Apparatus for monitoring a patient undergoing a magnetic resonance image scan
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- KONINKLIJKE PHILIPS NV
- Filing Date
- 2021-01-29
- Publication Date
- 2026-05-29
Smart Images
Figure CN115066626B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an apparatus, an imaging system, a method for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan, a computer program unit, and a computer-readable medium. Background Technology
[0002] Patient cooperation is a crucial determinant of the duration and diagnostic quality of MRI examinations. To acquire MR images, patients are required to enter the narrow chamber of an unfamiliar, noisy machine and remain still for approximately 15 to 60 minutes. Consequently, many patients experience anxiety-related reactions to MRI examinations (Meléndez, JC, & McCrank, E. (1993). Anxiety-related reactions associated with MRI examinations. Jama, 270(6), 745-747), and patient movement is highly common during MRI examinations (occurring in 7% to 29% of MRI examinations; Andre, JB, Bresnahan, BW, Mossa-Basha, M., Hoff, MN, Smith, CP, Anzai, Y., & Cohen, WA (2015). Towards quantifying the prevalence, severity, and cost associated with patient movement during clinical MRI examinations. Journal of the American College of Radiology, 12(7), 689-695). Patient movement can lead to parts of an MRI scan requiring re-examination (approximately 20% of MRI scans are re-examinations; Andre et al., 2015). Furthermore, patients with severe discomfort sometimes leave the scanner before the examination is completed (these are referred to as “incomplete examinations”).
[0003] Although such anxiety can be detrimental to MRI success, hospital staff currently have very limited information about the participants' mental state (e.g., level of anxiety or discomfort). They can only assess a patient's mental state, their stress level, and therefore the likelihood of motion artifacts through standard interpersonal contact. Furthermore, they can only do this before the actual scan.
[0004] Therefore, a patient's emotional and physiological state is not easily visible or available to the technician during an MRI examination or scan. However, if the patient becomes very anxious or in pain, or moves improperly, the scan should be stopped. In some cases, the patient may press an alarm button or allow himself / herself to hear it acoustically. However, the decision to press the button can be made very late. In the worst case, the button may malfunction due to technical problems, may be lost or outside the patient's reach, or the patient may have already lost consciousness without the technician noticing.
[0005] These problems need to be addressed.
[0006] US2020 / 008703 A1 discloses a method for monitoring a patient during a medical imaging examination, which detects at least one state signal of a patient and evaluates it using an evaluation algorithm to derive the patient's emotional state and the general trend of that emotional state.
[0007] EP 3 381 353 A1 relates to a method for planning an imaging scan protocol through a scanning imaging system, including recording a patient’s physiological data and modifying it during imaging, even if this results in an increase in noise emission attributable to an increase in the magnetic field.
[0008] US2013 / 188830 A1 relates to a system for adaptively compensating for object motion in real time in an imaging system.
[0009] From EP 2 921 100 A1, a method for adapting a medical system to the motion of an object is well known, including the use of the system for detecting and quantifying the motion of an object. Summary of the Invention
[0010] Because patients may become too anxious or move too much in their body parts, it would be advantageous to have improved means of determining when to stop an MRI scan. The object of the invention is addressed using one aspect of the subject matter, wherein further embodiments are included in other aspects. It should be noted that the aspects and examples described below of the invention are also applicable to devices or monitoring of patients undergoing magnetic resonance imaging scans, imaging systems, methods for monitoring patients undergoing magnetic resonance imaging scans, as well as computer program units and computer-readable media.
[0011] In a first aspect, an apparatus is provided for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan, the apparatus comprising:
[0012] Input unit;
[0013] Processing unit; and
[0014] Output unit.
[0015] The input unit is configured to provide the processing unit with at least one sensor data of a patient undergoing an MRI scan using an MRI scanner. The input unit is also configured to provide the processing unit with at least one scan parameter of the MRI scanner used for the MRI scan. Furthermore, the input unit is configured to provide the processing unit with at least one characteristic of the patient. The processing unit is configured to predict the patient's stress level and / or predict the patient's motion state, the one or more predictions comprising utilizing the at least one sensor data of the patient, the at least one scan parameter of the MRI scanner, and the at least one characteristic of the patient. The output unit is configured to output information related to the predicted stress level and / or the predicted motion state of the patient.
[0016] In this way, by analyzing sensor data, data about the ongoing scan, and data about the patient's emotional and physiological state, the patient's stress level and / or likelihood of movement are determined, and these states are predicted to develop in the future. Therefore, real-time feedback can be provided to technical experts, who can then decide that if the scan is predicted to enter a state of anxiety or movement that is inconsistent with the scanning protocol, the scan should be stopped, and the device can indeed automatically initiate such a stop.
[0017] Therefore, the device objectively measures and predicts the patient's anxiety / discomfort level during scanning, as well as the predicted likelihood of motion artifacts. This information can be used to automatically abort the scan if needed.
