phase that generates an X-ray imaging pulse, and tagging the X-ray image corresponding to the X-ray imaging pulse with measurements from the respiratory phase, heart rate phase, and 3D spatial position localizer.

[0047] In some embodiments, the 3D spatial position localizer is configured to measure the patient's body movements caused by respiration and heartbeat in real time and output them as a time series.

[0048] In some embodiments, the respiratory sensor is configured to measure one or more physiological indicators related to a patient's respiration, including tidal volume and its first-order time derivative.

[0049] In some embodiments, the heart rate phase sensor is configured to measure periodic and steady-state electrical signals generated by the patient's heart and having characteristics corresponding to the heart rate phase.

[0050] In some embodiments, after being synchronized with the X-ray unit's clock, two separate filters are used to remove signal drift and noise from the patient's biosignals.

[0051] In some embodiments, an X-ray unit is triggered using the patient's biosignals measured by one or more sensors to acquire X-ray images of the patient during specific respiratory and cardiac phases.

[0052] In some embodiments, the X-ray unit includes a clock, and the patient's biosignals, measured by one or more sensors, are synchronized with the X-ray unit's clock. In some embodiments, each value of the biosignal is recorded in association with the acquired X-ray image.

[0053] In some embodiments, the patient's biosignals are measured during the training session before the patient's X-ray images are taken, and the patient's biosignals measured during the training session include multiple complete respiratory cycles of the patient.

[0054] In some embodiments, multiple tidal volume percentiles in a complete respiratory cycle are identified using patient biosignals measured during the training period, with each tidal volume percentile corresponding to one of several different respiratory phases.

[0055] In some embodiments, patient biosignals measured during training sessions are used to construct an optimized respiratory prediction model to predict desired respiratory periods for triggering an X-ray unit to take X-ray images of the patient.

[0056] In some embodiments, the optimized respiration prediction model is based on an autoregressive integrated moving average (ARIMA) model.

[0057] In some embodiments, the desired respiratory period for taking an X-ray image of the patient is configured to coincide with a cardiac gate frame in which cardiac-induced lung motion is slowly changing.

[0058] In some embodiments, the cardiac gate frame is selected based on the positions of the T and P waves in the electrocardiogram (ECG) signal, thereby slowing down the cardiac-induced lung motion.

[0059] In some embodiments, different respiratory phases of the lungs at a specific projection angle are collected from at least two respiratory cycles.

[0060] Some further aspects and advantages of the embodiments of this disclosure can be understood from the following description or from the embodiment of this disclosure. [Brief explanation of the drawing]

[0061] To more clearly describe the embodiments of this disclosure and the technical solutions in the prior art, the drawings necessary for describing the embodiments or the prior art are briefly described below. It is only a part of the whole. A person skilled in the art can derive other drawings from the structure shown in the drawing without having to exert any creative effort.

[0062] [Figure 1] A schematic block diagram of a GREX imaging system, including a hardware box, acquisition software, and post-processing software, according to some embodiments of this disclosure.

[0063] [Figure 2] Schematic flowchart of the GREX image acquisition process according to some embodiments of this disclosure.

[0064] [Figure 3] A schematic block diagram showing a top view of a 3D spatial position localizer for a GREX imaging system according to some embodiments of this disclosure.

[0065] [Figure 4]Examples of two respirations that share the same time-based maximum inspiratory phase but have fundamentally different tidal volumes, according to certain embodiments of this disclosure.

[0066] [Figure 5] Examples of tidal volume percentiles (the two subplots below) for patients with normal breathing (upper left subplot) and patients with irregular breathing (upper right subplot), according to certain embodiments of this disclosure.

[0067] [Figure 6] Schematic flowchart of the synchronization process according to some embodiments of this disclosure.

[0068] [Figure 7] Synchronized electrocardiogram (ECG) signals and lung respiratory signals according to certain embodiments of the present disclosure

[0069] [Figure 8] Schematic flowchart of a drift and signal noise reduction process according to certain embodiments of the present disclosure

[0070] [Figure 9] Plot of the relationship between tidal volume and time in exemplary respiration according to some embodiments of this disclosure

[0071] [Figure 10] A plot of the relationship between airflow and tidal volume in the same respiration as shown in Figure 9, according to some embodiments of the present disclosure.

[0072] [Figure 11] Schematic flowchart of the respiratory phase prediction process according to certain embodiments of this disclosure

[0073] [Figure 12] Examples of time-based gate frames selected based on the positions of the T and P waves of the ECG signal so that cardiac-induced lung motion changes slowly, according to some embodiments of the present disclosure.

[0074] [Figure 13] A schematic flowchart of an alternative GREX procedure using heart rate phase prediction instead of real-time identification of the cardiac gate frame, according to certain embodiments of this disclosure.

[0075] [Figure 14] Examples of trigger images in the early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase (from left to right) according to some embodiments of this disclosure.

[0076] [Figure 15] Schematic flowchart of the image acquisition trigger process according to certain embodiments of this disclosure.

[0077] [Figure 16] Schematic block diagram of variables used in the GREX image reconstruction algorithm example according to certain embodiments of this disclosure

[0078] [Figure 17] Schematic block diagram of the operating components of the GREX imaging system during the imaging procedure according to some embodiments of this disclosure.

[0079] [Figure 18] Examples of transverse X-ray projection images at tissue depths of 0°(A) and 90°(B) according to certain embodiments of this disclosure.

[0080] [Figure 19] Examples of depth resolution from a narrow-angle projection angle range (1) and a wide-angle projection angle range (2) according to certain embodiments of the present disclosure

[0081] [Figure 20] Examples of the β imaging angle positions relative to the three main imaging planes (left) and the β imaging angle positions relative to the imaging isocenter of the GREX imaging system in the top view (right) according to some embodiments of this disclosure.

[0082] [Figure 21]An example block diagram illustrating how a GREX imaging system, according to certain embodiments of this disclosure, converts 2D X-ray projection images into still image cubes and ultimately into an image cube video.

[0083] [Figure 22] An illustration of the still image cube creation process shown in Figure 21, according to some embodiments of this disclosure.

[0084] [Figure 23] Schematic flowchart of an image filtering process according to certain embodiments of this disclosure

[0085] [Figure 24] An example of a closed-loop lung tissue orbit of a tissue fragment located in the left lung near the heart, resulting from the interaction between the heart and lungs, according to certain embodiments of the present disclosure.

[0086] [Figure 25] A schematic flowchart illustrating the operation of components of a biomechanical model according to certain embodiments of this disclosure.

[0087] [Figure 26] (A) Illustration of the motion trajectory of lung tissue elements during the respiratory cycle according to certain embodiments of the present disclosure.

[0088] [Figure 26] (B) Illustration of displacement vectors between different respiratory phases according to some embodiments of the present disclosure

[0089] [Figure 27] A schematic flowchart of a multi-resolution 3D optical flow algorithm according to certain embodiments of this disclosure.

[0090] [Figure 28] (A) A schematic flowchart illustrating video creation using biocompletion according to certain embodiments of the present disclosure.

[0091] [Figure 28](B) Block diagram showing the creation of an intermediate image cube using a biometric data matrix according to some embodiments of the present disclosure.

[0092] [Figure 29] Examples of standard radiographs of a healthy person and a patient with a stage 1b left upper lung tumor (arrow).

[0093] [Figure 30] Examples of GREX parameter maps showing health status indicators for healthy individuals and GREX parameter maps showing health status indicators for patients, according to certain embodiments of this disclosure.

[0094] [Figure 31] (A) and (B) are flowcharts of a method for imaging a patient's lung according to some embodiments of the present disclosure.

[0095] [Figure 32] (A) and (B) are flowcharts of a method for controlling the gate of a radiation source according to certain embodiments of the present disclosure.

[0096] [Figure 33] (A) to (C) are flowcharts of a method for determining a biophysical model of a patient's lung according to some embodiments of this disclosure.

[0097] [Figure 34] (A) to (C) are flowcharts of a method for generating a 3D X-ray image cube video according to some embodiments of this disclosure.

[0098] [Figure 35] Examples of patient positioning and restraining devices (PPFs) (e.g., swivel chairs) that support patients, according to certain embodiments.

[0099] [Figure 36] Example of a Biological Event Monitoring Process (BEMP) card according to some embodiments Detailed description of the invention

[0100] Embodiments of this disclosure are described in detail below. Identical or similar elements, or elements having the same or similar function, are assigned the same reference numerals throughout the specification. The embodiments described herein with reference to the drawings are merely examples to enhance understanding of this disclosure and should not be construed as limiting the disclosure.

[0101] In this specification, unless otherwise specified, relative terms such as "center," "vertical," "horizontal," "front," "back," "right," "left," "inside," "outside," "down," "up," "horizontal," "vertical," "upward," "downward," "peak," and "bottom," as well as their derived terms (e.g., "horizontal direction," "downward," "upward"), should be interpreted as referring to the directions described or illustrated in the relevant drawings. The above relative terms are used for explanatory purposes only and do not imply construction or operation in a specific direction.

[0102] In this invention, unless otherwise specified, terms such as "attachment," "connection," "joining," and "fixing" are used broadly, and these may correspond to various forms that a person skilled in the art can understand depending on the specific situation, such as fixed connection, detachable connection, integrated, mechanical and electrical connection, direct connection, indirect connection via an intervening structure, and internal communication between two elements.

[0103] In the present invention, unless otherwise specified, a structure in which the first feature element is "above" or "below" the second feature element includes embodiments in which the first feature element is in direct contact with the second feature element, and may also include embodiments in which the first feature element is not in direct contact with the second feature element, or embodiments in which the first feature element is in contact with the second feature element with an additional element in between. Furthermore, when the first feature element is "above," "above," or "above" the second feature element, this may include embodiments in which the first feature element is directly above or diagonally above the second feature element, or it may simply mean that it is in a higher position than the second feature element. On the other hand, when the first feature element is "below," "below," or "below" the second feature element, this may include embodiments in which the first feature element is directly below or diagonally below the second feature element, or it may simply mean that it is in a lower position than the second feature element.

[0104] Figure 1 is a schematic block diagram of a GREX imaging system 100, which includes a hardware box 102, acquisition software 104 (e.g., stored in non-temporary memory), and post-processing software 106 (e.g., stored in non-temporary memory), according to some embodiments of the present disclosure.

[0105] The hardware box 102 contains three measurement sensors (e.g., a 3D spatial position localizer 300, a respiratory phase sensor 110, and a heart rate phase sensor 112) that independently collect the patient's biological signals as a time series. The collected time series is input to the acquisition software 104. The acquisition software 104 processes and filters the biological time series measurements to generate an imaging trigger signal (e.g., a signal to activate the X-ray unit 108). The imaging trigger signal targets a specific respiratory phase, but may also target the patient's heart rate phase if necessary. The respiratory phase and heart rate phase are defined by the biological time series measurements, respectively. The connecting cable transmits the imaging trigger signal from the acquisition software 104 to the X-ray unit 108, which then acquires radiographic images of the respiratory phase and heart rate phase at the target respiratory phase. The series of images acquired at the target respiratory phase and heart rate phase, which define the complete respiratory cycle, are input to the post-processing software 106. The post-processing software constructs a biomechanical model of lung motion from images acquired during the target respiratory and cardiac phases. Subsequently, the post-processing software generates other diagnostic results using the biomechanical model. While this application uses X-ray images as an example, those skilled in the art will understand that the method disclosed herein can be applied to other medical images with little effort. For example, the biomechanical model construction process is not limited to X-ray images, but can also be used with other medical images (e.g., CT scans or MRI).

[0106] Figure 2 is a schematic flowchart of the GREX image acquisition process 200 according to some embodiments of the present disclosure (for example, performed by the acquisition software 104 in Figure 1). Specifically, the acquisition software 104 synchronizes and processes time-series measurements from the hardware box 102 to remove potential signal drift and noise. The acquisition software 104 then implements a phase prediction algorithm to predict the respiratory and heart rate phases from the synchronized, drift- and noise-free time-series input. Based on the prediction results of the respiratory and heart rate phases, the acquisition software 104 uses a logic algorithm 118 to search for a match with the target respiratory and heart rate phase. The match with the target respiratory and heart rate phase is a condition that enables the generation of an imaging trigger and the acquisition of images by the X-ray unit 108. Once the GREX images are acquired, the post-processing software uses the GREX images to diagnose new lung diseases. GREX images were not widely used by healthcare professionals until recently. For example, using acquired GREX images, a biomechanical model is constructed that defines the shape of the chest as a function of the respiratory and cardiac phases, without necessarily using time parameters.

[0107] This specification describes the GREX-based imaging system in three parts. Chapter 1 relates to an embodiment of the hardware box 102. Chapter 2 deals with an embodiment of the acquisition software 104. Chapter 3 describes an embodiment of the post-processing software 106. Each chapter provides a more detailed description of the components and functions of the hardware box 102, acquisition software 104, and post-processing software 106, as shown in Figure 1.

[0108] Chapter 1. Hardware Box 102

[0109] In some embodiments, the hardware box 102 performs at least two functions: one is to collect biosignals that define the anatomical shape of the chest; the other is to communicate with a digital diagnostic X-ray unit (e.g., X-ray unit 108).

[0110] The biosignal inputs defining the anatomical shape of the chest include chest dimensions (measured by the 3D spatial localizer 300; Figure 3), respiratory phase (measured by the respiratory phase sensor 110), and heart rate phase (measured by the ECG monitor). In the hardware box 102, the biosignals are sampled in real time at 100 Hz to create time-series curves for each signal input. The output of the hardware box 102, i.e., the time-series curves, is sent to the acquisition software.

[0111] Figure 3 is a schematic block diagram showing a top view of a 3D spatial position localizer 300 of a GREX imaging system 100 according to some embodiments of the present disclosure. 1.1.3D Spatial Position Localizer 300

[0112] The 3D spatial position localizer 300 measures the patient's body movements caused by respiration and heartbeat in real time and outputs them as a time series in a coordinate space (e.g., Cartesian coordinate system, polar coordinate system, hyperspherical coordinate system, etc.). As illustrated in Figure 3, the 3D spatial position localizer 300 includes three 3D cameras (e.g., cameras 302a-302c) fixed to a balanced circular path on the ceiling of the room. The three cameras 302a are spaced 120° apart from each other. The patient is positioned between the X-ray unit 108 and the detector panel 304, and is therefore positioned at the center of the ceiling-mounted 3D spatial position localizer 300. The system measures the patient's chest expansion (e.g., rise and fall) in a predetermined coordinate space. Each camera 302 has a field of view that allows an unobstructed view of the patient's torso and one of the X-ray unit 108 and the detector panel 304.

[0113] Using a real-time depth map, each 3D camera 302 creates a rendering of the patient's surface. Information from the three 3D cameras 302 is combined to form volumetric skin surface position measurements that change in real time based on the patient's respiratory and cardiac phases (e.g., using light irradiation techniques). The 3D spatial position localizer 300 uses the volumetric skin surface position measurements in at least two ways: (i) defining the patient's spatial boundaries, and (ii) determining the tissue locations of the patient. Based on prerequisite information such as skin thickness, rib thickness, muscle thickness, and skeletal position obtained from a standard medical internal dose (MIRD) anatomy database, the 3D spatial position localizer 300 estimates the real-time spatial location of the lungs within the patient. The 3D spatial position localizer 300 calculates the spatial boundary conditions of the lungs using MIRD data. These spatial boundary conditions will later be used in post-processing software 106. For example, the estimates of the 3D spatial position localizer 300 for the spatial boundary conditions of the lungs create an initial chest shape that the post-processing software 106 uses to simulate the cumulative tissue density along the rays generated by the X-ray unit 108. 1.2. Respiratory phase sensor 110

[0114] The respiratory sensor 110 of the hardware box 102 measures key physiological indicators related to respiration, namely tidal volume and its first-order time derivative (e.g., the rate of change of tidal volume over time, or flow rate). There are two methods for measuring tidal volume: direct tidal volume measurement and indirect tidal volume measurement. Direct tidal volume measurement uses an oral spirometer consisting of a turbine in a tube that rotates at a speed proportional to the patient's inspiratory or expiratory flow rate. Indirect tidal volume measurement uses an abdominal belt (or other geometric measurements of the patient's chest, described below) to measure the patient's waist circumference change during respiration (see Figure 1). A larger waist circumference indicates inhalation, and a smaller waist circumference indicates exhalation. The abdominal belt cannot directly measure tidal volume. To convert the waist circumference change into a physiologically meaningful quantity, the hardware box 102 correlates the waist circumference change with the estimated lung volume determined by the 3D spatial position localizer 300. For example, chest expansion is proportional to abdominal expansion during respiration. It is also possible to estimate the air content of the lungs by using measurements from both the abdominal belt and the 3D spatial position localizer 300.

