Time-series signal prediction device, radiation therapy device, time-series signal prediction method, program, and storage medium

Abstract

A time-series signal prediction device according to an embodiment includes a time-series signal acquisition unit, a block conversion unit, a feature vector acquisition unit, and a feature vector conversion unit. The time-series signal acquisition unit acquires a time-series signal relating to biological information of a patient. The block conversion unit divides the time-series signal into partial signals at predetermined intervals, and converts the partial signals into block signals having a shorter time series than the time series of the time-series signal. The feature vector acquisition unit outputs, from the block signal, a feature vector indicating a time-series signal at a more future time than when the time-series signal is acquired. The feature vector conversion unit converts the feature vector into a prediction signal relating to biological information of the patient.

Description

Technical Field

[0001] The embodiments of the present invention relate to a time series signal prediction device, a radiation therapy device, a time series signal prediction method, a program, and a storage medium. Background Technology

[0002] Radiation therapy is a treatment method that uses radiation to destroy lesions in a patient's body. However, if the lesions are not accurately located, even normal tissues can be damaged. Therefore, CT scans are performed beforehand to determine the location of lesions in three dimensions, allowing for the planning of the direction and intensity of radiation to minimize irradiation of normal tissues.

[0003] In radiotherapy, to ensure radiation is delivered according to the plan, the patient's position needs to be aligned both during the treatment plan and during treatment. To ensure the location of lesions, bones, etc., within the patient is consistent with the treatment plan, the following method is used: A fluoroscopic image of the patient lying on the examination table in the treatment room before treatment is performed, and a DRR (Digitally Reconstructed Radiograph) image, which is virtually reconstructed from a 3D CT image taken during the treatment plan, is compared. The patient's positional shift between the images is determined, and the examination table is moved based on this shift. The patient's positional shift is then determined by searching for the position of the CT image reconstructed from the DRR that most closely resembles the fluoroscopic image. Although many methods have been proposed to automate this search using computers, the search results still need to be confirmed by radiotherapy practitioners through comparison of the fluoroscopic image and the DRR.

[0004] A known method of respiratory-synchronized irradiation involves determining the location of a tumor, such as in the lungs or liver, which move due to respiration and heartbeat, during irradiation. The irradiation is performed when the determined location matches the location determined by the treatment plan. Methods for estimating tumor location include sensing the movement of the chest and abdominal body surfaces that expand / contract with respiration and estimating the tumor location based on sensor values. Alternatively, methods include capturing fluoroscopic images of the irradiated patient and tracking the tumor location within the fluoroscopic images. Hereinafter, the signal obtained from body surface movement, the trajectory of the tumor location, will be referred to as a "time-series signal." While controlling the timing of the irradiation of the treatment beam can reduce unnecessary irradiation of the patient, the need for tumor location prediction arises due to the computational time and response time of devices required for tumor location estimation.

[0005] For example, in the method disclosed in Non-Patent Document 1, LSTM (Long Short Term Memory), a type of recursive network, is used to learn time-series signals from multiple patients to predict signal values ​​at a specified time. Furthermore, the method disclosed in Non-Patent Document 2 differs from that of Non-Patent Document 1 by using CNN (Convolutional Neural Network), a type of feedforward network. This disclosed technique predicts signal values ​​at more future times (T+M, M > 0) based on time-series signals up to a certain time (T). This method can expect higher accuracy as the input time-series signal length increases; however, this also increases computation time, making it sometimes unsuitable for prediction. Moreover, compared to teaching only the signal value at time T+M, teaching the waveform signal from time T+1 to T+M increases the amount of information used for learning, thus expecting improved prediction accuracy.

[0006] Existing technical documents Non-patent literature Non-patent literature 1: "Towards real-time respiratory motion prediction based on long short-term memory neural networks", Phys Med Biol. 2019 Apr 10; 64(8): 085010. Non-patent document 2: "Respiratory Prediction Based on Multi-Scale TemporalConvolutional Network for Tracking Thoracic Tumor Movement", Front.Oncol., 27May 2023 Summary of the Invention

[0007] The problem that the invention aims to solve The problem to be solved by the present invention is to provide a time series signal prediction device, a radiotherapy device, a time series signal prediction method, a program, and a storage medium capable of making high-speed and high-precision predictions of time series signals related to the patient's biological information.