[0018] The problem addressed by this device is reducing patient anxiety / discomfort, which could otherwise negatively impact the patient experience. Scanning can be stopped before patient movement, associated with reduced image quality, has already impaired image quality.
[0019] The patient's motion state can also be referred to as the patient's movement state.
[0020] In an example, the processing unit is configured to implement at least one machine learning algorithm to predict the patient's stress level and / or the patient's motion state, wherein the at least one machine learning algorithm is trained based on: at least one sensor data of one or more reference patients undergoing one or more reference MRI scans by one or more reference MRI scanners; at least one scan parameter of the reference MRI scanner used for the one or more reference MRI scans; and at least one characteristic of each of the reference patients.
[0021] Therefore, AI-based algorithms are used to infer the patient's emotions and physical state from sensor data, scanning protocol data, and background information related to the patient, enabling the development of the patient's stress state and mobility state to be predicted.
[0022] In other words, the predictive algorithm takes data about the patient that can be collected before the examination, either in a hospital or at home, and uses this together with information collected from sensors during the scan and details about the scan to predict the patient’s anxiety and mobility levels during the scan, wherein the determination may also utilize the results of previous examinations.
[0023] In the example, the patient's at least one sensor data is acquired by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, processing unit, pressure sensor, and weight sensor.
[0024] In the example, the at least one sensor data of the one or more reference patients is acquired by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, processing unit, pressure sensor, weight sensor.
[0025] In the example, the patient's at least one sensor data includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary display, and weight distribution on the intracavitary examination table.
[0026] In the example, the at least one sensor data from the one or more reference patients includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary display, and weight distribution on the intracavitary examination table.
[0027] In the example, the at least one scanning parameter of the MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
[0028] In the example, the at least one scan parameter of the reference MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
[0029] In the example, the patient’s at least one characteristic includes one or more of the following: age, weight, body mass index, information about a previous diagnosis, the patient’s physical condition, the patient’s mental condition, completed questionnaire information, and patient feedback.
[0030] In the example, the at least one characteristic of each of the reference patients includes one or more of the following: age, weight, body mass index, information about a previous diagnosis, the patient's physical condition, the patient's mental condition, and completed questionnaire information.
[0031] In an example, the input unit is configured to provide the processing unit with information about one or more previous MRI scans performed on the patient, and wherein the prediction of the patient's stress level and / or the patient's motor state includes utilizing information about one or more previous MRI scans performed on the patient.
[0032] In the example, the training of the machine learning algorithm includes utilizing information about one or more previous MRI scans performed by at least one of the one or more reference patients.
[0033] In an example, the processing unit is configured to determine whether the predicted stress level of the patient and / or the predicted movement state of the patient will exceed a stress threshold level or a movement threshold level, and wherein the information relating to the predicted stress level of the patient and / or the predicted movement state of the patient includes an indication of whether either threshold is predicted to be exceeded.
[0034] In a second aspect, an imaging system is provided, comprising:
[0035] Magnetic resonance imaging scanner;
[0036] At least one sensor; and
[0037] According to the first aspect, a device for monitoring patients undergoing magnetic resonance imaging (MRI) scans.
[0038] The at least one sensor is configured to provide at least one sensor data of a patient undergoing an MRI scan to the processing unit of the device. The device is configured to automatically stop the MRI scan based on information related to the predicted stress level and / or the predicted motion state of the patient.
[0039] In a third aspect, a method for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan is provided, the method comprising:
[0040] Provide the processing unit with at least one sensor data of a patient who has undergone an MRI scan by an MRI scanner;
[0041] The processing unit is provided with at least one scanning parameter of the MRI scanner for the MRI scan;
[0042] The processing unit is provided with at least one characteristic of the patient;
[0043] The processing unit predicts the patient's stress level and / or predicts the patient's motion state, the one or more predictions including utilizing the patient's at least one sensor data, the MRI scanner's at least one scan parameter, and the patient's at least one characteristic; and
[0044] The output unit outputs information related to the predicted stress level and / or the predicted motor state of the patient.
[0045] According to another aspect, a computer program unit is provided for controlling one or more of the apparatus or system as previously described, which is adapted to perform one or more of the methods as previously described when the computer program unit is run by a processing unit.
[0046] According to another aspect, a computer-readable medium is provided that stores computer units as previously described.
[0047] The computer program unit may be, for example, a software program, and may be an FPGA, PLD, or any other suitable digital module.
[0048] Advantageously, the benefits provided by any of the above aspects apply equally to all other aspects and vice versa.