[0115] In this specification, "tidal volume" refers to the difference between the current lung volume and a predetermined baseline (for example, the amount of air exhaled maximally during normal, effortless breathing, the amount of air inhaled maximally during normal, effortless breathing, or other baseline volume). According to the law of ideal gases, due to the density difference between room air and internal air, the air in the lungs expands by 11% beyond the tidal volume. According to the law of conservation of mass, the body expands by 11% beyond the inhaled volume. Therefore, the tidal volume of the lungs... This can be calculated using external body measurements calibrated to match the internal air volume. Furthermore, the 3D spatial position localizer 300 determines the patient's volume expansion rate during respiration and performs a secondary validation of the accuracy of the air volume and tidal volume measurements. In other words, it compares the body's volume expansion rate with estimated air volumes of the trachea, lungs, and bronchi obtained from X-ray images. 1.3. Heart rate phase sensor 112

[0116] As illustrated in Figure 1, the heart rate phase is measured using either an electrocardiogram (ECG) monitor or a blood volume pressure device (e.g., a heart rate phase sensor 112). For example, a clinician might use an ECG monitor, attaching leads to the patient's arm and a ground lead to the lower left side of the patient's abdomen (away from the diaphragm and abdominal belt). The human heart generates periodic, steady-state electrical signals with characteristics corresponding to the phases of the heartbeat. The steady-state signal is a stochastic process whose joint probability distribution does not change over time. The blood volume pressure device uses a light source and a light sensor to measure the optical attenuation of the patient's finger. The amount of circulating blood pumped out with each heartbeat changes the optical attenuation of the finger. The optical attenuation is directly proportional to the heart rate phase.

[0117] Typically, a digital diagnostic X-ray unit (e.g., X-ray unit 108) is operated by an analog plunger attached to a plug port. The plug port is uniquely configured to accept a plunger having a specific pin structure. Each pin structure includes an "acquisition pin" that receives a voltage signal to operate the X-ray unit (e.g., gate control). When the user presses the plunger, a voltage pulse is sent to the digital diagnostic X-ray unit. The digital diagnostic X-ray unit activates an imaging beam that penetrates the patient's body.

[0118] As illustrated in Figure 1, the hardware box 102 transmits the same voltage pulse (e.g., a gate signal) from the acquisition software to the digital diagnostic X-ray unit 108 using a connecting cable. When the voltage pulse signal exceeds a predetermined voltage threshold, it activates the digital diagnostic X-ray beam (e.g., gates the X-ray beam), and when the voltage pulse signal at the pin falls below the predetermined voltage threshold, the X-ray unit 108 turns off. In some embodiments, the hardware box 102 generates a square wave signal with a pulse height greater than the predetermined voltage threshold and maintains this pulse height during image exposure. That is, it generates a square pulse of voltage Y lasting for X seconds so that the X-ray unit 108 can acquire an X-ray image of the patient. Since a subthreshold voltage below Y does not trigger the scanner, the value of Y must be greater than the predetermined voltage threshold in order to start X-ray imaging. The pulse duration X refers to the image exposure time, which begins the moment the voltage Y exceeds the predetermined voltage threshold and ends when the voltage falls below the predetermined voltage threshold. When the voltage falls below the predetermined voltage threshold, the X-ray unit 108 turns off. The pulse duration is defined in the manufacturer's specifications, but it is usually around a few milliseconds.

[0119] Chapter 2. Acquired Software 104

[0120] The acquisition software 104 collects spatial, cardiac, and lung time series measured by the hardware box 102 and determines the timing to trigger the X-ray unit 108 to acquire X-ray images at specific respiratory or cardiac phases. The acquisition software 104 precisely overlays (e.g., synchronizes) the measured cardiac and lung phases, processes (e.g., filters) the collected biological time series, determines the appropriate imaging time, and generates an electronic trigger signal for the digital diagnostic X-ray unit 108. The electronic trigger signal (e.g., gate signal) activates the X-ray unit 108 to acquire a snapshot image of the chest shape. The spatial, cardiac, and lung values ​​associated with the snapshot are recorded, defining the surface shape of the chest when the image was taken. The process is fully automated and does not depend on user input. As illustrated in Figures 1 and 2, the process implemented by the acquisition software 104 consists of four auxiliary components, namely 2.1 (synchronization module 114), 2.2 (signal). It includes processing module 116), 2.3 (logical algorithm 118), and 2.4 (trigger generation module 120). 2.1. Synchronization Module 114

[0121] The inputs to the synchronization module 114 include signals from the 3D spatial position localizer 300, the respiratory phase sensor 110, and the heart rate phase sensor 112. The synchronization module 114 synchronizes these inputs with the clock of the X-ray unit 108. Physiological biological signals are acquired asynchronously and therefore require synchronization later. The reason for this asynchronous nature is that the respiratory cycle is slower than the cardiac cycle and completely independent of it. As mentioned above, the respiratory cycle and cardiac cycle are measured separately by different sensors. The synchronization module 114 is configured to synchronize the respiratory phase sensor and heart rate phase sensor with acquired images. When an image is taken, it displays the anatomical shape of the chest at a certain point in time. This point in time is recorded by the X-ray unit 108's intrinsic timing system, but it is not necessarily synchronized with the biological time series of the respiratory and cardiac sensors.

[0122] Time alone cannot distinguish between periods of irregular breathing and periods of normal breathing. In other words, if time is the only dimension defining the respiratory period, then images taken during normal breathing and images taken during abnormal breathing (such as coughing) are computationally indistinguishable. Figure 4 shows an example of two respirations 400 (e.g., 400a and 400b) that share the same maximum inspiratory phase in time but have fundamentally different tidal volumes, according to some embodiments of the disclosure. When superimposed, it can be seen that the two respirations 400 are actually different despite having similar maximum inspiratory phases. This is because the magnitude of the tidal volume is different. In some embodiments, tidal volume refers to a measured volume value (e.g., in ml) relative to a reference volume value. For example, the reference volume value is the minimum lung volume value (e.g., during the maximum expiratory phase) during normal breathing (e.g., without extra effort or forced exhalation) of the patient. In some embodiments, the reference volume value is different for each patient. In some embodiments, the reference volume value is 0. In some embodiments, tidal volume is measured at a certain point in time. Figure 4 illustrates the change in tidal volume over a certain period.

[0123] To overcome the limitations of the time dimension, the GREX imaging system 100 defines the respiratory phase by physiological values ​​obtained from various physiological sensors in the hardware box 102. This can be considered a more useful respiratory phase dimension than time. The function of the synchronization module 114 is primarily to enable seamless and smooth transitions between the X-ray unit 108 and the GREX imaging system 100. In some embodiments, the acquisition software 104 uses a 30-second training frame (described in detail in 2.2 below) to calculate tidal volume percentiles from time-series tidal volume observations collected during the training frame. The acquisition software 104 defines the respiratory phase using tidal volume percentiles rather than the interval between peaks of the periodic cosine wave. The tidal volume percentiles of the acquisition software 104 are far more useful for defining lung shape than the interval between peaks of the periodic cosine wave because tidal volume fluctuates with each breath.

[0124] Figure 5 shows examples of tidal volume percentiles (the two lower subplots) for patients with normal breathing (upper left subplot) and patients with irregular breathing (upper right subplot) according to some embodiments of the present disclosure. For quantitative evaluation of the tidal volume histogram for irregular breathing, the ratio of normal inspirational tidal volume to abnormal inspirational tidal volume is used as an indicator of the respiratory phase. This ratio has a threshold that defines the probability of irregular breathing in the patient. As shown in the two lower subplots of Figure 5, the vertical lines indicate the positions of the 85th, 90th, 95th, and 98th percentile tidal volumes in the tidal volume histogram. During normal breathing (lower left subplot of Figure 5), the percentile values ​​of normal tidal volume (85th and 90th) are higher than the percentile values ​​of abnormal tidal volume (95th and 98th) during irregular breathing (lower right subplot of Figure 5). It is nearby.

[0125] Figure 6 is a schematic flowchart illustrating the synchronization process 600 between various sensors of the hardware box 102 and the diagnostic X-ray unit 108, according to some embodiments of the present disclosure. The hardware box 102 continuously collects respiratory phases defined by tidal volume percentiles, heart rate phases defined by the ECG, and chest shape in coordinates defined by the 3D spatial position localizer 300. The sensor signals that give rise to these measurements must first be synchronized with each other before they can be synchronized with the X-ray unit 108. To this end, differences in the measurement channels of the various sensors and differences in the electrical resistivity of the cables are compensated for by matching the trace lengths using impedance matching in the digital-to-analog converter. Since the clock of the X-ray unit 108 (e.g., timing system) is not typically synchronized with the clock of the hardware box 102, the connection cable is connected to a data acquisition board, which is configured to interface the board with the clock of the X-ray unit 108. The clock of the X-ray unit 108 passes through a distribution buffer and is trace-matched for each channel of the analog-to-digital converter. The converted digital signals are sent to a field-programmable gate array, where all signals are synchronized. When the acquisition software 104 properly synchronizes the signals from the respiratory phase sensor and heart rate phase sensor, the result will be similar to the example shown in Figure 7.

[0126] To prevent patients from being exposed to unnecessary radiation, the acquisition software 104 will not send a trigger signal to activate the X-ray unit 108 via the connecting cable unless the components of the GREX imaging system 100 are properly synchronized. In some embodiments, a 30-second training frame is used to verify the synchronization between the clock of the hardware box 102 and the clock of the X-ray unit 108. Therefore, the 30-second training frame must contain a sample equivalent to 30 seconds. If all the sensors of the hardware box 102 and the clocks of the X-ray unit 108 do not show a sample equivalent to exactly 30 seconds, they are not synchronized. If a discrepancy is found from the verification procedure described above, the synchronization system restarts to correct the discrepancy. However, the 30-second training frame is merely an example, and those skilled in the art will understand that the length of the training frame is not particularly important as long as there is enough data to perform the synchronization process. 2.2. Signal Processing Module 116

[0127] After the spatial position, respiratory phase, and heart rate phase signals are synchronized, the acquisition software 104 processes the sensor signals to remove noise and ultimately predicts the patient's tidal volume with high accuracy. The noise present in the lung and heart measurement time series is due to the sensor's electronics, electrodes, and background electrical signals. A series of filters ensures temporal accuracy by removing noise from the lung and heart measurement time series. The biological time series filtered by the series of filters can maintain temporal accuracy.

[0128] In some embodiments, two separate filters (e.g., wavelet filters) are used to remove signal drift and noise from the biological time series. Signal drift introduces temporal distortion to the measurements, causing the measurements at the start of data acquisition to no longer match the measurements at the end of data acquisition. Signal noise is not inherently physiological and can pose a serious problem when calculating patient airflow from tidal volume measurements. Figure 8 is a schematic flowchart of a drift and signal noise reduction process 800 according to some embodiments of this disclosure.

[0129] The acquisition software 104 requires a smooth tidal volume time series to calculate the first time derivative of tidal volume, for example, airflow. If the tidal volume time series is not smooth, the first time derivative of tidal volume cannot generate a smooth curve. The curve will contain discontinuities that contradict the biophysical reality of respiration. Figure 9 shows some embodiments of the present disclosure. The image shows a plot of the relationship between tidal volume and time in an exemplary respiration. Figure 10 shows a plot of the relationship between airflow and tidal volume in the same respiration as shown in Figure 9, according to some embodiments of the present disclosure, i.e., a continuous closed loop used by the post-processing software 106 for biomechanical modeling (see 3.1).

[0130] The acquisition software 104 performs two different functions using filtered and time-accurate time-series curves. The first function of the acquisition software 104 is to generate respiratory phase predictions using a short prediction range. Figure 11 is a schematic flowchart of the respiratory phase prediction process 1100 according to some embodiments of the present disclosure. The short prediction range is a "look-ahead" of the prediction algorithm. The prediction algorithm predicts future moments in which a desired tidal volume and airflow (e.g., respiratory phase) will occur. The "desired" respiratory phase is set as a target that can trigger the diagnostic X-ray unit 108 to acquire the desired chest shape.

[0131] A short prediction range reduces further sources of inaccuracy in respiratory prediction, such as per-breath variations in respiratory amplitude and respiratory duration. As the limit of the prediction range approaches zero, lung shape changes also approach zero (e.g., lung shape becomes substantially constant). In other words, there is little possibility that lung shape will change dramatically during a short prediction range. Therefore, a shorter prediction range reduces the impact of human respiratory variability on the prediction accuracy of respiratory motion models.

[0132] A time series of filtered tidal volume with high temporal accuracy serves as input to a respiratory prediction algorithm. This algorithm enables fast, real-time, and highly accurate prediction of respiratory phases. For example, the respiratory prediction algorithm is based on an autoregressive integrated moving average (ARIMA) model. The ARIMA model is suitable for respiratory prediction because it does not assume stationary input values ​​and is composed of polynomials. The polynomial coefficients of the ARIMA model are estimated during a 30-second training window starting from the beginning of the imaging examination. The number of polynomial coefficients in the ARIMA model, e.g., the model's order, is validated using nonlinear optimization to minimize the information-based search function and reduce or eliminate overfitting. If the model's order is optimal for the collected training data, a tidal volume histogram (see 2.1) is created, and the probability density function is calculated using a log-likelihood objective function. As described in 2.1, the tidal volume distribution is used to identify irregular breathing. If irregular breathing is detected, the training data is discarded and reacquired. If irregular breathing is not detected, the ARIMA model coefficients are estimated using the maximum likelihood method with the training data and probability density function. The 30-second training window also serves as instrument validation before imaging. Figure 11 is a flowchart of the respiratory phase prediction process.

[0133] The second function of the acquisition software 104 is to identify the heart rate phase and ensure that the heart is in phase with each desired chest shape. The duration of the prediction range is a critical parameter in the acquisition software 104's efforts to accurately predict human respiration, as human respiration is a quasi-random function (each breath has its own unique aspects). In some embodiments, the duration of the prediction range is longer than the sum of the latency of the digital diagnostic X-ray unit 108 and the exposure time for X-ray imaging. The sum of the latency of the digital diagnostic X-ray unit 108 and the exposure time for X-ray images is very short, about 10 milliseconds. As a result, the prediction period is also short (about 1-2 sensor measurement sample values ​​at an operating frequency of 100-1000 Hz).

[0134] When the acquisition software 104 searches for a match between the cardiac and respiratory phases, the likelihood of the phases, represented by a single point in each time series, aligning is low. As a result, the imaging method that searches for a single-point match takes longer to complete because it has to wait for the acquisition software 104 to find a match with an extremely low probability. However, the cardiac gate frame expands the size of the match frame, thus reducing the time required for imaging.

[0135] To further reduce the computation time of the signal processing software, the signal processing software does not predict the heart rate phase. Instead, the signal processing software targets a specific gate frame in which the heart does not cause rapid displacement of the lungs. Figure 12 shows an example of a time-based gate frame selected based on the T-wave and P-wave positions of the ECG signal so that cardiac-induced lung motion changes slowly, according to some embodiments of the present disclosure. The dotted line represents lung motion due to the heart rate. The rate of change of lung motion (e.g., velocity) is represented by the slope of the dotted line. A small slope of the dotted line indicates a small rate of change, and the corresponding heart rate phase is an ideal gate frame. Figure 12 shows that the gate frame occurs (consistently) between the T-wave and the P-wave and is more distorted towards the P-wave. Lung motion due to the heart rate is minimal in the gate frame.

[0136] The above describes a method for identifying an ideal cardiac gate frame that minimizes the physical impact of the heart on the lungs while maintaining a frame in which the target respiratory phase can coincide with the desired cardiac phase. In some embodiments, the GREX imaging system 100 predicts the cardiac phase instead of gate-controlling it, based on the difference in signal processing that distinguishes the cardiac phase from the respiratory phase (e.g., the cardiac phase is periodic and stable). Because the cardiac phase is periodic and stable, an unsupervised multilayer perceptron using backpropagation can predict the next heartbeat based on pattern extraction instead of a time-series prediction process.