[0008] Methods for solving problems The time-series signal prediction apparatus of this embodiment includes a time-series signal acquisition unit, a block conversion unit, a feature vector acquisition unit, and a feature vector conversion unit. The time-series signal acquisition unit acquires a time-series signal related to the patient's biological information. The block conversion unit divides the time-series signal into partial signals at predetermined intervals, converting them into block signals of time series shorter than the original time-series signal. The feature vector acquisition unit acquires feature vectors from the block signals, the feature vectors representing time-series signals at more future times than when the time-series signal was acquired. The feature vector conversion unit converts the feature vectors into a prediction signal related to the patient's biological information.

[0009] Invention Effects According to the present invention, a time series signal prediction device, a radiotherapy device, a time series signal prediction method, a program, and a storage medium are provided that can perform high-speed and high-precision prediction of time series signals related to the patient's biological information. Attached Figure Description

[0010] Figure 1 This is a block diagram illustrating a schematic configuration of a treatment system equipped with the time-series signal prediction device of the first embodiment.

[0011] Figure 2 This is a block diagram showing the general configuration of the time series signal prediction device 100 according to the first embodiment.

[0012] Figure 3 This is a diagram showing an example of a waveform signal acquired by the time series signal acquisition unit 110.

[0013] Figure 4 This diagram illustrates an example of the conversion from a time-series signal to a block signal performed by the block conversion unit 120.

[0014] Figure 5 This is a diagram illustrating another example of the conversion from a time-series signal to a block signal performed by the block conversion unit 120.

[0015] Figure 6 This is a diagram illustrating an example of how the characteristic vector acquisition unit 130 acquires a characteristic vector.

[0016] Figure 7 This is a diagram illustrating another example of how the feature vector acquisition unit 130 acquires feature vectors.

[0017] Figure 8 This is a flowchart illustrating an example of the processing flow performed by the time series signal prediction device 100 of the first embodiment.

[0018] Figure 9This is a block diagram showing the general configuration of the time series signal prediction device 200 according to the second embodiment.

[0019] Figure 10 This is a flowchart illustrating an example of the processing flow performed by the time series signal prediction device 200 of the second embodiment.

[0020] Figure 11 This is a block diagram showing the general configuration of the time series signal prediction device 300 according to the third embodiment.

[0021] Figure 12 This is a flowchart illustrating an example of the processing flow performed by the time series signal prediction device 300 according to the third embodiment. Detailed Implementation

[0022] Hereinafter, the time series signal prediction apparatus, radiotherapy apparatus, time series signal prediction method, program, and storage medium according to embodiments will be described with reference to the accompanying drawings.

[0023] (First Implementation) (Overall composition) Figure 1 This is a block diagram illustrating a schematic configuration of a treatment system equipped with the time-series signal prediction device of the first embodiment. The treatment system 1 includes, for example, a treatment device 10, a sensor S, and a time-series signal prediction device 100. The treatment device 10 includes, for example, an examination table 12, a computed tomography (CT) device 14 (hereinafter referred to as "CT imaging device 14"), and a treatment beam irradiation gate 16.

[0024] The examination table 12 is a movable treatment table that holds a patient (P) receiving radiation-based treatment in a lying position using, for example, a fixation device. The examination table 12 moves within a ring-shaped CT imaging device 14 with an opening, while the patient P is fixed in place, under control from an examination table control unit (not shown). Based on a movement signal, the examination table control unit controls a translation mechanism and a rotation mechanism provided on the examination table 12 to change the direction of the treatment beam B irradiated onto the patient P fixed to the examination table 12. The translation mechanism can drive the examination table 12 in three-axis directions, and the rotation mechanism can drive the examination table 12 around three axes. Therefore, the examination table control unit, for example, controls the translation and rotation mechanisms of the examination table 12 to move the examination table 12 in six degrees of freedom. The degrees of freedom of the examination bed control unit to control the examination bed 12 may not be six degrees of freedom, but may be fewer than six degrees of freedom (e.g., four degrees of freedom) or more than six degrees of freedom (e.g., eight degrees of freedom).