[0049] The above aspects and examples will become apparent and will be illustrated with reference to the embodiments described below. Attached Figure Description
[0050] Exemplary embodiments will be described below with reference to the following figures:
[0051] Figure 1 A schematic setup of an example device for monitoring a patient undergoing a magnetic resonance imaging scan is shown;
[0052] Figure 2 A schematic setup of an example imaging system is shown;
[0053] Figure 3 A method for monitoring patients undergoing magnetic resonance imaging scans is shown;
[0054] Figure 4 It shows Figure 1 Devices or having such devices Figure 2 The high-level system architecture of a detailed embodiment of the imaging system; and
[0055] Figure 5 An example of a stress level indicator is shown. Detailed Implementation
[0056] Figure 1 An example of a device 10 for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan is shown. Device 10 includes an input unit 20, a processing unit 30, and an output unit 40. The input unit is configured to provide the processing unit with at least one sensor data of the patient undergoing an MRI scan using an MRI scanner. The input unit is also configured to provide the processing unit with at least one scan parameter of the MRI scanner used for the MRI scan. The input unit is further configured to provide the processing unit with at least one characteristic of the patient. The processing unit is configured to predict the patient's stress level and / or predict the patient's motion state, one or more predictions comprising utilizing at least one sensor data of the patient, at least one scan parameter of the MRI scanner, and at least one characteristic of the patient. The output unit is configured to output information relating to the predicted stress level and / or the predicted motion state of the patient.
[0057] According to the example, the processing unit is configured to implement at least one machine learning algorithm to predict the patient's stress level and / or the patient's motion state. The at least one machine learning algorithm is trained based on: at least one sensor data from one or more reference patients undergoing one or more reference MRI scans using one or more reference MRI scanners; at least one scan parameter from the reference MRI scanners used in the one or more reference MRI scans; and at least one characteristic of each of the reference patients.
[0058] In this example, the machine learning algorithm includes one or more neural networks.
[0059] In the example, the machine learning algorithm includes one or more classic machine learning algorithms.
[0060] In the example, the machine learning algorithm includes one or more support vector machines (SVMs).
[0061] In this example, the machine learning algorithm includes one or more decision trees.
[0062] In this example, at least one machine learning algorithm comprises two parts. The first part of the machine learning algorithm is used to determine a person's stress state and mobility state. Here, mobility state can refer to the probability of movement, which can have different levels even for a stationary patient. Thus, a stationary patient can be determined to have a low probability of movement, or a stationary patient can be determined to have a high probability of movement. The second part of the machine learning algorithm then operates to determine a prediction of the future of the person's stress state and mobility state. The first part can be, for example, a standard neural network, such as a convolutional neural network (CNN), and the second part can be a recurrent neural network (RNN) or a long short-term memory (LSTM) neural network version of an RNN. Therefore, a neural network here can refer to a combination of neural networks.
[0063] According to the example, at least one sensor data of the patient is collected by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, processing unit, pressure sensor, weight sensor.
[0064] According to the example, at least one sensor data of one or more reference patients is acquired by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, processing unit, pressure sensor, weight sensor.
[0065] According to the example, at least one sensor data of the patient includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary monitor, and weight distribution on the intracavitary examination table.
[0066] According to the example, at least one sensor data from one or more reference patients includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary monitor, and weight distribution on the intracavitary examination table.
[0067] According to the example, at least one scanning parameter of the MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
[0068] According to the example, at least one scan parameter of the reference MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
[0069] According to the example, at least one characteristic of the patient includes one or more of the following: age, weight, body mass index, information about previous diagnosis, the patient's physical condition, the patient's mental condition, completed questionnaire information, and patient feedback.
[0070] According to the example, at least one characteristic of each of the reference patients includes one or more of the following: age, weight, body mass index, information about previous diagnosis, the patient's physical condition, the patient's mental condition, and completed questionnaire information.
[0071] According to the example, the input unit is configured to provide the processing unit with information about one or more previous MRI scans performed on the patient. Prediction of the patient's stress level and / or prediction of the patient's motor state may then include utilizing the information about the one or more MRI scans performed on the patient.
[0072] As an example, training a machine learning algorithm involves using information about one or more previous MRI scans performed by at least one of one or more reference patients.
[0073] According to the example, the processing unit is configured to determine whether the predicted stress level of the patient and / or the predicted motor state of the patient will exceed a stress threshold level or a motor threshold level. The information relating to the predicted stress level of the patient and / or the predicted motor state of the patient may then include indications as to whether one or both of these thresholds are predicted to be exceeded.
[0074] Figure 2 An example of an imaging system 100 is shown. The imaging system 100 includes a magnetic resonance imaging scanner 110 and at least one sensor 120. The system 100 also includes a device 10 for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan, as described above. Figure 2As described. At least one sensor 120 is configured to provide at least one sensor data of a patient undergoing an MRI scan by an MRI scanner to the processing unit of the device. The device 10 is configured to automatically stop the MRI scan performed by the magnetic resonance imaging scanner 110 based on information relating to the predicted stress level and / or the predicted motion state of the patient.