[0137] Figure 13 is a schematic flowchart of another GREX process 1300 that uses heart rate phase prediction instead of real-time identification of the cardiac gate frame, according to some embodiments of the present disclosure. In this case, 20 seconds of the training frame (heart rate 20-22) is used to train the algorithm, and the remaining 10 seconds (heart rate 10-11) is used to validate the weights of the multilayer perceptron nodes. The node weights are determined iteratively by gradient descent optimization until the model error on the training set is minimized. The trained model is applied to 10 seconds of validation data, and if the multilayer perceptron's heart rate phase prediction on the validation data is insufficient, the node weights are recalculated. 2.3. Logic Algorithms 118

[0138] Medical algorithms and systems already exist that identify the T wave and P wave of an ECG during the cardiac cycle. Since the heart rate phase sensor 112 continuously measures the cardiac cycle, the time interval between the T wave and the subsequent P wave (equal to a constant proportion of the cardiac cycle and therefore proportional to the heart rate) is also well known. The logic algorithm 118 uses the ECG features within the time interval to introduce a short time lag before starting the gate frame, such that the gate frame starts, for example, midway between the T wave and the P wave, and closes after the logic algorithm 118 has identified the P wave.

[0139] Figure 14 shows an example of triggered X-ray image frames corresponding to different respiratory phases of the respiratory cycle, such as early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase (from left to right), according to some embodiments of the present disclosure. In this case, the acquisition software 104 identifies at least six respiratory phases representing a single respiratory cycle. During the training period, the acquisition software 104 generates a sample distribution, and the logic algorithm 118 calculates tidal volume percentiles that define the target respiratory phase of the logic algorithm 118. The arrows shown in Figure 14 point to six cardiac gate frames that coincide with the target respiratory phases from which the logic algorithm 118 should generate imaging trigger signals. Once the acquisition software 104 acquires a respiratory phase, it generates an automatic validation to prevent duplicate imaging of the same respiratory phase.

[0140] Figure 15 is a schematic flowchart of the image acquisition trigger process 1500 according to some embodiments of the present disclosure. The heart rate phase sensor measurement is used to identify the cardiac gate frame, as previously described with reference to Figure 12. The respiratory phase sensor measurement is used to predict the respiratory phase, as previously described with reference to Figure 11. The logic algorithm 118 is used to predict the heart rate phase Identify the match between the frame and the predicted respiratory period. If a match is found, validate the list of respiratory periods to determine if they have already been acquired. If the respiratory period has been acquired previously, do not generate an imaging trigger pulse. If the respiratory period has not yet been acquired, generate an imaging trigger pulse and take a snapshot of the patient's anatomical structure. Record the respiratory period, heart rate phase, and 3D spatial position localizer 300 measurements and tag them to the image. Once all respiratory periods have been acquired, update the respiratory period list to avoid duplicate image acquisition. 2.4. Trigger generation (gate control)

[0141] The X-ray unit 108 has a port containing a series of electrical pins. One of the pins receives an electrical impulse that defines the timing and duration of radiation exposure. Based on the identified respiratory period of the logic algorithm 118 within the cardiac gate frame, the trigger generator generates a square wave trigger as an electrical impulse. The generated trigger signal is transmitted to the X-ray unit 108 via a fiber optic cable with a vendor-specific plug. Chapter 3. Post-processing software

[0142] The imaging triggers, based on biometric information identified by the hardware box 102 (Chapter 1) and acquisition software 104 (Section 2), input high-quality data (and remove low-quality input) through the image reconstruction algorithm. Specifically, the process of acquiring targeted images based on biometric data during normal respiration improves image reconstruction and post-processing techniques, thereby improving quality. Because the fundamental assumptions of underlying radiological mathematics presuppose anatomical equivalence at imaging angles with various probes, the process of acquiring targeted images based on biometric data during normal respiration allows for more accurate correlation of multiple images of anatomical shapes taken from different angles and at different times (e.g., different breaths) of the same patient. The improved quality images function as observational data, and the biometric signals function as input data for complex biomechanical models of chest shape. 3.1. Acquisition of GREX images using digital tomosynthesis

[0143] 3D reconstruction requires multiple imaging angles. In the case of GREX imaging, each angle must be acquired for each respiratory phase. The acquisition software 104 (Chapter 2) creates a trigger signal that allows the X-ray unit 108 to repeatedly image the chest in a specific shape. The same shape, acquired at different imaging angles and with different respiratory phases, becomes a set of 2D projection images used to reconstruct 3D. Many methods for reconstructing 3D from multiple 2D projection images are already known. An example is the convolutional backprojection algorithm, which directly reconstructs the 3D density function using a series of 2D projections. This is called the "FDK image reconstruction algorithm" and is disclosed in Feldkamp LA, Davis LC, Kress JW, "Practical cone beam algorithm," J Opt Soc Am 1, 612-619 (1984).

[0144] Figure 16 is a schematic block diagram of the variables used in the image reconstruction algorithm according to some embodiments of the present disclosure. When a single projection image is acquired, the X-ray unit 108 moves by an angle β, and the detector plane moves so as to remain perpendicular to the X-ray unit 108. In some embodiments, the patient moves, and the X-ray unit 108 remains in the same position during the acquisition of the projection image. Figure 17 schematically illustrates an example of the process by which the X-ray unit 108 and detector panel 304 of Figure 3 move to obtain multiple imaging angles. For example, the X-ray unit 108 moves from position 1700a to position 1700b, and then to position 1700c. The detector panel 304 moves from position 1702a to position 1702b, and then to position 1702c. The detector plane rotates around an axis parallel to the detector plane, and the imaging plane rotates around its parallel axis z. The positions of pixels in the detector plane and their corresponding pixels in the imaging plane are separated by a distance s. The anatomical information (f(x, z / y)) of the imaging plane ((x, z) plane) at an arbitrary depth y is calculated using the following equation (1).

number

[0145] In equation (1), N O β is the total number of projected images, β is the angle of each projected image, d is the distance from the light source to the image plane, s is the distance from the pixel to the detector, p is the detector axis perpendicular to the rotation axis, ξ is the detector axis parallel to the rotation axis, R(β,p,ξ) corresponds to the cone-beam projection data (for example, the function R is the detector reading for a specific angle, and p and ξ are coordinates), h is the convolution filter, and W(p) is the weight function. Essentially, equation (1) represents a combination of convolution, back projection, and weighting.

[0146] Information about a point on the central plane can be calculated from projection data along the intersection of the detector plane and the central plane (y=0). Projection images that intersect the detector plane along lines parallel to the central plane but do not exist within the central plane (constant, non-zero y) define a plane themselves. This plane is treated as another inclined central plane. Once a complete set of projection images is obtained ("complete" means all rotation angles around the normal are obtained), the density of the inclined plane is reconstructed using the Radon transform. To obtain a complete set of projection images, the light source must be rotated 360° around the subject along the circle of the inclined plane. For example, in CT imaging, the light source is rotated a full 360° around the subject. The bolded "subject" implies (more precisely, explicitly states) that deviating from the fundamental assumption of the radon transformation (which is also incorporated into the R(β,p,ξ) term in equation (1)), namely, the assumption that the radon transformation input is viewed as various angular probes to a stationary object of a single, fixed, non-moving anatomical structure, where time and space do not change, would impair the representativeness of the reconstruction of the anatomical structure in the ground validation data underlying the imaging. GREX imaging targets only highly accurate (occurring in different respirations but geometrically identical respiratory periods) radon transformation inputs based on biological data (i.e., "pre-selection"). Here, since the respiratory period (i.e., change in the time domain) is used in GREX, which is geometrically, anatomically, and physically rigorously defined (e.g., a definitionally correct formula), and in practice uses the short prediction range inherent to GREX imaging and prospective targeting based on biological data, it can be said that the fundamental assumption of the transformation is adhered to both definingly and practically.

[0147] However, 360° rotation is not practical in GREX imaging because the numerous projection images required for 3D reconstruction of the torso increase the clinical procedure time and the patient's radiation dose. Instead, GREX imaging uses a projection angle range of -45°≦β≦45° or 0°≦β<90°, with a maximum of 90°. Those skilled in the art will understand that while experimental testing can determine a more optimal projection angle range, the theoretical projection angle range will not exceed 90°. In some cases, -45°≦β≦45° is preferable to 0°≦β<90°, in order to keep the radiation dose of the -45°≦β≦45° imaging procedure as low as possible. X-ray photons penetrate human tissue less frequently in the -45°≦β≦45° projection angle range compared to the 0°≦β<90° projection angle range, and therefore the photon energy is lower and the dose is reduced, as illustrated in Figures 18A and 18B.

[0148] To generate high-quality 2D images without excessive radiation exposure, the photon energy must be high enough to partially penetrate the patient's body, but not high enough to completely penetrate it. Patients with greater body thickness require higher photon energy than those with less. When |β|>45°, the human body is considerably thicker than when |β|<45°. Generally, as β→0°, the photon energy decreases. In some embodiments, the GREX imaging system 100 acquires six respiratory phases at five different projection angles, resulting in a total of 30 projections, but the number of projections may vary depending on the application of the GREX imaging technique. For example, in breast tomosynthesis, the symmetrical curvature of the breast means that the breast surface is essentially equidistant from the light source at all projection angles. In other words, tomosynthesis is suitable for the breast. Furthermore, the breast does not move when placed on the cradle. This can be considered a typical clinical tomosynthesis, similar to lung imaging while holding your breath.

[0149] The challenges related to lung and cardiac motion can be addressed by GREX-based geometric definitions for the respiratory phase, in addition to highly accurate and fast GREX-based targeting prediction algorithms (2.1 and 2.2). Therefore, the GREX imaging stem 100 can handle changing torso curvature using biosurface information collected by the 3D spatial localizer (1.1). This biosurface information assists in image post-processing to address the attenuation of tissue density in the imaging field. This quantifies previously ignored attenuation sources, ultimately resulting in more faithful image reconstruction. In summary, GREX imaging technology enables imaging of the lungs and heart ("dynamic") without breath-holding through digital tomosynthesis.

[0150] GREX-based 3D reconstruction may be tailored to individual patients (custom medicine using individualized angles and corresponding arcs), or it may be used as a "general minimum procedure time and general minimum delivery dose" (considering past reconstruction method statistics and the number of previous GREX-based datasets for the patient across approximately five discrete angles ± three angles).

[0151] As shown in Figure 19, when the rotation of the X-ray unit 108 is within a narrow range of angles, the depth resolution is lower than when the rotation of the X-ray unit 108 is within a wider range of angles. Motion, regardless of size, introduces image artifacts into the reconstructed image, generally resulting in the detection of false-positive cancers. However, the GREX imaging system 100's unique hardware box 102 (Chapter 1) collects biosignals that notify the acquisition software 104 (Chapter 2) of the timing to acquire images so that the shape of the chest is clinically identical, thereby enabling smart chest digital tomosynthesis.

[0152] In some embodiments, GREX imaging techniques can vary the projection angle during a single respiratory phase, but multiple projection angles captured at different points in time (defined by the quantitative definition of a "respiratory phase" in GREX imaging) all correspond to the same respiratory phase. In other words, they are all considered to be imaging a single chest shape. Furthermore, in GREX-based tomosynthesis, depth information can be obtained because the final number of photons in the detector across all pixels and the distribution of photons in space on the detector surface reflect a single chest shape probed at multiple angles.

[0153] In GREX imaging technology, the X-ray unit 108 and detector panel 304 can also be mounted on a non-motorized arm or stand. An example of a manual arm in the GREX imaging procedure is illustrated in Figure 20. 1) At the β1 position, all six essential chest shapes ("quantitatively defined respiratory phases of GREX") are imaged. 2) The clinician physically changes the orientation of the X-ray unit 108 and the detector panel 304 to image the patient at the β2 position. 3) The orientation of the X-ray unit 108 and detector panel 304 relative to the patient is verified by the 3D spatial localizer, so imaging can be performed with β2. 4) In β2, if only four of the six chest shape images required are acquired during the first breath, the patient should continue breathing normally in the β2 position until the remaining two chest shape images are acquired. 5) Since all six respiratory patterns were acquired using β1 and β2, the X-ray arm was repositioned and each imaging angle was obtained in the order of (β2→β3; β3→β4; β4→β5). 6) As indicated by the arrows in the workflow, the X-ray arm moves only four times in total until the end of the procedure (β1→β2; β2→β3; β3→β4; β4→β5).

[0154] With GREX imaging technology, the arm is moved only four times during the procedure, so the clinician interacts with the X-ray arm only by moving it from β1 to β5 over six respiratory phases (for example, changing the orientation of the X-ray machine 30 times). This minimizes procedure time, interaction between the clinician and the equipment during the procedure, and the degree of wear on the X-ray arm.

[0155] In some embodiments, the 3D spatial positioning localizer enables pre-position verification and safety interlocks for three individual elements: the patient's posture and position, the position of the X-ray unit 108, and the position of the detector in space, and can maintain collective consistency in the position / placement of each element relative to the positions of axes and other elements. Since GREX imaging geometrically defines the respiratory phase, pre-position verification and safety interlocks are also possible. Thus, the GREX imaging system 100 (expected respiratory phase prediction algorithm) can provide the user with software-based safety and quality assurance control functions that prevent the trigger algorithm from initiating the "beam on" state (in the case of safety interlocks) even if either (or both) the X-ray arm or the detector, or the patient's posture, is mispositioned (or inconsistent) in space relative to a specific angle β.

[0156] In some embodiments, the 3D spatial positioning localizer records all of the instrument's coordinates during the procedure, so the image reconstruction technique has the advantage of being able to (post-mortem) quantify each β angle and its associated uncertainty.

[0157] Figure 21 is an example block diagram showing how the GREX imaging system 100 reconstructs a still image cube from 2D projection image data captured from each imaging angle position, according to some embodiments of the present disclosure. Specifically, Figure 21 shows an example of GREX imaging where the coronal or sagittal plane itself forms the outer limit / boundary of the imaging angle position (e.g., projection angle). The X-ray projection image of the i-th plane is used for each of the six breathing shapes, with a single imaging angle position (β iThe images are taken from ). The sum of the planar projections from i=1 to i=n reconstructs the depth information at (x,y,z) of the (v1,f1) shape probed using the projections from i=1 to i=n (Equation 1), and each has a focal plane at a unique depth (compared to the other projections).

[0158] 2D projection image data is acquired six times for each projection angle (s) (early intake (EI), late intake). Inspiration (LI), maximum inspiration (MI), early expiration (EE), late expiration (LE), maximum expiration (ME). Simply put, the projection angle ranges from 0° to 90°. X-ray unit 108 moves to the next projection angle only if all respiratory phases have been acquired at the previous projection angle. Since 2D projections are classified according to the respiratory phase, the still image cubes are reconstructed (for example, β probed with (v1, f1)). i = 1 to β i The cube represents the volume of the chest during each respiratory phase (up to a projection angle of =n). The static image cube is interpolated over time using the method described in 3.3 below.

[0159] X-ray projection images acquired at each angle β are also acquired at the target respiratory phase identified using the ARIMA model (2.2). After the target respiratory phase is acquired, the X-ray unit 108 moves to the next imaging angle position. Figure 22 illustrates an example where only the maximum inspiratory and maximum expiratory phases are acquired using the acquisition software 104. In this example, after the maximum inspiratory and maximum expiratory phases are acquired at β1, the X-ray unit 108 moves to β2 and acquires the maximum inspiratory and maximum expiratory phases at β2 as well. Once the target respiratory phase is acquired at all imaging angle positions, the images are classified according to each respiratory phase. The X-ray images are sorted according to the respiratory phase, and the projection images are grouped according to the six biologically defined respiratory phases. Although the respiratory phases are acquired during different breaths, the highly accurate ARIMA model ensures that the tidal volume and airflow parameters are identical even in X-ray projection images taken at different projection angles. GREX imaging biologically defines the "same chest shape" so that the lungs exist in the "same chest shape" at multiple points in time. Since the ARIMA model (2.2) is a fast prediction method, GREX imaging probes the same respiratory period at different angles. This will result in minimizing chest shape prediction errors using the short prediction range of the ARIMA model. The combined equivalence of depth information at individual angles (to generate actual structural depth information as a whole) depends on the equivalence (consistency, within approximate tolerance) of the geometric position of the chest across various probe angles. Thus, the accuracy of respiratory phase prediction ensures successful image reconstruction.

[0160] Using projection images classified by respiratory phase, still image cubes are created using the aforementioned standard FDK image reconstruction algorithm (or a similar cone-beam shape image reconstruction algorithm) given by equation (1). The image reconstruction algorithm creates static image cubes classified by respiratory phase using projection images classified by respiratory phase. Each respiratory phase has its own individual image cube. A still image cube is called a still cube because it represents an anatomical structure in a single respiratory phase. A 3D image cube video is created from 3D still image cubes by combining the still cubes representing all target respiratory phases and interpolating them in time (see 3.3).