[0025] The CT imaging apparatus 14 is an imaging device for performing three-dimensional computed tomography. Multiple radiation sources are arranged inside a circular opening of the CT imaging apparatus 14, and radiation for imaging the body of patient P is irradiated from each radiation source. That is, the CT imaging apparatus 14 irradiates radiation from multiple locations around patient P. The radiation irradiated from each radiation source in the CT imaging apparatus 14 is, for example, X-rays. The CT imaging apparatus 14 detects radiation irradiated from corresponding radiation sources and reaching the body of patient P by multiple radiation detectors arranged inside the circular opening. Based on the energy of the radiation detected by each radiation detector, the CT imaging apparatus 14 generates a CT image obtained by imaging the body of patient P. The CT image of patient P generated by the CT imaging apparatus 14 is a three-dimensional digital image in which the energy of the radiation is represented by digital values. For example, the three-dimensional imaging of the body of patient P in the CT imaging apparatus 14, i.e., the irradiation of radiation from each radiation source, and the generation of CT images based on the radiation detected by each radiation detector, are controlled by a imaging control unit (not shown).

[0026] The therapeutic beam irradiation gate 16 irradiates a therapeutic beam B, which is used to destroy a tumor (lesion) located in the body of patient P that is the target of treatment. The therapeutic beam B is, for example, X-rays, gamma rays, an electron beam, a proton beam, a neutron beam, a heavy particle beam, etc. The therapeutic beam B is directed linearly from the therapeutic beam irradiation gate 16 to patient P (more specifically, the tumor within patient P's body). The irradiation of the therapeutic beam B in the therapeutic beam irradiation gate 16 is controlled, for example, by a therapeutic beam irradiation control unit (not shown).

[0027] In this method, radiotherapy is performed by irradiating the tumor in patient P's body with a treatment beam B through a treatment beam irradiation gate 16. However, when the tumor, which is the lesion of patient P, is located in an organ that moves due to respiratory or cardiac activity, such as the lungs or liver, it is necessary to irradiate the tumor with X-rays from the CT imaging device 14 and the treatment beam B from the treatment beam irradiation gate 16 at the same time when the tumor is within the irradiation range of the treatment beam B. However, in the prior art, for example, due to gaps in device response, sometimes X-rays and the treatment beam B are irradiated even when the tumor is outside the irradiation range of the treatment beam B, resulting in radiation that is unnecessary for patient P. Against this background, the present invention utilizes a sensor S and a time-series signal prediction device 100 to predict the timing of the tumor being within the irradiation range of the treatment beam B at high speed and with high accuracy.

[0028] Sensor S, for example, intermittently senses the movement of the chest and abdominal body surface that expands / contracts with the patient P's breathing, and outputs a one-dimensional waveform signal with the obtained sensor value as the amplitude to the time series signal prediction device 100. Sensor S can be a contact sensor placed on the upper abdominal surface of the patient P lying on the examination table 12, or a non-contact sensor that measures the distance to the chest and abdomen using a laser.

[0029] Figure 2 This is a block diagram illustrating the schematic configuration of the time series signal prediction apparatus 100 according to the first embodiment. The time series signal prediction apparatus 100 includes, for example, a time series signal acquisition unit 110, a block conversion unit 120, a feature vector acquisition unit 130, and a feature vector conversion unit 140. Some or all of the components of the time series signal prediction apparatus 100 are implemented by executing programs (software) using a hardware processor such as a CPU (Central Processing Unit). Some or all of these components can be implemented using hardware (including circuitry) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and GPU (Graphics Processing Unit), or through a combination of software and hardware. Some or all of the functions of these components can also be implemented using a dedicated LSI. The program can be pre-saved in a storage device such as ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), or flash memory (a storage device with non-transient storage media) provided in the time series signal prediction device 100. Alternatively, it can be saved in a removable storage medium such as a DVD or CD-ROM (a non-transient storage medium). The program is then installed into the HDD or flash memory of the time series signal prediction device 100 by mounting the storage medium onto the drive device provided in the time series signal prediction device 100. The program can also be downloaded from other computer devices via a network and installed into the HDD or flash memory of the time series signal prediction device 100.