[0075] Figure 3 A method 200 for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan during its basic steps is shown. Method 200 includes:
[0076] In step 210, also referred to as step a), at least one sensor data of a patient undergoing an MRI scan by an MRI scanner is provided to the processing unit;
[0077] In step 220, also referred to as step b), at least one scanning parameter of an MRI scanner for MRI scanning is provided to the processing unit;
[0078] In step 230, also referred to as step c), at least one characteristic of the patient is provided to the processing unit;
[0079] In prediction step 240, also referred to as step d), the processing unit predicts the patient's stress level and / or predicts the patient's motion state, one or more predictions including utilizing at least one sensor data of the patient, at least one scan parameter of an MRI scanner, and at least one characteristic of the patient; and
[0080] In the output step 250, also referred to as step e), the output unit outputs information relating to the predicted stress level and / or the predicted motor state of the patient.
[0081] In this example, the machine learning algorithm includes one or more neural networks.
[0082] In the example, the machine learning algorithm includes one or more classic machine learning algorithms.
[0083] In the example, the machine learning algorithm includes one or more support vector machines.
[0084] In this example, the machine learning algorithm includes one or more decision trees.
[0085] In the example, step d) includes the processing unit implementing at least one machine learning algorithm to predict the patient's stress level and / or the patient's motion state, wherein the at least one machine learning algorithm is trained based on one or more of the following: at least one sensor data of one or more reference patients undergoing one or more reference MRI scans by one or more reference MRI scanners; at least one scan parameter of the reference MRI scanner used for one or more reference MRI scans; at least one characteristic of each of the reference patients.
[0086] In this example, at least one sensor data of the patient is collected by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, processing unit, pressure sensor, and weight sensor.
[0087] In the example, at least one sensor data of one or more reference patients is acquired by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, processing unit, pressure sensor, weight sensor.
[0088] In this example, at least one of the patient's sensor data includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary monitor, and weight distribution on the intracavitary examination table.
[0089] In the example, at least one sensor data from one or more reference patients includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary monitor, and weight distribution on the intracavitary examination table.
[0090] In the example, at least one scan parameter of the MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
[0091] In the example, at least one scan parameter of the reference MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
[0092] In the example, at least one characteristic of the patient includes one or more of the following: age, weight, body mass index, information about a previous diagnosis, the patient's physical condition, the patient's mental condition, completed questionnaire information, and patient feedback.
[0093] In the example, at least one characteristic of each of the reference patients includes one or more of the following: age, weight, body mass index, information about previous diagnoses, the patient's physical condition, the patient's mental condition, and information from completed questionnaires.
[0094] In an example, the method includes providing a processing unit with information about one or more previous MRI scans performed on the patient, and wherein step d) includes utilizing the information about one or more previous MRI scans performed on the patient.
[0095] In this example, training the machine learning algorithm involves using information about one or more previous MRI scans performed by at least one of one or more reference patients.
[0096] In the example, step d) includes the processing unit determining whether the predicted stress level of the patient and / or the predicted movement state of the patient will exceed a stress threshold level or a movement threshold level, and wherein, in step e), the information relating to the predicted stress level of the patient and / or the predicted movement state of the patient includes an indication of whether either threshold is predicted to be exceeded.
[0097] Now, with regard to specific embodiments, an apparatus, imaging system, and method for monitoring patients undergoing magnetic resonance imaging (MRI) scans are described in more detail, wherein reference is made to... Figure 4-5 .
[0098] Figure 4A high-level system architecture is illustrated. The patient is undergoing an MRI scan, and the biometric and physiological sensors indicated at "A" provide corresponding data to the processing unit indicated at "PU". Scan parameters for the MRI scan, indicated at "B", are also provided to the processing unit. Here, ExamCard results can also be provided to the processing unit, including information such as how long the examination or scan will take and how much time remains. Patient-related information, indicated at "C", is also provided to the processing unit. The processing unit implements a first neural network, indicated at "D", such as a convolutional neural network, to predict the patient's stress level and mobility level. The processing unit also implements a second novel network, indicated at "E", such as a recurrent neural network or a new long short-term memory network, which is used to predict the patient's future stress level and mobility level. The current and predicted stress levels are then displayed to the operator, indicated at "F", and an emergency decision-making logic, indicated at "G", uses the predicted stress level and predicted mobility level to determine whether an emergency stop of the scan should be made, indicated at "H". Therefore, the processing unit predicts a patient's stress level and mobility status based on multiple parameters ("features") obtained from sensors, scan settings, and patient information. Patient feedback regarding the stress and mobility experienced by this patient and other patients can also be provided as part of the training of an internal machine learning algorithm, which can be combined with sensor data and scan parameter information, as well as patient information about those patients who underwent those scans.