[0161] The GREX imaging system 100 maintains the exposure dose at the lowest level reasonably achievable through statistical image reconstruction. Each acquired image increases the overall dose of the imaging process (clinically undesirable), but provides additional information used for image reconstruction (clinically desirable). Conventional image reconstruction methods based on Fourier transforms and filtered back projection tend to produce image artifacts because they cannot handle missing information, such as missing projection angles β. For example, if projection is performed every 10° instead of 5°, the amount of information available for creating a still image cube is halved, and the dose is also halved compared to the former. Statistical iterative image reconstruction can address the loss of information resulting from incomplete image datasets.

[0162] The GREX imaging system 100 may complete the image reconstruction task using various existing statistical image reconstruction algorithms (for example, constructing a still image cube). Alternatively, the GREX imaging system 100 can improve upon conventional statistical image reconstruction algorithms by implementing its own feedback steps and adhering to boundary conditions based on the law of conservation of mass.

[0163] Because GREX images are biophysically defined by biophysical quantities governed by physical laws, fundamental principles of physics can be applied to GREX's statistical image reconstruction. By biotagging each image and the resulting image cube, and by collecting a continuous biodata stream during the procedure even when the patient is not being imaged, respiratory dynamics of mass exchange (inspiration and expiration) and volume changes can be understood (solving the problem of invariant lung tissue volume that is consistent throughout the scanning process). Since static image cubes can only be constructed from moving organs such as the lungs, the law of conservation of mass can be applied. This is due to the speed and accuracy (2.2 and 2.3) of the predictive / trigger algorithms that accurately label and acquire the same chest shape at different points in time. In other words, the mass of tissue in a static image cube should not change between any static image cube and the next static image cube (for example, it does not change due to the law of conservation of mass). Based on the law of ideal gases, the ratio of air at room temperature to air in the lungs is 1.11. By obtaining the tidal volume from the image cubes using sensor data, and employing the ratio of 1.11 and the mass / volume air curve of the deviation relative to room temperature, the inspiratory mass (absolute value, and relative value as a ratio of image cubes at different respiratory phases) can be determined.

[0164] The mass conservation law-based boundary conditions in GREX imaging are extremely useful because, for example, the presence of air can artificially darken voxels, negatively impacting the ability of statistical image registration algorithms to accurately determine the density of objects. By correcting the acquired projection image for differences in air volume and consistently separating quantities that need to be constant throughout the scanning process (e.g., the mass of lung tissue), GREX imaging can be used to generate still image cubes. It generates image reconstructions with higher accuracy than what could be achieved otherwise.

[0165] Considering two GREX projection images acquired at different imaging angles and during the same biologically defined respiratory phase, even though the lung air volume is the same, the way the air displaces tissue may differ between the two projection images. This can cause the brightness of nodules in the affected and erroneous second projection image to be dimmer than in the first projection. As a result of this error, parts of the tissue ("nodules") that were visible in the first projection image may not be visible in the second projection image, ultimately leading to a duller (or misinterpreted as background) appearance of the "nodules" in the resulting image cube. The mass conservation law-based boundary condition in GREX imaging is implemented as a feedback step to ensure that lung mass is conserved between the aforementioned image cube with dimmer nodules and another image cube obtained from a later respiratory phase (which correctly represents the brightness of the "nodules" anatomically). The feedback step in GREX imaging corrects the second projection image of the erroneous image cube during reconstruction by updating the expected shape based on a simulation using the first projection image as the absolute reference. This allows GREX statistical image reconstruction to generate more accurate still image cubes.

[0166] In addition to existing statistical image reconstruction algorithms, GREX post-processing software 106 improves anatomical imaging by incorporating edge definition filters (see 3.2), spatial boundary conditions (see 3.2), and smooth transitions between respiratory phases (described in section 3.3) into digital diagnostic X-ray images. 3.2. Image Filters

[0167] The quality of digital diagnostic X-ray images depends on the settings of the X-ray unit 108 and the anatomical area being imaged. Each patient and anatomical area has a different electron density, and X-rays penetrate this density to generate an image. For example, because the lungs are mainly air and the femur is made of bone, imaging the femur requires more X-ray energy than imaging the chest. Considering that high-energy X-rays penetrate the body more deeply than low-energy X-rays, the number of X-rays emitted from the body to reach the flat-panel detector differs between high-energy and low-energy X-rays, even when imaging the same anatomical shape. Too many X-rays emitted from the body result in overexposure of the flat-panel detector, similar to overexposure in optical photography. If the settings of the X-ray unit 108 are not optimal for the anatomical area being imaged, the image quality will be significantly reduced. In clinical practice, commercial vendors have devised imaging protocols for digital diagnostic X-ray units that roughly estimate the optimal X-ray unit settings for selected anatomical areas. However, the approximation of the optimal tube setting has the drawback of not being customized to account for potentially significant anatomical variations in each patient's anatomical site (for example, the stomach of an obese man is different from that of a man of average weight). In fact, existing imaging protocol settings from vendors are only rough estimates and are not suitable for forming high-quality images.

[0168] When the optimal X-ray unit settings and other imaging parameters are unknown before imaging (as is the case in the current medical field), digital image filters can be strategically used to improve image quality even if the X-ray unit settings are not optimal. Improved images enhance the visibility of anatomical features that are barely visible to the human eye. For example, in the coronal plane, not all ribs may be visible in digital diagnostic X-ray images. Post-processing software 106 filters the coronal image with edge-enhancing filters, such as the Laplacian filter, to display all rib boundaries in the resulting image, even if the rib boundaries in the original image were not visible to the human eye (e.g., a radiologist). Post-processing software 106 overlays the filtered image with the original image, highlighting the previously invisible rib edges that have been enhanced (e.g., filtered) in the original image. Image filters that the user can apply include the Laplacian filter, Hanning filter, Butterworth filter, Parzen filter, and Wi This includes oar filters, Metz filters, Ramp filters, nonlinear spatial average filters, and hybrid filters.

[0169] In some embodiments, post-processing software 106 uses skin surface measurements from the 3D spatial position localizer 300 to calculate optimal imaging parameters for images acquired during each respiratory phase. As the patient breathes, more air is inhaled and the chest circumference increases, changing the body's electron density. The patient's diameter increases, the distance between the patient and the X-ray unit 108 decreases, and the distance between the patient and the detector panel 304 also decreases, resulting in additional image noise in the resulting X-ray images. The 3D spatial position localizer 300 tracks the patient's skin surface position in each image relative to the X-ray unit 108 and the detector panel 304. Tracking the skin surface position can provide unique measurements used in digital diagnostic X-ray examinations.

[0170] Existing digital diagnostic radiology relies on scaling dose index readings measured in an ionization chamber to an estimated patient body mass index. Radiologists currently perform only two measurements: one using a solid water cylindrical phantom equivalent to a 16cm diameter tissue density, and another using a substantially identical phantom with a 32cm diameter. X-ray unit 108 incorporates a "one size fits all (patients)" protocol for various anatomical locations, according to Bender's definition. For example, Bender provides only a single protocol incorporating imaging settings that radiologists can select, regardless of the patient's chest diameter. This means that images are acquired with the same imaging settings whether the man has a thick chest or not.

[0171] The GREX imaging system 100's 3D spatial position localizer 300 generates individual measurements of the patient's chest diameter in real time. Measurement begins when the technician selects a customized X-ray unit setting for the patient. As the patient breathes, the chest diameter changes. When the patient's chest diameter changes, the technician may not be able to set the optimal imaging parameters to match the patient's chest diameter. On the other hand, using real-time and customized chest diameter measurements allows for post-processing to remove image noise and simulate an X-ray image acquired with optimal image parameters.

[0172] Figure 23 is a schematic flowchart of an image filtering process 2300, according to some embodiments of this disclosure, which calculates a noise-free lung image and simulates an image acquired with optimal X-ray unit settings. As described in Chapter 2, after image acquisition, the trachea is identified using a line profile of a segment passing laterally (left to right) through the neck region. The neck consists of muscles, bones, and arteries, but the trachea contains only air and is significantly less dense than the tissue, making it stand out from all other tissues. The line segment indicates the location of air, and a small region containing pixels assigned to air is identified. Image noise in the X-ray image is calculated by dividing the entire image into small patches. Gaussian noise is estimated individually for each patch, and the patch with the least amount of noise is used for texture mapping. In the texture mapping method, the initial texture level of each patch is estimated using a gradient covariance matrix. For the patch with the lowest initial noise level, the noise level is re-estimated in an iterative process that continues until the noise estimate of the patch converges with additional iterations of the gradient covariance matrix. Weak texture patches are assumed to be in air. Air is thought to be located away from the patient, for example, in the upper corner of the X-ray image. Estimating the noise level of the patch gives a baseline noise level for the entire image. Next, the baseline noise level identified by the iterative gradient covariance matrix is ​​subtracted from the entire image to obtain an estimate of the noise-free air density in the trachea.

[0173] The post-processing software 106 estimates the initial position of the lungs using 3D spatial position localization. A human skeletal model (scaled individually for each patient) is overlaid on the surface position estimation provided by Iza 300. The individually scaled patient skeletons are rigidly registered to the X-ray image using visible landmarks present on the skin surface (e.g., clavicle, rotator cuff, scapula, vertebrae). Rigid registration of the skeleton to the X-ray image allows the position of the rib cage to be determined from the skeleton. The rib cage itself provides boundary conditions of pixel values ​​close to the estimates of the lung ends and the unnoised air density of the trachea. Both the lung ends (via the position of the rib cage) and the pixel values ​​close to the estimates of the unnoised air density of the trachea are automatically identified as seed positions by the region formation algorithm (positions from which the region formation algorithm starts and which then grow radially outward). The region formation method is a region-based segmentation method. Segmentation first identifies an initial set of seed points in the image, then examines the neighboring pixels of the initial seed points to determine whether neighboring pixels should be added to the region. This process is repeated in the same way as general data clustering algorithms. In other words, the region formation algorithm uses the initial placement of seed pixels to expand outward using a statistical process of adding "similar" pixels. The region formation algorithm continues until the discriminant pixels are statistically different from the merged clusters (continuing to merge similar pixels).

[0174] In practice, the region formation algorithm in the GREX imaging system "stops" at important anatomical landmark interfaces (e.g., lungs surrounded by thoracic cage enhancement pixels) (e.g., detecting pixel dissimilarity). Pixels that the region formation algorithm does not identify as belonging to lung tissue are masked (masked images are defined as images that enhance structure when subtracted from the original image), forming two separate images: (i) segmented lungs and (ii) other human tissues. To provide accurate and noise-free segmented lung volumes (e.g., lung volumes not visually obscured by non-lung tissues) to improve diagnostic visibility for radiologists, tissues related to body masks (e.g., unrelated to the lungs and therefore containing less visual information) are removed from the lung images. For example, pixels belonging to the intercostal muscles in each simulated imaging ray projection are completely subtracted from the segmented lung images. Body masks are subtracted from the lung images as described above to improve visualization of lung tissue, and then a second validation is performed using the body masks against the patient surface position calculated by the 3D spatial position localizer 300. For example, the post-processing software 106 calculates the number of pixels that the region formation algorithm identifies as a body mask, and then calculates the body diameter at various positions along the height of the torso. This body diameter calculation must closely match the estimate of the 3D spatial position localizer 300 for the patient's body diameter. If it does not, it may indicate that the 3D spatial position localizer 300 needs to be recalibrated to improve accuracy.

[0175] If clinical users desire a more accurate view of the body mask (for example, for clinical or educational reasons), the body mask image is simulated with optimal X-ray unit settings, removing noise morphologies and potential artifact sources from the body mask. The body mask and segmented lungs are then combined again to form artifact- and noise-free X-ray images, further improving overall clinical applications such as structural contouring. 3.3. Biomechanical Modeling

[0176] The biomechanical model used in the post-processing software 106 is based on the first principles of physics, namely the law of conservation of mass and the ideal gas law. The goal of the biomechanical model of the post-processing software 106 is to determine biophysical indicators that enhance the clinical doctor's disease diagnosis ability. The biophysical indicators in question include, but are not limited to, the stress and strain of lung tissue elements.

[0177] Applying a force causes a mechanical system to generate stress. In the case of the lungs, the elements of the mechanical system are composed of lung tissue. The visible lung tissue from medical images is composed of the parenchyma (including alveolar sacs, alveolar walls, bronchi, and blood vessels). The parenchyma is directly involved in lung function. The tissue elements suitable for biomechanical modeling need to be small enough to be homogeneous internally and further statistically stable in response to respiratory stimuli. The general voxel size in the medical imaging range of the lungs is 1mm 3 ~3mm 3 and corresponds to 125 to 375 alveoli. The voxel is considered to contain alveoli with a nearly uniform density and a stable response to respiratory stimuli. The alveoli are arranged in a hexagonal pattern and expand due to the expansion of the normal stress from each shared alveolar wall. From the sum of the normal stresses of the expanding lung tissue elements, an estimated value of the pressure generated by the respiratory stimulus experienced by the alveoli can be obtained. The expansion stress is offset by the reaction stress of the alveolar wall, and when the airflow through the tissue element is 0, the two stresses are in equilibrium. When the position of the lung tissue element moves spatially in an arbitrary direction due to a change in the tidal volume of the lungs, it can be modeled by the response of the tissue element to the normal stress applied to the corresponding surface of the lung tissue element. In other words, the biomechanical model includes a vector term that describes the response of the lung tissue element to an increase in tidal volume, and the vector term is related to the normal stress.

[0178] Strain is defined as the response of a mechanical system to stress. From the perspective of an element (e.g., tissue), stress is a deformable force, and strain is a restorative force. The stress vector on the surface of lung tissue includes two elements: (i) normal stress (related to the outward or inward movement of lung tissue elements resulting in expansion or contraction, respectively), and (ii) shear stress, which is orthogonal to the normal stress and arises from pressure imbalances induced by airflow. By definition, the orthogonal component of shear stress does not contribute to changes in lung volume. The relationship between normal stress / shear stress and tidal volume / airflow is illustrated in the "Airflow-Tidal Volume Relationship" plot shown in Figure 10. Exhalation corresponds to the lowest tidal volume where the airflow is 0 (left end of the curve). During inhalation, the airflow is positive and the tidal volume is increasing (top of the curve). As the airflow slows, the tidal volume is maximum. As tidal volume approaches its maximum (right end of the curve), airflow decreases sharply, and the amount of air expelled from the lungs begins to increase. As the lungs expel more air, tidal volume also begins to decrease. Tidal volume decreases when airflow is negative and continues to decrease until tidal volume reaches the remaining capacity and airflow becomes zero (e.g., exhalation). In short, the motion of lung tissue elements is defined by tidal volume and airflow respiratory stimulation. The motion of lung tissue elements is due to the normal stress and shear stress acting on the elements. Changes in the shape of lung tissue elements (e.g., compression and extension) are modeled by strain, not stress.

[0179] Medical analysis using models relies on the quantitative analysis of forces, thus enabling new diagnoses in the medical field. Because it allows for the clinical visualization and analysis of the functional behavior of the lungs (e.g., the movement of lung tissue), lung health is essentially assessed by the balance of stress and strain forces in each voxel of the lung.

[0180] The first step in model building is to identify which tissues are part of the lung and which are not. For this purpose, the GREX imaging technique acquires several snapshots of the chest shape in multiple configurations (six different respiratory phases). The lungs are segmented based on a region formation segmentation algorithm that uses the air density of the trachea as the initial starting point for the region formation algorithm. The GREX imaging system 100 registers two deformable images to build a biomechanical model. One acts on the lung tissue, and the other on non-lung tissues (e.g., chest wall, ribs, liver, heart, trachea, esophagus, etc.). The results obtained from the region formation algorithm distinguish between lung and non-lung tissues before registration of the deformable images.

[0181] The GREX imaging system, which uses two deformable registrations—one for the lungs and one for other tissues—is based on the fact that the motion and material properties of the lungs differ from those of other tissues. The deformable image registrations are for the lungs and other tissues as a whole. When only one registration is used, the registration assigns a greater weight to other tissues than to lung tissue, meaning it assigns an unrealistically low weight to lung tissue. Consequently, the range of motion becomes unrealistically limited. GREX imaging technology was built to visually elucidate insidious lung motion, such as insidious motion at the lung surface, which is computationally complex.