[0030] The time series signal acquisition unit 110 acquires, for example, a time series signal representing a waveform from the sensor S via wireless communication. Figure 3This is a diagram illustrating an example of a waveform signal acquired by the time-series signal acquisition unit 110. Figure 3 In the chart, the horizontal axis represents time, and the vertical axis represents amplitude. Figure 3 In the waveform shown, the peak P1, which becomes a peak, represents the point where patient P has the greatest inspiration, and the peak P2, which becomes a trough, represents the point where patient P has the greatest expiration. The time series signal acquisition unit 110 outputs the acquired time series signal to the block conversion unit 120.

[0031] As another example, the time-series signal acquired by the time-series signal acquisition unit 110 may also be a trajectory representing the position of the tumor as it moves with the patient P's breathing. Here, the trajectory representing the position of the tumor refers to a time-series signal representing the temporal change of the center of gravity position of the tumor within the fluoroscopic dynamic image of the patient P. In this case, the time-series signal acquisition unit 110 may also acquire the time-series signal from the CT imaging device 14 instead of from the sensor S.

[0032] Furthermore, as another example, the time-series signal acquired by the time-series signal acquisition unit 110 may be a time-series signal representing the temporal changes of the diaphragm of patient P, or it may be the trajectory of a marker placed near the tumor in patient P's body that moves with respiration. The trajectory of the marker is a time-series signal representing the temporal changes of the marker's position within the patient's fluoroscopic dynamic image. The position of the tumor, diaphragm, and marker (hereinafter, the tracking object) within the fluoroscopic dynamic image is estimated at each time step, for example, by template matching using images of the tracking object taken before tracking as templates. Alternatively, for example, the position of the tracking object at each time step is estimated based on feature quantities obtained by transforming images of the tracking object taken using a pre-prepared model. In this case, the model can be constructed using machine learning that learns from multiple images of the tracking object taken before tracking.

[0033] The block conversion unit 120 divides the time series signal acquired by the time series signal acquisition unit 110 into partial signals at each predetermined interval, and converts them into block signals of time series shorter than the time series of the time series signal. Figure 4 This diagram illustrates an example of the conversion from a time-series signal to a block signal performed by the block conversion unit 120. As an example, Figure 4 This illustrates the conversion of a time series signal consisting of 12 sequences of signal values ​​y(t-11) to y(t) from time t-11 to time t (i.e., the current time) into a block signal of 3 sequences. The partial signals at this point are vector data consisting of signal values ​​y(t-3) to y(t) from time t-3 to time t, signal values ​​y(t-7) to y(t-4) from time t-7 to time t-4, and signal values ​​y(t-11) to y(t-8) from time t-11 to time t-8.

[0034] Figure 5 This diagram illustrates another example of the conversion from a time-series signal to a block signal performed by the block conversion unit 120. As an example, Figure 5 This illustrates the conversion of eight sequences of signal values ​​y(t-7) to y(t), consisting of eight values ​​from time t-7 to time t, into three block signals. In this case, some signals are vector data consisting of signal values ​​y(t-3) to y(t) from time t-3 to time t, signal values ​​y(t-5) to y(t-2) from time t-5 to time t-2, and signal values ​​y(t-7) to y(t-4) from time t-7 to time t-4. This allows for the repetition of signal values ​​between sequences without altering the order of the time series.

[0035] The feature vector acquisition unit 130 acquires feature vectors representing a future time series signal from the block signal converted by the block conversion unit 120. More specifically, in order to predict a future time series signal based on the input time series signal, the feature vector acquisition unit 130 is configured in advance as a predictor, which is obtained by statistically modeling the changes in the signal based on time series signals related to the biological information of multiple patients.

[0036] As a method for modeling the feature vector acquisition part 130, there is, for example, a method for constructing a state-space model. Specifically, if time t and time t are used to model the feature vector acquisition part 130, then... The two waveform values ​​y(t) and When plotted on a two-dimensional coordinate system, the breathing waveform, being periodic, exhibits elliptical motion at the points plotted along with the time interval. This characteristic is represented using a state-space model. Since this model is nonlinear, an Extended Kalman Filter (EKF) is used, approximating the Kalman filter in a manner corresponding to the nonlinear state-space model. The eigenvector acquisition unit 130 is constructed by optimizing the parameters of this state-space model using collected time-series signals.