[0099] Figure 5 Two examples of stress level indicators determined by the processing unit are shown, presented to the operator or technician operating the MRI scanner, indicating the current and future stress levels, how the stress level has progressed to the present, and how its future progression can be predicted. A motion (or movement) status level indicator can also be provided in a similar manner, indicating the patient's present, past, and future likelihood of movement. Stress level information and / or movement status level information enable the technician or operator to initiate an emergency stop when needed, and the system itself can also initiate an emergency stop if the stress level is predicted to become too high or the likelihood of patient movement is too great.
[0100] The emergency detection logic operates based on information about the current and predicted stress levels, as well as the current and predicted movement status. This logic determines when to stop the MRI scan. Ideally, this logic is configured such that the scan stops when the expected stress level increases above a critical level during scan execution, or when there is an anticipated likelihood that the patient will move to an unacceptable level. This would mean that the patient would otherwise be highly likely to press the emergency button or move in a manner that would impair the integrity of the scan image or lead to a potentially dangerous situation.
[0101] Further details regarding the possible inputs to the processing unit are:
[0102] Sensors used to monitor a patient's mood and physical condition may include:
[0103] Camera-based heart rate and respiration detection. This can be used to measure instantaneous heart rate, which increases with anxiety. Heart rate / respiration rate measured at home using the Preparation app can be used as a baseline for this (see below for details on the Preparation app).
[0104] Camera-based facial recognition and emotion state determination (technologists will realize this is an available AI technology).
[0105] Camera-based motion detection technology
[0106] Microphones used for speech and voice emotion recognition (technologists will realize this is an available AI technology).
[0107] skin electrical activity sensor
[0108] Skin resistance sensor
[0109] Skin temperature sensor
[0110] Skin moisture sensor
[0111] Skin accelerometer
[0112] Other pulse and respiration sensors
[0113] Camera-based motion detection
[0114] Camera-based blink frequency detection
[0115] RF pilot tone
[0116] RF radar sensor
[0117] EEG and ECG data
[0118] Hand movement sensors, for example, can establish micro-movements or minimal movements of the hand, which can be unconscious, and can be linked to hand tension, thereby indicating an increase in stress levels.
[0119] Additional or supplementary physiological or psychological measures may be taken to reflect increased anxiety and discomfort, including heart rate, respiratory rate, and skin conductance.
[0120] Monitor the weight distribution on the patient's table. It was determined that if the patient begins to feel that parts of their body will move involuntarily, they can tense their entire body or major body parts to compensate for this impending movement, and this can be detected.
[0121] Sedation monitoring was conducted via vital sign monitoring to determine when sedated patients began to awaken. It has been established that as patients begin to awaken, this can lead to anxiety and movement.
[0122] Scan parameter information may include:
[0123] Contrast type (T1, T2, DWI, etc.)
[0124] Timing parameters
[0125] Gradient strength setting
[0126] SAR (RF settings)
[0127] k-space sampling mode
[0128] Remaining scan time
[0129] Information about what is shown to patients on an intracavitary monitor is important because it has been determined that what is shown to patients can affect their stress levels and mobility.
[0130] Patient information (e.g., from electronic medical records) may include:
[0131] age
[0132] weight
[0133] Body Mass Index
[0134] Previous diagnosis
[0135] Other physical and mental conditions
[0136] Based on data obtained in the preparation room or at home, before scanning (camera, interviews, questionnaires),
[0137] Regarding this Prepare data More information is detailed below.
[0138] Therefore, in summary, the prediction of stress level and mobility status is achieved by an artificial intelligence (AI) algorithm using features as input and stress level / mobility status as output. This algorithm can be a combination of machine learning methods (such as support vector machines or neural networks) for deriving the current stress level and mobility status and machine learning algorithms (such as RNNs or LSTMs) for predicting the development of stress level over the next few minutes. Supervised training of the AI algorithm is achieved by using feedback from the patient and other patients (e.g., emergency button status or self-estimated stress level) as labels, and via associated sensor data, scan parameter data for those scans, and patient information from other patients for those scans. Training the algorithm on patient feedback allows the stress level / mobility status to be standardized to an individual severity scale. This indicator can be defined as a number ranging from 0 (low stress level / low mobility level - i.e., low probability of movement) to 1 (critical stress level / critical mobility level - i.e., high probability of movement), where reaching the critical stress level means the patient is very likely to press the emergency stop button, and reaching the critical mobility level means the patient is very likely to move and impair the scan. The processing unit outputs the patient's current stress level value and a prediction of its assessment over the next few minutes, as well as the patient's current mobility level value and a prediction of its probability of movement over the next few minutes. The processing unit outputs are provided to an indicator display for technical experts and emergency decision-making logic.
[0139] It should be noted that, generally speaking, not all of the above data will be available for every individual patient. However, the available data can be used as input for anxiety and mobility prediction algorithms. Based on data collected for the current MRI examination, data about the patient, and data from potential earlier examinations (if available), predictions about the expected level of anxiety and the likelihood of mobility can be made. Furthermore, based on general data from earlier scans and individual data, predictions about the expected image quality (mobility interference) and its relationship to anxiety / mobility likelihood can be made.