[0182] Optical flow algorithms are examples of deformable image registration algorithms that can track tissue motion in images. Figure 27 is a schematic flowchart of a multi-resolution 3D optical flow algorithm 2700 according to some embodiments of this disclosure. Specifically, the 3D multi-resolution optical flow algorithm performs deformable image registration that identifies structures in two different images based on the brightness, contrast, or both of each structure. Considering that non-lung tissues include anatomical structures (e.g., ribs, pectoral muscles, sternum, etc.) that are brighter (e.g., at least 10 times denser) than lung tissue, applying a single 3D multi-resolution optical flow algorithm to both non-lung and lung regions prevents the algorithm from artificially prioritizing the less dense lung tissue. The algorithm's computational resources are preferentially allocated to bright, contrasting structures (e.g., ribs, etc.) present in non-lung tissue. Since the allocation of computational resources preferentially ignores the dynamics of lung tissue, this directly violates the objective of GREX imaging, which is to visually elucidate the insidious motion of lung tissue.

[0183] Because prioritizing resource allocation to high-contrast structures is a characteristic of 3D multi-resolution optical flow algorithms, the deformable image registration work in the chest is divided into two auxiliary tasks (e.g., two internally more homogeneous regions): (i) image registration of the lungs and (ii) image registration of non-lung tissues. In some embodiments, to analyze two separate work regions from the chest as a whole, identification of the lung surface (e.g., the boundary between the lungs and non-lung tissues) is required before 3D multi-resolution optical flow or deformable image registration. The lung surface can be identified by a region formation algorithm that starts from the inside of the lung in an air-filled (e.g., visually dark) region, expands outward towards the lung surface boundary, and encounters high-pixel contrast at the lung surface boundary. First, the region formation algorithm is executed.

[0184] The lungs are not attached to the chest wall. Therefore, lung movement is relatively independent of chest movement. In other words, other tissue dynamics are at work, rather than a predictable push-pull response at the lung surface boundary. For example, because horizontally adjacent chest voxels move vertically downward, lung voxels move horizontally into the space where the chest voxels previously occupied.

[0185] To accurately model complex motion dynamics, the GREX imaging system 100 quantifies the shear force acting on the surfactant layer at the lung surface boundary. Two separate segmentations are applied to the lung tissue and non-lung tissue to obtain the basis data for force estimation. The force estimation procedure can be performed by subtracting the segmented lung pixels from the X-ray image (e.g., deleting the assigned values) (performed for each reconstructed respiratory phase image). The pre-segmented lungs are masked from the original image, which provides an image containing all other tissues. The segmented chest shape from each image needs to be aligned with the chest shape of each image in other images to determine the position of each lung tissue element in all images. The multi-resolution optical flow algorithm calculates a displacement vector field showing the displacement of all pixels in two images acquired at different respiratory phases and performs image registration. In fact, knowing the displacement vector field allows for a highly accurate description of the spatial position of all lung tissue elements in the chest shape. The difference between the displacement vector field of segmented lung registration and the displacement vector field of non-lung tissues reveals the magnitude and direction of shear forces in the lungs and chest wall.

[0186] Observations of over 150 patients revealed a linear relationship between displacement and tidal volume. The relationship between displacement and airflow is also linear. The output of the multi-resolution optical flow algorithm is a displacement vector in coordinate space for each measured tidal volume and airflow magnitude. Calculating the displacement vector for all respiratory phases yields a closed-loop trajectory (see Figure 10), which corresponds to the observed values ​​in the biomechanical model. As will be explained in more detail below, the parameters of the biomechanical model (which may include parameters representing normal stress related to tidal volume, normal stress related to airflow, and shear stress related to airflow) are obtained by individually applying QR decomposition to each lung tissue element. The parameters are specific to each lung tissue element (e.g., each lung tissue element has its own unique solution) and collectively describe the response of the lung tissue elements to respiratory stimuli. The parameters of the biomechanical model are vectors that are globally scaled by the measured tidal volume and airflow (e.g., tidal volume and airflow are scalar values). The relationship between biomechanical model parameters, such as the angle between two (vectorized) parameters, is useful for diagnosing the latent incidence of lung disease. Based on the intrinsic biomechanical model vector parameters of lung tissue elements (e.g., each part of the tissue has different vector parameters), the displacement of lung tissue elements is scaled to the chest shape. The chest shape is defined by tidal volume and airflow measurements. The biomechanical model can be calculated for each patient or shared across multiple patients.

[0187] In some embodiments, the biomechanical model describes the movement of lung tissue as the tidal volume (T) of the lung. v ), airflow (A f ), heart rate phase (H c It approximates a function of multiple factors, including ). These values ​​are global values. For example, the heart rate phase is the same for all tissue elements in the chest. The global values ​​are treated as scalar numbers and are measured by the hardware described in Chapter 1. Tidal volume, airflow, and heart rate phase are all measurements that change over time. The intrinsic stress and strain values ​​for each tissue element are given by vector p1 → , p2 → p3 → p4 →It can be expressed using the following formula.

number

[0188] Figure 24 is an example of a closed-loop lung tissue motion trajectory of a tissue fragment located in the left lung near the heart, partially resulting from the heart-lung interaction, according to some embodiments of the present disclosure. The wave behavior illustrated in Figure 24 arises from the heart-lung interaction. Figure 24 is a schematic block diagram illustrating a method for determining the position of a lung tissue fragment moving within a closed-loop lung tissue motion trajectory based on the biomechanical model described above, according to some embodiments of the present disclosure. Figure 24 illustrates a method for calculating the displacement of a single tissue element from the origin to an arbitrary position on the closed-loop trajectory by summing the three vectors of the biomechanical model described by the above equations.

[0189] Interpolating images during the acquired respiratory phase using a biomechanical model based on physiology offers the significant advantage of verifying the accuracy of the biomechanical model's output using quantitative physical quantities. According to the ideal gas law, the ratio of lung volume change to tidal volume is 1.11 at room temperature. In other words, the ratio of room air density to lung air density is 1.11. Therefore, the volume integral of the divergence of the normal stress vector must also be 1.11 (for example,

number

[0190] An example of how biomechanical modeling in GREX imaging technology enhances diagnosis is early-stage lung tumors that are invisible to the radiologist during imaging. Because the tumor is much smaller due to image sensitivity, it is invisible to the radiologist. Even though the tumor is invisible to the radiologist, its electron density is greater than that of healthy lung tissue, and its presence can affect the balance of forces within the lung. A higher electron density in the tumor means that differences in the tumor's material and mechanical properties (e.g., different characteristic stress and strain parameters) will affect both the tumor's motion and the motion of the surrounding local area (e.g., healthy tissue near the tumor). The effect of the tumor on local healthy lung tissue can be roughly analogous to the effect of mass on the spacetime continuum under general relativity. When a massive object exists, spacetime around it is curved, and light around the object behaves differently than it would if its mass were absent. The same analogy applies to the lung; the tumor distorts the trajectory of adjacent healthy tissue, resulting in different motion than that of healthy lung tissue. From the displacement vector map of the biomechanical model, physicians can immediately grasp changes in the composition and biomechanical properties of lung tissue. That is, if a tumor is pre-defined, the displacement vector field will show unnatural vector curls and other altered properties. The creation of parameter maps using GREX imaging post-processing software 106 (see 3.4) visually displays important diagnostic information that was previously invisible to the user. GREX imaging parameter maps are an example of the new diagnostic advancements the GREX platform is bringing to the medical field.

[0191] Figure 25 is a schematic flowchart illustrating the operation of components of a biomechanical model according to some embodiments of this disclosure.

[0192] The biomechanical model process uses quantitative means to biologically interpolate two images (2D or 3D) acquired at different respiratory phases. Equation (2) shows that the solution to the biomechanical model is the displacement of the two respiratory phases (U → -U0 → ). As mentioned above, the displacement between the two respiratory phases (U → -U0 → ) is detected by performing deformable image registration and cross-indexing the two respiratory phases. This typically consists of the following three steps: Step 1: Perform region-based segmentation to draw the structural boundaries between the lungs and non-lung tissues. Step 2: Perform intensity-based structure mapping using 3D multi-resolution optical flow to match the "same" structures in the two image cubes. Step 3: In some embodiments, an initial estimate of the displacement between two respiratory phases (U → -U0 → ) is repeatedly improved until it is optimized.

[0193] Figure 26A illustrates the motion trajectory of lung tissue elements during a respiratory cycle according to some embodiments of the present disclosure. Figure 26A contains six images at different image angles, corresponding from left to right to each of the six respiratory phases of a complete respiratory cycle, as illustrated in Figure 14: early inspiratory phase (EI), late inspiratory phase (LI), maximum inspiratory phase (MI), early expiratory phase (EE), late expiratory phase (LE), and maximum expiratory phase (ME). According to finite strain theory, the vector connecting the positions of particles in an undeformed configuration and a deformed configuration is called the particle displacement vector. Using a specific voxel within the EI image cube as a reference, the six images show the deformation of the reference voxel in its original position, size, and shape during the respiratory cycle. As air enters the lungs, the reference voxel begins to "bubble." In other words, the lung tissue corresponding to the reference voxel deforms as the voxel "bubbles." The displacement of the two respiratory phases (U → -U0 →This quantifies the degree of voxel foaming and corresponding lung tissue deformation. Deformable image registration assumes that reference voxels in the EI image cube move to new positions in the LI image cube, and simultaneously deform due to more air inhalation into the lung tissue. As the respiratory cycle progresses... The reference voxels maintain their motion trajectories, as illustrated by the MI, EE, LE, and ME image cubes, respectively. In other words, for every voxel in the reference image cube, a set of displacement vectors is calculated across all six image cubes.

[0194] Figure 26B illustrates that there are five deformable registrations (2→1, 3→1, 4→1, 5→1, 6→1) between the EI image cube and the other five image cubes. These six image cubes represent the anatomical structures of the patient's chest at the corresponding six predefined respiratory phases. For each voxel, a vector (U) is generated in the image cube corresponding to each of the six respiratory phases. → ) exists. Assuming the vector corresponding to the EI image cube is zero, the six displacement vectors of a particular voxel are:

number

[0195] Similarly, for a specific respiratory phase "n", the tidal volume of the lung (T v ), airflow (A f ), heart rate phase (H c The biometric data matrix, which includes ),

number

[0196] For each voxel in the image cube, the biomechanical model of the parameter matrix.

number

number

number

number

[0197] As described above, there are many deformable image registration algorithms that can perform image registration of GREX imaging, including the 3D multi-resolution optical flow algorithm. The 3D multi-resolution optical flow algorithm calculates a smooth (e.g., fluid-like) transition between images taken at various observed tidal volumes. The displacement (U) calculated between two respiratory phases → -U0 → ) will be the observation point in the tissue trajectory shown in Figure 24. The parameters of the biomechanical model (p1 → , p2 → p3 → p4 → For example, if we solve this using least squares regression, we get the tidal volume (T v ), airflow (A f ), heart rate phase (H c The ) changes, generating the entire closed-loop trajectory. Obtaining the closed-loop trajectories of all tissue elements in the chest yields a new image cube that is considered bio-interpolated. Table 1 summarizes each component of GREX's biomechanical model and how it is detected. [Table 1]

[0198] Biointerpolation is performed on the acquired image cubes using post-processing software 106 to input data for all potential respiratory phases and create a complete video of the chest. Generally, the video requires at least 30 simulated images for smooth transitions between frames. Figure 28A is a schematic flowchart showing video creation 2800 using biointerpolation according to some embodiments of the present disclosure. As illustrated in Figure 28B, the biodata matrix corresponding to the EI image cube is T v =20 ml, (A f ) = 20 ml / s, H c It is assumed to be = 0.10.

[0199] Each voxel in the intermediate simulation image cube represents a specific moment in the respiratory cycle, e.g., EI+Δt, EI+2Δt, EI+3Δt, etc., and corresponds to the biomechanical model of the parameter matrix [P → This can be calculated using the biometric data matrix corresponding to a specific moment in time. 3.4. Parameter Map

[0200] One of the clinical advantages of GREX imaging technology is its unique parameter mapping. GREX imaging technology uses 2D plots, 2D color washes, 3D plots, and 3D vector field maps to present users with previously unavailable patient chest health information.

[0201] Figure 29 shows examples of standard radiographs of a healthy individual (left) and a patient with a stage 1b left upper lung tumor (right, indicated by the arrow). Both represent standard radiographs used in the current field of radiography. The affected left lung on the right side of Figure 29 is an early-stage lung tumor, which is difficult to detect in a standard radiograph because it is difficult to read and only displays anatomical information. In contrast, Figure 30 shows an accompanying GREX parameter map showing health status indicators for two of the same patients, according to one embodiment of this disclosure. The parameter map on the right clearly shows that the patient has disease in the left upper lung, whereas the standard radiograph is ambiguous. By changing the frame level and image contrast, the area of ​​poor ventilation in the left upper lung is eventually revealed. However, if the user does not follow these steps, the likelihood of detecting this early-stage lung cancer tumor is low. In other words, parameter maps generated by GREX can significantly reduce the risk of the user missing a disease.

[0202] First, let's look at p1 in Figure 30, which represents the normal stress related to the tidal volume. → Let us consider the 2D color map. p1 → Based on the definition, the user is p1 → It can be expected that the displacement near the diaphragm will be larger than that at the apex of the lung (for example, as tidal volume increases, the tissue displacement of the diaphragm will be larger than that at the apex of the lung). Generally speaking, p1 → The size changes smoothly throughout a healthy lung. p1 → If the gradient is calculated over the entire lung, then the smooth function is Obtainable. According to the example illustrated in Figure 30, for both patients, p1 → Even with similar parameter distributions, healthy individuals exhibit twice the tissue displacement of patients.

[0203] Next, we represent the sum of normal stress and shear stress related to airflow, (p2 → +p3 →Consider a 2D color wash of the following. Generally speaking, parameters are higher near the area where the bronchial trees take in air into the lungs at a faster rate (the center of the lung). However, the presence of a tumor in an affected lung substantially alters the elastic behavior of the lung, potentially making the tumor visually distinguishable from the functionally healthy lung tissue distribution. In the example illustrated in Figure 30, the presence of a lung tumor is clearly visible in the left lung due to its significantly different size. When disease affects a lung area, other lung tissue areas "take care of the slack" by ventilating the healthy area rather than the diseased area. However, as illustrated in the left-hand diagram of Figure 29, a healthy lung is more elastic than a diseased lung. In other words, airflow resistance increases significantly at the location of the tumor, highlighting its presence and allowing for a quantitative analysis of the tumor's effect on the patient's ability to ventilate properly during respiration. The ratio of parameters related to tidal volume to parameters related to airflow clearly shows that the upper left lung has a considerably different pattern from that of a healthy person. Physically, a higher percentage can be interpreted as the tissue moving in a more circular pattern than the example of tissue trajectory illustrated in Figure 24. Tidal volume (T) is shown below the 2D color wash. v p1 → ) and airflow (A f (p2 → +p3 → The observed total motion component histogram, resulting from )), also shows a clear indicator of disease in the left lung. A healthy lung has a bimodal distribution A. f (p2 → +p3 → ) has. However, if one lung is diseased, the overall distribution pattern differs between the two lungs. GREX imaging techniques include the Bayes process, which leverages patient history (information voluntarily provided by the patient before the examination) and the distribution of parameter histograms of biomechanical models to better classify the pathogenesis of the disease. In summary, consistent indicators of lung disease for multiple new GREX parameters (e.g., the presence of tumors, as shown in Figure 30) are of great help in early detection and disease diagnosis. 3.5. Diagnostic Disease Pointers

[0204] Lung cancer, chronic obstructive pulmonary disease (COPD), lower respiratory tract infections, and tuberculosis all have disease pointers that can be identified in existing digital diagnostic X-ray images. However, diagnostic disease pointers are not always displayed at the early stages of the disease. However, as shown in FIGS. 29 and 30, GREX parameters can serve as diagnostic disease pointers at the early stages of the disease. If the medical team only examines standard X-ray photographs, the tumors of lung cancer are not large enough to be clearly displayed in the standard X-ray photographs, so the presence of lung cancer may not be detected. GREX imaging is used in combination with multiple targeted radiographs based on biological data of biological signals, and utilizes previously unavailable information regarding lung movement to diagnose diseases. In the GREX process, it can directly address early-stage lung cancer, which tends to have low specificity in digital diagnostic X-ray images. The existing sensitivity limitations of digital diagnostic X-ray-based diagnosis are addressed by imaging based on biological information that uses patient-specific respiratory information indicating the disease. Clinicians can diagnose at an early stage of the onset of the disease and continue to monitor patients according to the progression of the disease through visualization of lung tissue movement and parameter maps. By applying machine learning and finite element analysis techniques to find latent patterns and changes in baseline of lung function, users can detect the disease at the time of onset, increase the survival rate of patients, and expand treatment options.

[0205] Another example of how GREX brings ripple effects to diagnostic methods is the classification of the etiology of lung diseases. In biopsies, the symptoms of lung diseases caused by asbestos and those caused by tobacco are different. Asbestos is a natural silicate mineral composed of long, thin fibrous crystals made up of millions of fine fibers. Inhaled asbestos fibers penetrate the alveoli and ultimately form a dense network, impairing alveolar function and reducing lung function. The dense network collects cancerous tissue and is called mesothelioma. In standard X-ray photographs, mesothelioma can only be identified when plaques have accumulated. The accumulation of plaques appears as high-density sclerosis (blurring) in the image of the lung. GREX allows the user to Before visually identifying the presence of a vest, latent asbestos may be detected through a parameter map. When asbestos forms a network, the elasticity of the lungs decreases locally. A local and slight decrease in elasticity is represented in the parameter map (2D color wash, parameter ratio, histogram). For example, in tissue that has lost elasticity, the trajectory of the tissue during breathing becomes more circular than elliptical. This means that the ratio of the size of p1 → to (p2 → + p3 → ) is higher than that of healthy tissue. Referring to the color map, the area of the lungs showing a pattern indicating the disease can be clearly seen.

[0206] Inhaled particles (non-fibrous) are completely different from fibrous asbestos particles. Inhaled particles deposit in the lungs, "block" the airways, form scar tissue, form tumors instead of a network, and reduce lung function. GREX indicates that particle deposition has started by detecting minute local changes in the dynamics of lung movement that are inconsistent with healthy characteristics. Tobacco contains tar and radon that adhere to the alveoli and dramatically change lung function. GREX provides a unique tool that can track the decline in lung function and show users the destructive power of smoking habits to patients. COPD is another disease that can be more effectively elucidated with a parameter map than current clinical methods. COPD is currently detected by spirometry tests and standard X-rays. These methods are not particularly sensitive and cannot detect COPD early. The earlier a disease is detected, the more opportunities there are to provide preventive treatment before it's too late and give patients a chance to change bad habits. 3.6. Video Presentation

[0207] The advantage of a graphical user interface (GUI) for video presentations is that it provides a clean and intuitive canvas for displaying the results of both 2D and 3D videos. Features such as rotation, video pause, and rulers are located in bins represented by tabs at the top of the screen. When a feature is selected, the tooltip changes, indicating the feature being used to the user. The GUI is designed to be lightweight, so it works without problems even on older computer systems. Finally, the video presentation GUI works in conjunction with annotations / outlines, possible disease pointers, and cloud computing platforms.

[0208] Healthcare professionals prefer to highlight and add notes directly to medical images rather than attaching additional documentation. They require imaging capabilities such as image annotation ("attention" arrows and treatment notes) and anatomical contouring (drawing virtual lines around anatomical structures). These users are often part of large care teams that collaboratively treat patients. The following example illustrates the current workflow and technical capabilities of a clinical care team.

[0209] When imaging tests are performed on lung cancer patients, radiologists interpret the images, which then provide instructions for radiation therapy to radiation oncologists and form the basis for surgical intervention by surgeons. Specifically, radiologists outline the tumor area and provide the outlined image to the radiation oncologist.

[0210] Radiation oncologists write instructions for radiation therapy and / or tumor resection surgery based on imaging.

[0211] Based on the outline drawn by the radiologist and the instructions for tumor resection from the radiation oncologist, the surgeon removes the tumor.

[0212] Post-processing software 106 provides users with tools to directly annotate and outline videos, making healthcare workflows more streamlined. This user-friendly feature enhances healthcare workflows and reduces the potential for medical errors. The interactive nature of the user interface ensures that treatment instructions and medical concerns are clearly visible to all users, and clinical notes are correctly displayed in their corresponding anatomical areas of interest.

[0213] Figures 31A-31B are flowcharts of a patient's lung imaging method 3100. In some embodiments of this disclosure, some or all of the operations described below can be performed without human intervention (e.g., without intervention by a technician). In some embodiments, method 3100 is performed by any of the apparatus described herein (e.g., the GREX imaging system 100 shown in Figure 1). Some operations of method 3100 are performed by a computer system including one or more processors and a memory that stores instructions causing one or more processors to perform the operations of method 3100 when performed by one or more processors. Some operations of method 3100 may be combined in any order and in any order.

[0214] The method includes positioning the patient in a first orientation relative to the X-ray imaging device (e.g., the GREX imaging system in Figure 1) (3102). In some embodiments, method 3100 is performed on an existing medical imaging system modified to perform specific operations described below. In some embodiments, positioning the patient in a first orientation includes moving (e.g., rotating) the patient in a first orientation (see, for example, Figure 35), while the X-ray imaging system (e.g., the X-ray unit and detector panel) remains in a fixed position. For example, in some embodiments, the patient is seated or standing in a patient positioning frame (PPF), such as the PFF3501 described with reference to Figure 35. In some embodiments, positioning the patient in a first orientation relative to the X-ray imaging device includes rotating the patient positioning frame. For example, in some embodiments, at the start of method 3100, the patient positioning frame is rotated so that the patient's sagittal plane is positioned at a predetermined angle with respect to the optical axis of the X-ray imaging device (e.g., the axis through which the X-ray imaging device transmits X-rays). In some embodiments, the default angle is selected from the group consisting of -45 degrees, -22.5 degrees, 0 degrees, 22.5 degrees, and 45 degrees.

[0215] In some embodiments, positioning the patient in a first orientation involves moving (e.g., rotating) the X-ray imaging system, while the patient remains in a fixed position (see, for example, Figure 17).

[0216] In some embodiments, while the patient is positioned in a first orientation relative to the X-ray imaging device, the patient positioning device maintains the patient in a fixed position (e.g., stabilizes the patient's position), and a 3D image of the patient's lungs can be reconstructed based on the assumption of a “stationary object” (see, for example, this specification).

[0217] The method includes obtaining a patient respiratory volume measurement (3104) (for example, the patient is breathing normally and does not need to hold their breath). In some embodiments, the patient respiratory volume measurement is a measurement of the patient's lung volume (e.g., instantaneous lung volume) (e.g., direct measurement) or a derivative of the patient's lung volume (e.g., flow rate). In some embodiments, as will be described in more detail below, the patient respiratory volume measurement is a measurement that can be converted to the patient's tidal volume (e.g., measuring the rise and fall of the chest). As used herein, tidal volume means the difference between the current lung volume and a predetermined baseline (e.g., the amount exhaled maximally during normal breathing without extra effort, the amount inhaled maximally during normal breathing without extra effort, or other baseline volumes).

[0218] In some embodiments, the patient's respiratory volume measurement is a geometric (spatial or positional) measurement of the patient's respiration.

[0219] In some embodiments, the patient's respiratory volume measurement includes the patient's chest elevation (and / or chest depression) measurement (3106). In some embodiments, the patient's respiratory volume measurement is obtained using one or more volume respiratory phase sensors (e.g., respiratory phase sensor 110 in Figure 1) from the group consisting of a 3D scanner, a spirometer, and an abdominal belt (3108). In some embodiments, The method further includes creating a point cloud on the surface of the patient's chest (3110). The patient's respiratory volume measurement is determined from the point cloud on the surface of the patient's chest.

[0220] In some embodiments, the 3D point cloud is used to determine the patient's respiratory volume measurement without generating a mesh reconstruction of the patient's chest (e.g., the raw output of the 3D point cloud is used to generate the patient's respiratory volume measurement without generating a mesh). In some embodiments, the point cloud on the surface of the patient's chest is acquired using 3D imaging techniques to measure one or more locations on the patient's chest (3112). For example, 3D imaging techniques include laser scanning techniques such as light detection / distancing (LIDAR). Laser scanning techniques are very beneficial because some lasers can accurately measure the position of the patient's chest even when the patient is wearing clothing (e.g., the patient wears LIDAR-penetrating clothing while undergoing method 3100).

[0221] In some embodiments, method 3100 further includes identifying one or more anatomical landmarks on the surface of the patient's chest (e.g., externally visible landmarks such as the clavicle, trachea, sternal notch, sternum, xiphoid process, and spine) using a point cloud on the surface of the patient's chest (3114). The method further includes inferring the location of one or more internal anatomical landmarks within the patient's chest (e.g., the location of the patient's lungs) from the point cloud on the surface of the patient's chest. In some embodiments, the method includes generating a reconstruction of the patient (e.g., generating a computer model for the patient, also called a “virtual patient”). In some embodiments, generating a reconstruction of the patient includes generating a reconstruction of the patient’s internal anatomical structure. In some embodiments, the reconstruction of the patient’s internal anatomical structure includes a computer model of tissue density (e.g., a 3D model). In some embodiments, the reconstruction includes or is used to determine absorption cross-sections (e.g., X-ray absorption cross-sections) at multiple locations within the body. In some embodiments, the patient reconstruction is used to determine the X-ray dose delivered by the X-ray imaging device (e.g., for each image).

[0222] The method further includes determining the patient's respiratory period from the patient's volumetric measurements while the patient is positioned in a first orientation relative to the radiation source and while obtaining the patient's respiratory measurements (3116) (for example, in real time when the patient is breathing normally). In some embodiments, the patient's respiratory period is defined by the volume of the lungs. Thus, in some embodiments, determining the respiratory period of the lungs includes determining the volume of the lungs.

[0223] In some embodiments, operations 3102 onward are performed as part of the imaging period of Method 3100. Method 3100 further includes providing a training period prior to the imaging period in which information about the patient's normal respiration is obtained. For example, during the training period, the patient's respiratory volume measurements are acquired at regular intervals over multiple respiratory cycles of the patient (e.g., 15, 20, and 50 cycles of the patient's respiratory cycle are considered to correspond to one breath). The volume measurements from the training period are then used to associate specific volume measurements with specific respiratory phases. For example, a volume measurement corresponding to a tidal volume of 400 ml is associated with the maximum inspiratory phase, and a volume measurement corresponding to a tidal volume of 0 ml is associated with the maximum expiratory phase. Furthermore, statistics on the patient's respiration (e.g., a histogram) may be acquired during the training period and used to confirm that the respiration acquired during imaging is "normal" respiration (e.g., no deep breaths or abnormal respiration).

[0224] During the imaging period, in some embodiments, the determined (3118) patient respiratory period is a future (e.g., predicted) respiratory period. That is, in some embodiments, determining the patient's respiratory period based on the patient's respiratory volume measurements involves predicting a future respiratory period from one or more current and / or past respiratory periods. For example, the prediction is based on a time series of respiratory periods. In some embodiments, predicting a future respiratory period from one or more current and / or past respiratory periods Prediction involves generating an autoregressive integrated moving average (ARIMA) model. In some embodiments, prediction uses data from the training period.

[0225] When it is determined that the respiratory phase of the patient matches a predetermined respiratory phase (for example, the lung volume matches a predetermined lung volume), the method includes performing gate control of the X-ray imaging device and generating an X-ray projection image of the patient's lungs (which may also be referred to as a projection image) (3120). In some embodiments, the X-ray projection image is an image taken at a specific angle (for example, determined by the orientation of the patient relative to the X-ray imaging device). In some embodiments, the X-ray projection image is obtained using a single X-ray exposure. In some embodiments, the method includes abandoning the gate control of the X-ray imaging device (for example, abandoning exposing the patient to X-ray radiation) when it is determined that the respiratory phase of the patient does not match the predetermined respiratory phase.

[0226] In some embodiments, the method includes determining whether the current respiration is irregular from the patient's respiratory volume measurement. The method further includes abandoning the gate control (for example, acquisition) of the X-ray imaging device (for example, continuing to wait for a suitable respiration for acquiring an X-ray projection image with respect to the respiratory phase) when it is determined that the current respiration is irregular.

[0227] In some embodiments, the predetermined respiratory phase is the first predetermined respiratory phase among a plurality of predetermined respiratory phases (3122). In some embodiments, the method further includes performing gate control of the X-ray imaging device and generating X-ray projection images of the patient's lungs respectively when it is determined that the respiratory phase of the patient matches any of the plurality of predetermined respiratory phases while acquiring the patient's respiratory volume measurement. In some embodiments, the X-ray projection image of the patient's lungs (for example, X-ray measurement value) is acquired only when the respiratory phase of the patient determined by the patient's respiratory volume measurement matches any one of the plurality of predetermined respiratory phases (3124). Accordingly, the total amount of X-ray exposure of the patient is reduced.

[0228] In some embodiments, the multiple predefined respiratory phases include the early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase in the patient's complete respiratory cycle (3216) (see, for example, Figure 14). In some embodiments, an X-ray projection image is obtained for each of the multiple predefined respiratory phases while the patient is positioned in a first orientation relative to the X-ray imaging device.

[0229] In some embodiments, the X-ray projection image is a first X-ray projection image (3128), and the method further includes repositioning the patient to a second orientation relative to the X-ray imaging device (e.g., rotating the patient or the X-ray imaging system). In some embodiments, the method includes continuing to determine the patient's respiratory phase based on the patient's respiratory volume measurements while the patient is positioned to the second orientation relative to the X-ray imaging device and while continuing to acquire the patient's respiratory volume measurements. In some embodiments, the method includes gating the X-ray imaging device to generate a second X-ray projection image of the patient's lungs when it is determined that the patient's respiratory phase coincides with a predetermined respiratory phase.

[0230] In some embodiments, the method further includes generating a still image cube corresponding to a predetermined respiratory period using a first and a second X-ray projection image (see, for example, Figure 21). In some embodiments, the still image cube is a three-dimensional reconstruction of the volume of the patient's lungs. In some embodiments, the X-ray projection image is obtained for each of a plurality of predetermined respiratory periods while the patient is positioned in each of a plurality of orientations (including a first and a second orientation) relative to the X-ray imaging device. In some embodiments, the plurality of orientations include at least five orientations (e.g., -45 degrees, -22.5 degrees, 0 degrees, 22.5 degrees, 45 degrees). In some embodiments, the plurality of orientations include more than five orientations (e.g., six, seven, eight, or more orientations). In some embodiments, the patient is positioned in each of the X-ray imaging devices While positioned in multiple orientations (including the first and second orientations), no X-ray projection images will be acquired except during multiple predetermined respiratory periods. Therefore, since X-ray projection images are acquired for five orientations and six different respiratory periods, a total of 30 X-ray projection images will be acquired (for example, the X-ray imaging device will acquire only these images through gate control as described above).

[0231] As already described herein, these 30 X-ray projection images can be used to reconstruct a 3D biomechanical model of lung movement. In some embodiments, a static image cube corresponding to a predetermined respiratory period (see, for example, Figure 21) is generated from fewer than 10 X-ray projection images acquired from various angles during the predetermined respiratory period (3130).

[0232] Those skilled in the art will understand that Method 3100 can also be applied to movements other than lung movements due to respiration. For example, in some embodiments, the Method includes positioning a patient in a first orientation relative to an X-ray imaging device. The Method further includes obtaining 3D measurements of parts of the patient's body (e.g., 3D measurements of the position of the patient's body parts). The Method further includes determining whether a trigger criterion for triggering radiation exposure by the X-ray imaging device is met based on the 3D measurements of the patient's body parts while the patient is positioned in the first orientation relative to the X-ray imaging device and while the 3D measurements of the patient's body parts are being obtained, and if it is determined that the trigger criterion is met, gate control of the X-ray imaging device to generate an X-ray image of the patient. In some embodiments, the X-ray image is an image of a part of the patient's body (e.g., the patient's legs, abdomen, skull, etc.). In some embodiments, the trigger criterion includes a criterion that is met when the 3D measurements of the patient's body parts indicate that the patient's body parts are in a predetermined position (e.g., position relative to the imaging device). In some embodiments, the method includes abandoning gate control of the X-ray imaging device (for example, abandoning exposure of the patient to X-rays) if it is determined that the trigger criteria are not met. Furthermore, in some embodiments, method 3100 can also be applied to other imaging methods that are not strictly X-ray based, such as positron emission tomography (PET) imaging and MRI imaging.