[0140] In seeking to improve MRI scans, the inventors realized that two major limitations were stress generated within the patient during the scan and patient movement. In researching these areas, the inventors determined that movement and stress were only weakly correlated (and in some studies, they were completely uncorrelated). For example, extremely anxious patients might show very little movement on the scanner (they are too “rigid” or “petrified” to move). And relaxed patients might move because they have itching / because parts of their body are asleep, or they are so relaxed about the procedure that they forget they shouldn’t move.
[0141] Another example is:
[0142] A sudden decrease in movement of body parts may increase the risk of exercise, while anxiety curves may show very different patterns.
[0143] Therefore, if a person's arm falls asleep, it may temporarily reduce the movement of that arm in the short term, but increase the likelihood of movement in the longer term (once the numbness turns into pain and the person repositions themselves).
[0144] If a person is fully asleep, there will be almost no movement before they wake up. How long a person has been asleep, how deep their sleep is, and when they are expected to wake up (combined with factors such as MR noise level) can be a better predictor of future movement.
[0145] Therefore, this implies a complex relationship between anxiety and movement, and it is possible that reducing anxiety can increase the likelihood of movement (e.g., when transitioning from a "petrified" state to a slightly less anxious state) or decrease the likelihood of movement, depending on multiple factors.
[0146] However, the inventors determined that it is possible to use sensors to monitor the patient and train a machine learning algorithm using scan parameters from current and future scans, as well as patient-related information, to determine how stress levels will develop and how the likelihood of patient movement will develop.
[0147] Therefore, the inventors recognized the AI-based tools developed in recent years that utilize sensor information to identify human emotions. These include facial emotion recognition (e.g., https: / / azure.microsoft.com / en-us / services / cognitive-services / emotion / ), voice emotion recognition (e.g., http: / / www.good-vibrations.nl / , https: / / vokaturi.com / ), and emotion recognition via other physiological sensors (see, for example, https: / / www.wareable.com / health-and-wellbeing / future-of-emotion-sensing-wearables-111, https: / / biosay.com / ). The inventors leveraged their understanding of these technologies, combined with MRI scanner scan information using specialized predictive algorithms, patient background information, and additional sensor combinations, to help detect and respond to emergencies during MRI examinations and increase patient comfort and safety. In this way, the patient's emotional and physiological state is actually analyzed and its development is predicted. Real-time feedback is provided to technical experts, and automated decisions can be made regarding when to stop the scan based on the patient's current and predicted emotional and physiological state. In this way, the patient experience is improved (less anxiety or pain), patients are safer due to the system's automated emergency stop, and the workflow can be improved because the scan can be stopped and repeated earlier.
[0148] Data collected prior to the current MRI examination is used to determine the predicted stress level and the predicted level of motor function, as discussed above. A specific method for collecting this information, or at least a portion of it, relevant to the patient, is described in detail below. The collected data is referred to as “preparation data.”
[0149] Prepare data
[0150] Prior to an MRI scan, patients receive online information and training (e.g., in the form of a smartphone app) to help them prepare for the upcoming MRI scan at home or in the hospital. This app may include several elements such as a) patient-specific information; b) training exercises; c) questionnaires about the patient; and d) measurements of their heart rate and respiratory rate, skin color, and / or reaction speed.
[0151] Patient-specific information may include text, video, or audio, which provides patients with information about an upcoming scan or medical procedure (e.g., a video with information about an MRI scan, the MRI machine, and what is happening).
[0152] Training exercises provide participants with specific guidance or training to improve their skills that allow them to go through medical procedures; among other things, exercises include improving the ability to remain still for extended periods of time and / or the ability to hold their breath / follow a specific breathing pattern for a certain period of time.
[0153] The questionnaire will include demographic questions, such as about the patient's age, number of previous scans, education level, etc., as well as other relevant questions (e.g., about the patient's claustrophobia, anxiety level, etc.). Additionally, other questionnaire inputs can be collected, such as self-reports. For example, the patient's perception of the upcoming scan and their own expectations. Finally, each piece of information or activity can be followed by questions related to the patient's perception of the scan.
[0154] Heart rate and respiratory rate measurements can be collected using a mobile phone. This can provide an indication of a person's baseline heart rate, possibly before and after the user accesses training materials and questionnaires. Such measurements can also be followed based on the patient's perception and expectations. Heart rate and respiratory rate measurements can also be performed only before the examination and compared with previously conducted baseline measurements.
[0155] Skin color can be determined using a telephone, such as by measuring facial areas before or only before the examination.
[0156] By creating a small game app, it is possible to determine a patient's reaction time at home and when they are sitting in the waiting room.