[0233] Furthermore, Method 3100 can also be applied to radiotherapy and radioimaging. For example, in some embodiments, the method includes positioning a patient in a first orientation relative to a radiotherapy source. The method further includes obtaining 3D measurements of the patient's body parts (e.g., 3D measurements of the positions of the patient's body parts). The method further includes determining, based on the 3D measurements of the patient's body parts, whether a trigger criterion for triggering radiation exposure by the radiotherapy source is met while the patient is positioned in the first orientation relative to the radiotherapy source and while the 3D measurements of the patient's body parts are being obtained, and, if the trigger criterion is determined to be met, performing gate control of the radiotherapy source to expose the patient to radiation (e.g., exposing the patient's body parts to radiation). In some embodiments, the trigger criterion includes a criterion that is met when the 3D measurements of the patient's body parts indicate that the patient's body parts are in a predetermined position (e.g., position relative to the imaging device). In some embodiments, the method includes abandoning radiation exposure if the trigger criterion is determined to be not met.

[0234] The specific sequence of operations described in Figures 31A-31B is merely an example, and the described sequence does not mean that it is the only sequence in which the operations can be performed. Those skilled in the art will understand that the operations described herein can be rearranged in various ways. Furthermore, other processes described herein with respect to other methods described herein can also be applied in a manner similar to method 3100 described above with reference to Figures 31A-31B. Such processes are, for example, shown in Figures 2, 6, 8, 11, 13, 15, 21, 23, 25, 27, 28A-28B, 32A-32B, 33A-33C, and 34A-3 This document is written with reference to 4C. For the sake of brevity, detailed explanations will be omitted.

[0235] Figures 32A-32B are flowcharts of a method (3200) for controlling the gate of a radiation source in matching the respiratory and cardiac phases of a patient. In some embodiments, some or all of the operations described below can be performed without human intervention (e.g., without intervention by a technician). In some embodiments, method 3200 is performed by any of the apparatus described herein (e.g., the GREX imaging system 100 shown in Figure 1). Some operations of method 3200 are performed by a computer system including one or more processors and a memory that stores instructions to cause one or more processors to perform the operations of method 3200 when performed by one or more processors. Some operations of method 3200 may be combined in any order and in any order.

[0236] The method includes positioning the patient in a first orientation relative to the radiation source (3202). In some embodiments, the radiation source is an X-ray imaging device (3204). In some embodiments, the radiation source is a radiotherapy source (3206). For example, as illustrated in Figures 3 and 35, the patient (e.g., patient 3502) is positioned in a first position relative to the radiation source 3504 (e.g., X-ray unit 108). In some embodiments, patient positioning includes moving the patient (e.g., rotating), as in operation 3102 (Figure 31A), and in some embodiments, patient positioning includes moving the X-ray apparatus (e.g., X-ray source and detector) (e.g., rotating).

[0237] The method includes obtaining patient respiratory measurements (3208) (e.g., using a respiratory sensor). In some embodiments, the patient respiratory measurement is the patient's respiratory volume measurement in operation 3104 (Figure 31A). In some embodiments, the patient respiratory measurement is not the patient's respiratory volume measurement (e.g., a timing-based patient respiratory measurement).

[0238] The method includes obtaining measurements of the patient's cardiac function (3210). In some embodiments, the patient's cardiac function is measured using one or more sensors (3304). In some embodiments, an electrocardiogram (ECG) is used to measure the patient's cardiac function (e.g., a 3-lead or 12-lead ECG). In some embodiments, the method includes obtaining multiple measurements of the patient's cardiac function, which provide a time-series of electrical signals controlling the movement of the patient's heart.

[0239] Returning to imaging, the method includes determining the patient's respiratory phase from the patient's respiratory measurements (3214) while the patient is positioned in a first orientation relative to the radiation source and while acquiring the patient's respiratory measurements (3212), and determining the patient's heart rate phase (e.g., real-time) from the patient's cardiac function measurements (3216). For example, in some embodiments, the patient's heart rate phase is determined using a zero-instruction-set computer (ZISC) processor, as described with reference to Figure 36. The ZISC processor can identify the occurrence of a predetermined landmark in a cardiac cycle (e.g., an S wave or a T wave) during a single cardiac cycle or a portion of a cardiac cycle.

[0240] In some embodiments, the method further includes obtaining patient cardiac function measurements from multiple cardiac cycles of the patient (3218) before performing gate control of the radiation source to expose the patient to radiation (operation 3222, described later). In some embodiments, the method includes using the patient cardiac function measurements obtained from multiple cardiac cycles to determine the average interval between a predetermined heart rate phase and the start of a predetermined frame of the cardiac cycle. For example, the predetermined frame of the cardiac cycle represents the interval between the top of the R wave and the start of the gate frame. For example, in some embodiments, operations 3202 onward are performed as part of the imaging period of method 3200. Method 3200 further includes providing a training period prior to the imaging period in which information about the patient's cardiac function is obtained. For example, During the training period, the patient's ECG measurements are acquired at regular intervals over multiple cardiac cycles (e.g., 15, 20, and 50 cycles of the patient's cardiac cycle; one cycle corresponds to one complete period of cardiac movement, such as from any T wave to the next). Next, the ECG measurements from the training period are used to predict the sedation period of cardiac movement during the imaging period, as described below.

[0241] In some embodiments, the measurements obtained from multiple cardiac cycles of a patient are waveform measurements (3220) (e.g., ECG measurements) from multiple cardiac cycles, and the method includes verifying the waveform measurements from multiple cardiac cycles as being statistically stable.

[0242] The method further includes gate control (3222) of a radiation source based on the determination that the patient's respiratory phase coincides with a predetermined respiratory phase and that the patient's cardiac phase coincides with a predetermined frame of the cardiac cycle, thereby exposing the patient to radiation. In some embodiments, when the radiation source is gated, the patient's lungs are exposed to radiation. In some embodiments, the predetermined cardiac frame corresponds to a period of sedation of cardiac motion (e.g., a period in the cardiac cycle when cardiac motion is minimal, as described with reference to Figure 12). In some embodiments, the radiation source is gated within the same cardiac cycle as the determined cardiac phase. By thus gate control of the patient's lung exposure based on the coincidence of the respiratory phase and the predetermined cardiac phase frame, a precise area of ​​the lung (e.g., a precise area of ​​lung tissue) is exposed to radiation without perturbation or motion caused by cardiac motion.

[0243] In some embodiments, the radiation source is an X-ray imaging device. Gating the radiation source to expose the patient to radiation includes gateping the X-ray imaging device to generate an X-ray projection image of the patient's lungs (3224). In some embodiments, as in method 3100 of Figures 31A-31B, the X-ray projection image can be acquired for the X-ray imaging source and for multiple orientations of the patient for multiple respiratory phases. These X-ray projection images can be used to create videos of lung motion and / or biophysical models of the lungs (e.g., relating the movement of lung tissue to biophysical parameters such as stress, strain, and elasticity). In some embodiments, according to Method 3200, these images are acquired within a predetermined cardiac frame to minimize lung perturbation or motion due to cardiac movement.

[0244] In some embodiments, the radiation source is a radiotherapy source. Gating the radiation source to expose a patient to radiation involves gateping the radiotherapy source to deliver a therapeutic dose to the patient's lung region (3226). In these situations, it is important to deliver as much radiation as possible to the affected tissue (e.g., cancerous tissue) and as little radiation as possible to healthy tissue. Method 3200 improves the radiotherapy apparatus by precisely delivering a dose to diseased tissue while minimizing the radiation delivered to healthy tissue.

[0245] In some embodiments, determining whether the patient's heart rate phase matches a predetermined frame of the cardiac cycle involves predicting the predetermined frame of the cardiac cycle by detecting a predetermined heart rate phase in real time (for example, using the BEMP card in Figure 36) and waiting for a length of time corresponding to the average interval between the predetermined heart rate phase and the start point of the predetermined frame of the cardiac cycle (3228). For example, in some embodiments, the peak of the T wave is detected, the average interval from the T wave to the ideal gate frame is waited for, and the gate control of the radiation source is performed at that point.

[0246] The specific sequence of operations shown in Figures 32A-32B is merely an example, and the sequence described is not the only sequence in which the operations can be performed. Those skilled in the art will understand that the operations described herein can be rearranged in various ways. Also, refer to Figures 32A-32B for other processes described herein in relation to other methods described herein. This can be applied in a manner similar to method 3100 described above. Such a process is described, for example, with reference to Figures 2, 6, 8, 11, 13, 15, 21, 23, 25, 27, 28A-28B, 31A-31B, 33A-33C, and 34A-34C. For the sake of brevity, detailed explanations are omitted.

[0247] Figures 33A to 33C are flowcharts of a method 3300, according to some embodiments of the present disclosure, for generating a model of the mechanical properties of the lung by fitting data from aligned images. In some embodiments, some or all of the operations described below can be performed without human intervention (e.g., without intervention by a technician). In some embodiments, method 3300 is performed by any of the apparatus described herein (e.g., the GREX imaging system 100 illustrated in Figure 1). Some operations of method 3300 are performed by a computer system including one or more processors and a memory that stores instructions for one or more processors to perform the operations of method 3300 when performed by one or more processors. Some operations of method 3300 may be arbitrarily combined and their order may be arbitrarily changed.

[0248] The method includes extracting multiple displacement fields of lung tissue from multiple X-ray measurements (e.g., X-ray images or X-ray projection images) corresponding to different respiratory phases of the lung (3302). Each displacement field represents the movement of lung tissue from a first respiratory phase to a second respiratory phase, and each respiratory phase has a corresponding set of bioparameters. In some embodiments, the X-ray measurements are X-ray projection images (also referred to as X-ray projection images) obtained by method 3100 and / or method 3200. In some embodiments, extracting displacement fields from lung tissue includes identifying portions of lung tissue in one or more first X-ray projection images from a first respiratory phase as corresponding to the same portions of lung tissue in one or more second X-ray projection images from a second respiratory phase, and determining the displacement of portions of lung tissue from a first respiratory phase to a second respiratory phase (see, for example, Figure 28B). In some embodiments, the identification is performed using the deformable image registration algorithm described above.

[0249] In some embodiments, the multiple X-ray measurements include multiple X-ray images obtained during a first respiratory phase for each of multiple orientations of the X-ray imaging device relative to the patient, thereby forming multiple X-ray images corresponding to the first respiratory phase and multiple X-ray images corresponding to a second respiratory phase. In some situations, at least one X-ray image corresponding to the first respiratory phase is obtained during the respiratory cycle of a different patient (e.g., images obtained during the same respiratory phase but different breaths). In some embodiments, this method includes grouping the multiple X-ray images by respiratory phase.

[0250] In some embodiments, the method includes extracting a plurality of vector fields, one of which is a displacement field.

[0251] In some embodiments, one or more sensors are used to measure the patient's biosignals as one or more time-series sequences (3304), and one or more sensors include a 3D spatial position localizer (e.g., the 3D spatial localizer 300 in Figure 3), a respiratory phase sensor, and a heart rate phase sensor. In some embodiments, the 3D spatial position localizer is configured to measure the patient's body movements caused by respiration and heartbeat in real time and output them as a time series (3306). In some embodiments, the respiratory phase sensor is configured to measure one or more physiological indicators related to the patient's respiration, including tidal volume and its first-order time derivative (3308). In some embodiments, the heart rate phase sensor is configured to measure periodic and steady-state electrical signals (e.g., ECG signals) generated by the patient's heart (3310). For example, the heart rate phase sensor measures periodic and steady-state electrical signals having features corresponding to the heart rate phase.

[0252] In some embodiments, a patient's biosignals measured by one or more sensors are used to trigger an X-ray unit (3312) and to acquire X-ray images of the patient during specific respiratory and cardiac phases (see, for example, method 3200).

[0253] In some embodiments, the X-ray unit includes a clock (3314), and the patient's biosignals measured by one or more sensors are synchronized with the X-ray unit's clock. In some embodiments, each value of the biosignals is recorded in association with the acquired X-ray image.

[0254] In some embodiments, patient biosignals measured during a training session are used to construct an optimized respiratory prediction model to predict a desired respiratory period for triggering an X-ray unit to take an X-ray image of the patient (3316).

[0255] The method includes calculating one or more biophysical parameters of a lung biophysical model using multiple displacement fields of lung tissue between different respiratory phases of the lung and corresponding sets of bioparameters (3318). In some embodiments, calculating one or more biophysical parameters includes calculating one or more derivatives (e.g., curl, gradient, etc.) of the displacement fields. In some embodiments, the biophysical parameters are biomechanical parameters (e.g., stress, strain, modulus of elasticity, elastic limit, etc.). In some embodiments, one or more biophysical parameters define the physical relationship between bioparameters associated with different respiratory phases of the lung and multiple displacement fields of lung tissue (3320). In some embodiments, the set of bioparameters associated with each respiratory phase includes the tidal volume and airflow of the lung in each respiratory phase, as well as the heart rate phase corresponding to each respiratory phase (3322). In some embodiments, the physical relationship between bioparameters associated with different respiratory phases of the lung and multiple displacement fields of lung tissue is defined by the following equation.

number

[0256] p1 → The vector represents the normal stress caused by the tidal volume, p2 → This is the normal stress caused by airflow, and p3 → This is the shear stress caused by airflow, and p4 → This is tissue motion caused by cardiac movement, and the displacement of tissue at any point in a closed-loop trajectory (U → -U0 → ) is the tidal volume (T v ), airflow (A f ), heart rate phase (H c It is expressed as the sum of the stress, strain, and perturbational cardiac motion vectors, each scaled by ).

[0257] In some embodiments, the different respiratory phases of the lungs include the early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase in the patient's complete respiratory cycle (3324).

[0258] In some embodiments, the method further includes displaying a visual representation of the biophysical parameters. For example, in the example of Figure 30, the biophysical parameters are p1 → (p2 → +p3 → This is the ratio of ). In some embodiments, displaying a visual representation includes displaying an image of the lung, where the color of the location within the lung image corresponds to a biophysical parameter (for example, the lung image is displayed using a color map of biophysical parameters). Displaying such a visual representation improves the accuracy of the diagnosis and thus improves the X-ray imaging device itself. For example, as shown in Figure 29, it is clearer and easier to see in Figure 30 that the patient's left lung is affected than in a conventional X-ray image.

[0259] In some embodiments, the method generates multiple medical image cubes corresponding to different respiratory phases of the lung from multiple X-ray measurements corresponding to different respiratory phases of the lung (3326). This further includes (see, for example, Figure 21). In some embodiments, multiple displacement fields of lung tissue are extracted from multiple medical image cubes corresponding to different respiratory phases of the lung by image segmentation to depict the lung tissue from the remainder of a first medical image cube, and for each voxel in the first medical image cube, by determining the displacement vector between the voxel in the first medical image cube and the second medical image cube using intensity-based structural mapping between the first medical image cube and the second medical image cube (3328). Multiple displacement fields of lung tissue are extracted from multiple medical image cubes corresponding to different respiratory phases of the lung by iteratively adjusting the displacement vectors of different voxels in the first medical image cube and the corresponding voxels in the second medical image cube.

[0260] In some embodiments, the method includes selecting one or more medical image cubes from a plurality of medical image cubes as reference medical image cubes (3330), determining a set of bioparameters associated with each reference medical image cube, selecting one set of bioparameters based on lung biometrics between two sets of bioparameters associated with two reference medical image cubes, and simulating medical image cubes between the two reference medical image cubes by applying the set of bioparameters based on lung biometrics to a biophysical model.

[0261] The specific sequence of operations described in Figures 33A-33C is merely an example, and the described sequence does not mean that it is the only sequence in which the operations can be performed. Those skilled in the art will understand that the operations described herein can be rearranged in various ways. Furthermore, other processes described herein with respect to other methods described herein can also be applied in a manner similar to method 3300 described above with reference to Figures 33A-33C. Such processes are described, for example, with reference to Figures 2, 6, 8, 11, 13, 15, 21, 23, 25, 27, 28A-28B, 31A-31B, 32A-32B, and 34A-34C. For the sake of brevity, detailed explanations are omitted.

[0262] Figures 34A to 34C are flowcharts of the method 3400 for generating 3D X-ray image cube videos. In some embodiments, some or all of the operations described below can be performed without human intervention (e.g., without intervention by a technician). In some embodiments, method 3400 is performed by any of the apparatus described herein (e.g., the GREX imaging system 100 shown in Figure 1). Some operations of method 3400 are performed by a computer system including one or more processors and a memory that stores instructions to cause one or more processors to perform the operations of method 3400 when performed by one or more processors. Some operations of method 3400 may be combined in any order and in any order.