[0157] In this way, data preparation can include:
[0158] Usage data from mobile applications (e.g., number of logins, time spent in the application);
[0159] Data from questionnaires (e.g., questionnaires about patients' age, education level, anxiety, and perceptions and expectations);
[0160] Training-specific data (e.g., the amount of time participants were able to remain still while training at home; the number of seconds they could hold their breath; their heart rate, as measured by the app; the number of times they used different features of the app).
[0161] Physiological parameters (changes in heart rate or respiratory rate, changes in skin color, and differences in reaction time).
[0162] Training machine learning algorithms
[0163] Regarding the implementation of one or more machine learning algorithms, as discussed above, there are two distinct parts to the implementation:
[0164] Machine learning algorithms (classical SVM, decision trees, or neural networks) can be used to determine the current stress level and the current level of mobility based on sensor data and other information (not yet predicting the future). In this case, the algorithm would be trained using historical data labeled with subjective feedback or objective measurements from the patient, such as the frequency of patient-initiated scan interruptions or severe motion artifacts, or other information as discussed above.
[0165] Alternatively, stress levels and / or mobility status levels can be calculated from sensor data using known analytical functions.
[0166] For predicting the future development of stress levels and mobility levels, an implementation using recurrent neural networks (RNNs) has been found particularly suitable. Specifically, a long short-term memory (LSTM) implementation has been found well-suited for this purpose. This machine learning implementation is fed both background information (patient condition, scan type, etc.) and the development of stress levels (up to the present moment) as a function of time, and / or the development of mobility levels as a function of time. The network then predicts how stress levels will continue in the future, and how mobility levels will continue into the future. To train such a network, the temporal evolution of stress levels and mobility levels, along with background information for multiple subjects, needs to be recorded and used as training data.
[0167] It should be noted that the two parts discussed above can be operated individually for predicting stress level and mobility state level, or in combination. Therefore, at point 1, there can be an algorithm for determining the current stress level and different algorithms for determining the current mobility state level. Then at point 2, there can be an algorithm for predicting the future stress level and different algorithms for predicting the future mobility state level. However, the same algorithm at point 1 can determine both the stress level and the mobility state level, and the same algorithm at point 2 can predict both the future stress level and the mobility state level.
[0168] In another exemplary embodiment, a computer program or computer program unit is provided, characterized in that it is configured to perform method steps of the method according to one of the foregoing embodiments on a suitable device or system.
[0169] The computer program unit can therefore be stored on the computer unit, and it can also be part of the embodiments. The computing unit can be configured to perform or induce the execution of the steps of the methods described above. Furthermore, it can be configured to operate components of the apparatus and / or system described above. The computing unit can be configured to automatically operate and / or execute user commands. The computer program can be loaded into the working memory of the data processor. Therefore, the data processor can be equipped to execute the method according to one of the foregoing embodiments.
[0170] This exemplary embodiment of the invention covers both computer programs that use the invention from the outset and computer programs that convert existing programs into programs that use the invention through updates.
[0171] Furthermore, the computer program unit is capable of providing all the necessary steps of the process for implementing exemplary embodiments of the methods as described above.
[0172] According to another exemplary embodiment of the present invention, a computer-readable medium, such as a CD-ROM, a USB stick, etc., is provided, wherein the computer-readable medium has computer program units stored on the computer-readable medium, the computer program units being described in the preceding portion.
[0173] Computer programs can be stored / distributed on suitable media, such as optical storage media or solid-state media provided with or as part of other hardware, but computer programs can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.
[0174] However, the computer program may also exist on a network such as the World Wide Web and can be downloaded from such a network to the working memory of a data processor. According to another exemplary embodiment of the invention, a medium is provided for making a computer program unit available for download, wherein the computer program unit is arranged to perform the method described in one of the previously described embodiments according to the invention.
[0175] It must be noted that embodiments of the present invention are described with reference to different subjects. Specifically, some embodiments are described with reference to claims of the method type, while others are described with reference to claims of the device type. However, those skilled in the art will understand from the above and below description that, unless otherwise indicated, any combination of features relating to different subjects, in addition to any combination of features belonging to one type of subject, is also considered to be disclosed in this application. However, all features can be combined to provide synergistic effects beyond the simple sum of the features.
[0176] Although the invention has been described and illustrated in detail in the accompanying drawings and the foregoing description, such description and illustration are to be considered illustrative or exemplary rather than restrictive. The invention is not limited to the disclosed embodiments. Other variations of the disclosed embodiments will be understood and implemented by those skilled in the art in practicing the claimed invention by studying the drawings, description, and dependent claims.
[0177] In the claims, the word "comprising" does not exclude other units or steps, and the words "a" or "an" do not exclude multiple. A single processor or other unit may perform the functions of several items recited in the claims. Although specific measures are recited in different dependent claims, this does not indicate that combinations of these measures cannot be advantageously used. Any reference numerals in the claims should not be construed as limiting the scope.
Claims
1. A device (10) for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan, the device comprising: Input unit (20); Processing unit (30); as well as Output unit (40); The input unit is configured to provide the processing unit with at least one sensor data of a patient undergoing an MRI scan by an MRI scanner; The input unit is configured to provide the processing unit with at least one scanning parameter of the MRI scanner for the MRI scan; The input unit is configured to provide the processing unit with at least one characteristic of the patient; The processing unit is configured to implement at least one machine learning algorithm to predict the patient's future stress level and / or predict the patient's future mobility probability, the prediction including utilizing the patient's at least one sensor data, the MRI scanner's at least one scan parameter, and the patient's at least one characteristic, wherein the at least one machine learning algorithm is trained based on: at least one sensor data of one or more reference patients undergoing one or more reference MRI scans on one or more reference MRI scanners; at least one scan parameter of the reference MRI scanner used for the one or more reference MRI scans; at least one characteristic of each of the reference patients; at least one stress level and / or mobility probability level experienced by the one or more reference patients during the one or more reference MRI scans; and The output unit is configured to output information relating to the predicted future stress level of the patient and / or the predicted future mobility probability of the patient.
2. The apparatus according to claim 1, wherein, The at least one sensor data of the patient and the one or more reference patients is acquired by one or more of the following: camera, microphone, skin resistance sensor, skin temperature sensor, skin humidity sensor, skin accelerometer, pulse sensor, respiration sensor, radio frequency radar sensor, EEG sensor, pressure sensor, weight sensor.
3. The apparatus according to claim 1 or 2, wherein, The at least one sensor data of the patient and the one or more reference patients includes one or more of the following: respiratory rate data, heart rate data, voice data, skin resistance data, skin temperature data, skin humidity data, skin movement data, body part movement data, blink rate data, EEG data, information related to what is shown on the intracavitary display, and weight distribution on the intracavitary examination table.
4. The apparatus according to claim 1 or 2, wherein, The at least one scanning parameter of the MRI scanner and the reference MRI scanner includes one or more of the following: scan duration, remaining scan duration, current gradient intensity, future gradient intensity, contrast type, timing parameter, SAR (RF setting), and k-space sampling mode.
5. The apparatus according to claim 1 or 2, wherein, The at least one characteristic of the patient and each of the reference patients includes one or more of the following: age, weight, body mass index, information about a previous diagnosis, the patient's physical condition, the patient's mental condition, and completed questionnaire information.
6. The apparatus according to claim 1 or 2, wherein, The input unit is configured to provide the processing unit with information about one or more previous MRI scans performed on the patient, and wherein the prediction of the patient's stress level and / or the patient's likelihood of movement includes utilizing the information about one or more previous MRI scans performed on the patient.
7. The apparatus according to claim 1 or 2, wherein, The processing unit is configured to determine whether the predicted future stress level of the patient and / or the predicted future mobility probability of the patient will exceed a stress threshold level or a mobility probability threshold level, and wherein the information relating to the predicted future stress level of the patient and / or the predicted future mobility probability of the patient includes an indication of whether either threshold is predicted to be exceeded.
8. An imaging system (100), comprising: Magnetic resonance imaging scanner (110); At least one sensor (120); as well as The device (10) for monitoring a patient undergoing magnetic resonance imaging (MRI) scans according to any one of claims 1-7; The at least one sensor is configured to provide at least one sensor data of a patient undergoing an MRI scan to the processing unit of the device; The device is configured to automatically stop the MRI scan based on information related to the predicted stress level of the patient and / or the predicted likelihood of the patient's movement.
9. A method for monitoring a patient undergoing a magnetic resonance imaging (MRI) scan, the method comprising: Provide the processing unit with at least one sensor data of a patient who has undergone an MRI scan by an MRI scanner; The processing unit is provided with at least one scanning parameter of the MRI scanner for the MRI scan; The processing unit is provided with at least one characteristic of the patient; A processing unit implementing at least one machine learning algorithm predicts the future stress level and / or the future mobility probability of the patient, the prediction comprising utilizing at least one sensor data of the patient, at least one scan parameter of the MRI scanner, and at least one characteristic of the patient, wherein the at least one machine learning algorithm is trained based on one or more of the following: at least one sensor data of one or more reference patients undergoing one or more reference MRI scans on one or more reference MRI scanners; at least one scan parameter of the reference MRI scanner used for the one or more reference MRI scans; at least one characteristic of each of the reference patients; at least one stress level and / or mobility probability level of the one or more reference patients; and The output unit outputs information related to the predicted future stress level of the patient and / or the predicted future mobility probability of the patient.
10. A computer program unit for controlling an apparatus according to any one of claims 1 to 7 and / or a system according to claim 8, the computer program unit being configured, when run by a processor, to perform the method according to claim 9.