[0263] In some embodiments, one or more sensors are used to measure the patient's biological signals as one or more time-series sequences (3402), where one or more sensors include a 3D spatial position localizer, a respiratory phase sensor, and a heart rate phase sensor (see, for example, methods 3100 and 3200).

[0264] In some embodiments, the method further includes identifying a heart rate phase gate frame using measurements from one or more heart rate phase sensors, predicting a respiratory phase using measurements from one or more respiratory phase sensors, identifying a point of agreement between the heart rate phase gate frame and the predicted respiratory phase that generates an X-ray imaging pulse, and tagging the X-ray image corresponding to the X-ray imaging pulse with measurements from the respiratory phase, heart rate phase, and 3D spatial position localizer (3406) (see, for example, method 3200).

[0265] In some embodiments, the 3D spatial position localizer is used to determine the patient's position based on respiration and heart rate. The system is configured to measure the body's movements in real time and output them as a time series (3408) (see, for example, methods 3100 and 3200).

[0266] In some embodiments, the respiratory sensor is configured to measure one or more physiological indicators related to the patient's respiration, including tidal volume and its first-order time derivative (3410). For example, the tidal volume percentage changes with time and airflow.

[0267] In some embodiments, the heart rate phase sensor is configured to measure periodic, steady-state electrical signals generated by the patient's heart and having characteristics corresponding to the heart rate phase (3412).

[0268] In some embodiments, after being synchronized with the X-ray unit clock, two separate filters are used to remove signal drift and noise from the patient's biosignals (3414).

[0269] In some embodiments, a patient's biosignals measured by one or more sensors are used to trigger an X-ray unit and acquire X-ray images of the patient during specific respiratory and cardiac phases (3416).

[0270] In some embodiments, the X-ray unit includes a clock (3418). Patient biosignals measured by one or more sensors are synchronized with the clock of the X-ray unit. Each value of the biosignal is recorded in association with the acquired X-ray image.

[0271] In some embodiments, the patient's biosignals are measured during a training period (e.g., a training period) before taking X-ray images of the patient (3422), and the patient's biosignals measured during the training period include multiple complete respiratory cycles of the patient (see, for example, methods 3100 and 3200).

[0272] In some embodiments, multiple tidal volume percentiles in a complete respiratory cycle are identified using patient biosignals measured during the training period (3424), and each tidal volume percentile corresponds to one of several different respiratory phases.

[0273] In some embodiments, patient biosignals measured during a training session are used to construct an optimized respiratory prediction model to predict a desired respiratory period for triggering an X-ray unit to take an X-ray image of the patient (3426).

[0274] In some embodiments, the optimized respiratory prediction model is based on an autoregressive integrated moving average (ARIMA) model (3428).

[0275] In some embodiments, the desired respiratory period for taking an X-ray image of the patient is configured to coincide with a cardiac gate frame in which cardiac-induced lung motion is slowly changing (3430).

[0276] In some embodiments, the cardiac gate frame is selected based on the positions of the T wave and P wave in the electrocardiogram (ECG) signal (3432), thereby slowing down the cardiac-induced lung motion.

[0277] The method includes converting a first set of X-ray images of the lung taken at different projection angles into a second set of X-ray images of the lung corresponding to different respiratory phases (3434).

[0278] In some embodiments, converting a first set of X-ray images of the lung taken at different projection angles to a second set of X-ray images of the lung corresponding to different respiratory phases is done by using different projection angles. The method further includes taking a first set of X-ray images of the lungs at different projection angles (3436). Each set in the first set of X-ray images corresponds to a different respiratory phase of the lungs at a particular projection angle. In some embodiments, converting the first set of X-ray images of the lungs taken at different projection angles to a second set of X-ray images of the lungs corresponding to different respiratory phases further includes reorganizing the first set of X-ray images of the lungs into a second set of X-ray images of the lungs according to the associated respiratory phases. Each set in the second set of X-ray images corresponds to a different respiratory phase of the lungs.

[0279] In some embodiments, X-ray images within any particular set are geometrically resolved and temporally independent (3438).

[0280] In some embodiments, different respiratory phases of the lungs correspond to different tidal volume percentiles of lung movement (3440).

[0281] In some embodiments, the different respiratory phases of the lungs include the early expiratory phase, late expiratory phase, maximum expiratory phase, early inspiratory phase, late inspiratory phase, and maximum inspiratory phase in the patient's complete respiratory cycle (3442).

[0282] In some embodiments, multiple X-ray images of the lung taken at different projection angles all correspond to the same respiratory phase (3444).

[0283] In some embodiments, different respiratory phases of the lungs at a specific projection angle are collected from at least two respiratory cycles (3446).

[0284] The method includes generating still image cubes from each set of a second set of X-ray images for each respiratory phase using back projection (3448).

[0285] The method includes combining still image cubes corresponding to different respiratory phases of the lungs into a 3D X-ray image cube video using a time interpolation method (3450).

[0286] The specific sequence of operations described in Figures 34A-34C is merely an example, and the described sequence does not mean that it is the only sequence in which the operations can be performed. Those skilled in the art will understand that the operations described herein can be rearranged in various ways. Furthermore, other processes described herein with respect to other methods described herein are also applicable in a manner similar to method 3400 described above with reference to Figures 34A-34C. Such processes are described, for example, with reference to Figures 2, 6, 8, 11, 13, 15, 21, 23, 25, 27, 28A-28B, 31A-31B, 32A-32B, and 33A-33C. For the sake of brevity, detailed explanations are omitted.

[0287] Figure 35 shows an example of a patient positioning and restraining device (PPF) 3501 (e.g., a swivel chair) supporting a patient, according to some embodiments (3502). In some embodiments, the PPF 3501 rotates (e.g., along rotation 3503) to position the patient at multiple angles (e.g., orientations) relative to a radiation source (3504) (e.g., an X-ray imaging system or a radiotherapy system) (2502). For example, the PPF moves to move the patient to the desired position (e.g., the patient does not need to move independently), so that the radiation device 3504 can take X-ray images of the patient at various angles. In some embodiments, the patient rotates (e.g., along rotation 3502) over multiple angles without rotating the PPF 3501 (3502). In some embodiments, the PPF 3501 and / or the patient 3205 automatically rotate and / or move (e.g., using a motor) to the desired position. In some embodiments, the technician rotates and / or moves the patient 3502. In one embodiment, the flat panel detector unit 3505 is positioned behind the patient relative to the radiation device 3504.

[0288] In some embodiments, one or more cameras 3506-1 to 3506-m are used to detect objects within a predetermined area (e.g., a room) surrounding the PPF 3501. For example, camera 3506 captures whether an object collides with the PPF 3501 as the PPF 3501 moves around (and rotates within) the predetermined area. In some implementations, measures are taken to avoid collisions with the PPF 3501 (and the patient 3502).

[0289] In some embodiments, one or more 3D imaging sensors (e.g., LiDAR sensors or structured light sensors) are used to geometrically monitor the respiration of the patient 3502, as in method 3100.

[0290] Figure 36 shows an example of a Biological Event Monitoring Process (BEMP) card 3600. The BEMP card 3600 includes a programmable analog filter, a programmable analog-to-digital converter (ADC) / digital signal processor (DSP) 3605, and a Zero Instruction Set Computer (ZISC) processor 3606. The BEMP card 3600 receives an analog input signal 3602 containing the patient's biological signals. In some embodiments, the analog input 3602 includes an ECG signal. In some embodiments, the analog input is a 3-read ECG signal or a 12-read ECG signal. The ZISC processor 3606 detects a predetermined pattern in the analog input in real time. For example, the ZISC processor 3606 detects the R wave of the ECG signal to predict when the next sedation period in the patient's cardiac cycle will occur (e.g., within the TP interval). In some embodiments, the ZISC processor 3606 outputs a ZISC 4-bit weighted output that can be used to trigger a radiation source (e.g., an X-ray imaging system or a radiotherapy system). In some embodiments, the clock signal 3603 of the BEMP card 3600 is supplied from an external source such as a radiation source, thereby allowing the BEMP card 3600 to be synchronized with the radiation source.

[0291] The computational techniques described in Chapter 106 of the post-processing software consume a significant amount of computing resources. Many local hospitals and small clinics lack the necessary computer hardware to create videos, calculate biomechanical models, and provide users with fast and smooth experience results. Because the functions of post-processing software 106 run in the cloud, users can access the powerful visualization tools of post-processing software 106 from their treatment console, office desktop, or work laptop.

[0292] Throughout this specification, the terms “embodiment,” “some embodiment,” “one embodiment,” “another embodiment,” “example,” “specific example,” or “some example” mean that any material or property described in relation to a particular feature, structure, embodiment, or example is included in at least one embodiment or example of this disclosure. Accordingly, expressions such as “in some embodiment,” “in one embodiment,” “in an embodiment,” “in another embodiment,” “in an example,” “in a specific example,” or “in some examples” in this specification do not necessarily refer to embodiments or examples of this disclosure. Furthermore, particular features, structures, materials, and properties can be combined in any suitable way in one or more embodiments or examples.

[0293] While embodiments have been presented and described above, those skilled in the art will understand that these embodiments are not limiting to the present disclosure, and that modifications, substitutions, and alterations can be made as long as they do not deviate from the principles, spirit, and scope of the present disclosure.

Claims

1. A method for generating a 3D X-ray image cube video from a patient's 2D X-ray image, Converting a plurality of first sets of X-ray images of the lungs into a plurality of second sets of X-ray images of the lungs, Each first set of X-ray images includes multiple X-ray images taken at different tidal volume-based respiratory phases and at the same projection angle of the lung. Each of the multiple first sets of X-ray images is taken at a different projection angle than another first set of X-ray images in the multiple first sets of X-ray images. Each second set of X-ray images includes multiple X-ray images taken of the lung during the same respiratory period and at different projection angles. Each of the multiple second sets of X-ray images is taken during a different respiratory period than the other second set of X-ray images in the aforementioned multiple second sets of X-ray images. The conversion includes rearranging a plurality of first sets of X-ray images of the lungs into a plurality of second sets of X-ray images of the lungs based on associated tidal volume-based respiratory periods. Using the back projection method, still image cubes are generated from each set of multiple second sets of the X-ray images for each tidal volume-based respiratory period, A method comprising combining still image cubes corresponding to different tidal volume-based respiratory phases of the lung into a 3D X-ray image cube video by time interpolation.

2. The method according to claim 1, wherein X-ray images within any particular set are geometrically resolved and time-independent.

3. The method according to claim 1, wherein the different tidal volume-based respiratory phases of the lungs correspond to different tidal volume percentiles of the lung movement.

4. The method according to claim 1, wherein the different tidal volume-based respiratory phases of the lungs include an early expiratory phase, a late expiratory phase, a maximum expiratory phase, an early inspiratory phase, a late inspiratory phase, and a maximum inspiratory phase in the patient's complete respiratory cycle.

5. The method according to claim 1, wherein one or more sensors are used to measure the patient's biological signals as one or more time-series sequences, and the one or more sensors include one or more 3D spatial position localizers, a respiratory phase sensor, and a heart rate phase sensor.

6. Identifying the heart rate phase gate frame using measurements from one or more heart rate phase sensors, Predicting the respiratory period based on tidal volume using measurements from one or more respiratory period sensors, To identify the point of agreement between the heart rate phase gate frame that generates the X-ray imaging pulse and the predicted tidal volume-based respiratory phase, The method according to claim 5, further comprising tagging the X-ray image corresponding to the X-ray imaging pulse with the tidal volume-based respiratory phase, heart rate phase, and 3D spatial position localizer measurements.

7. The method according to claim 5, wherein the 3D spatial position localizer is configured to measure the patient's body movements caused by respiration and heartbeat in real time and output them as a time series.

8. The method according to claim 5, wherein the respiratory sensor is configured to measure one or more physiological indicators related to the patient's respiration, including tidal volume and its first-order time derivative.

9. The method according to claim 5, wherein the heart rate phase sensor is configured to measure periodic, steady-state electrical signals generated by the patient's heart and having characteristics corresponding to the heart rate phase.

10. The method according to claim 5, wherein the patient's biological signal is synchronized with the clock of the X-ray unit, and then signal drift and noise are removed from the patient's biological signal using two separate filters.

11. The method according to claim 5, wherein an X-ray unit is triggered using the patient's biosignals measured by one or more sensors, and an X-ray image of the patient is acquired during a specific respiratory and cardiac phase.

12. The aforementioned X-ray unit includes a clock, The patient's biological signals, measured by one or more of the aforementioned sensors, are synchronized with the clock of the X-ray unit. The method according to claim 11, wherein the values ​​of each of the biological signals are recorded in association with the acquired X-ray image.

13. The method according to claim 5, wherein the patient's biosignals are measured during a training session before an X-ray image of the patient is taken, and the patient's biosignals measured during the training session include multiple complete respiratory cycles of the patient.

14. The method according to claim 13, wherein multiple tidal volume percentiles in a complete respiratory cycle are identified using patient biosignals measured during a training frame, and each tidal volume percentile corresponds to one of different tidal volume-based respiratory phases.

15. The method according to claim 13, wherein the patient's biosignals measured during the training session are used to construct an optimized respiratory prediction model for predicting desired tidal volume-based respiratory periods, which trigger an X-ray unit to take an X-ray image of the patient.

16. The method according to claim 15, wherein the optimized respiration prediction model is based on an autoregressive integrated moving average (ARIMA) model.

17. The method according to claim 15, wherein the desired tidal volume-based respiratory period for taking an X-ray image of the patient is configured to coincide with a cardiac gate frame in which cardiac-induced lung motion is slowly changing.

18. The method according to claim 17, wherein the cardiac gate frame is selected based on the positions of the T wave and P wave in the electrocardiogram (ECG) signal, thereby causing cardiac-induced lung motion to change slowly.

19. The method according to claim 1, wherein different tidal volume-based respiratory periods of the lung at a specific projection angle are collected from at least two respiratory cycles.

20. A system for determining a patient's lung biophysical model from multiple X-ray measurements corresponding to different tidal volume-based respiratory phases of the lung, One or more processors, A memory for storing instructions, and when an instruction is executed by one or more processors, the one or more processors Converting a plurality of first sets of X-ray images of the lungs into a plurality of second sets of X-ray images of the lungs, Each first set of X-ray images includes multiple X-ray images taken at different tidal volume-based respiratory phases and at the same projection angle of the lung. Each of the multiple first sets of X-ray images is taken at a different projection angle than another first set of X-ray images in the multiple first sets of X-ray images. Each second set of X-ray images includes multiple X-ray images taken of the lung during the same respiratory period and at different projection angles. Each of the multiple second sets of X-ray images is taken during a different respiratory period than the other second set of X-ray images in the aforementioned multiple second sets of X-ray images. The conversion includes rearranging a plurality of first sets of X-ray images of the lungs into a plurality of second sets of X-ray images of the lungs based on associated tidal volume-based respiratory periods. Using the back projection method, still image cubes are generated from each set of multiple second sets of the X-ray images for each tidal volume-based respiratory period, A system that performs a series of operations including combining still image cubes corresponding to different tidal volume-based respiratory phases of the lung into a 3D X-ray image cube video using a time interpolation method.

21. A non-temporary computer-readable storage medium for storing instructions, wherein when the instructions are executed by a system comprising one or more processors, the one or more processors: Converting a plurality of first sets of X-ray images of the lungs into a plurality of second sets of X-ray images of the lungs, Each first set of X-ray images includes multiple X-ray images taken at different tidal volume-based respiratory phases and at the same projection angle of the lung. Each of the multiple first sets of X-ray images is taken at a different projection angle than another first set of X-ray images in the multiple first sets of X-ray images. Each second set of X-ray images includes multiple X-ray images taken of the lung during the same respiratory period and at different projection angles. Each of the multiple second sets of X-ray images is taken during a different respiratory period than the other second set of X-ray images in the aforementioned multiple second sets of X-ray images. The conversion includes rearranging a plurality of first sets of X-ray images of the lungs into a plurality of second sets of X-ray images of the lungs based on associated tidal volume-based respiratory periods. Using the back projection method, still image cubes are generated from each set of multiple second sets of the X-ray images for each tidal volume-based respiratory period, A storage medium that performs a series of operations including combining still image cubes corresponding to different tidal volume-based respiratory phases of the lung into a 3D X-ray image cube video using a time interpolation method.

Images