[0037] Furthermore, as another method for modeling the feature vector acquisition part 130, there is a method utilizing recurrent neural networks. Here, a recurrent neural network refers to a neural network model that obtains its output by causing the internal state to change through sequential input of multiple sequences of input data. As a representative example, there is LSTM (Long Short Term Memory).

[0038] Furthermore, as another method for modeling the feature vector acquisition unit 130, a feedforward neural network can be used. Here, a feedforward neural network refers to a neural network model that inputs input data together without dividing it into sequences and obtains the output. For example, when the feedforward network is a convolutional neural network (CNN), the block signal is treated as two-dimensional data for processing when it is input. Alternatively, the sequence of block signals can be input into the network as data for different channels.

[0039] When using these neural network models, the feature vector acquisition unit 130 is constructed by learning the model's intrinsic parameters through deep learning of the collected time-series signals. At this time, the collected time-series signals are allowed to be overlapped and divided into multiple partial signals. For each partial signal, an input teaching block signal and a target feature vector are separated. The data included in the target feature vector consists only of signals from moments more in the future than the teaching block signals.

[0040] The aforementioned learning refers to adjusting internal parameters to make the feature vector obtained as the output when the teach block signal is input to the model approximate the target feature vector. A function measuring the distance between the target feature vector and the target feature vector is used as the error function. To minimize the error function, optimization methods such as backpropagation are employed.

[0041] Figure 6 This is a diagram illustrating an example of how the characteristic vector is obtained by the characteristic vector acquisition unit 130. For example... Figure 6 As shown, the feature vector is, for example, a vector whose elements are the amplitude values ​​of the aforementioned partial signals. That is, when predicting the signal value at time t+M (M > 0, M is a natural number) for the teaching block signal up to time t, the target feature vector becomes a vector whose elements are the signal values ​​y(t+1) to y(t+M) from time t+1 to time t+M.

[0042] Figure 7 This is a diagram illustrating another example of how the characteristic vector is obtained by the characteristic vector acquisition unit 130. For example... Figure 7 As shown, the feature vector can also be a vector whose elements are the differences in amplitude values ​​of the aforementioned partial signals. That is, when predicting the signal value y(t+M) at time t+M (M > 0, M is a natural number) for the teach block signal up to time t, the target feature vector becomes a vector that starts from the teach block signal at time t and whose elements are the differences d(t+1) to d(t+M) in signal values ​​from time t+1 to time t+M.

[0043] The feature vector conversion unit 140 converts the feature vectors acquired by the feature vector acquisition unit 130 into predictive signals related to the biological information of the patient P. More specifically, for example, in the feature vector acquisition unit 130... Figure 6 In the case of the eigenvector shown, the signal value y(t+M) at time t+M is obtained from the eigenvector as the prediction signal. Furthermore, for example, in the case obtained by the eigenvector acquisition unit 130... Figure 7 In the case of the eigenvectors shown, the eigenvector conversion unit 140 obtains the signal value y(t) + d(t+1) + ... d(t+M) obtained by adding the sum of the elements d(t+1) to d(t+M) of the eigenvectors from time t+1 to time t+M to the signal value y(t+M) at time t+M, and uses this as the signal value y(t+M) at time t+M. The eigenvector conversion unit 140 outputs the calculated prediction signal to the monitor or the like of the time series signal prediction device 100. The operator of the treatment device 10 checks the prediction signal output to the monitor or the like, determines the timing of X-ray dynamic image acquisition and the timing of treatment beam B irradiation, and performs X-ray dynamic image acquisition and treatment beam B irradiation.

[0044] Furthermore, the feature vector conversion unit 140 calculates the reliability of the time series signal based on the feature vectors acquired by the feature vector acquisition unit 130. For example, the more discontinuous the elements of the feature vector in the time series signal, the smaller the reliability value calculated by the feature vector conversion unit 140. Here, discontinuity can be determined, for example, based on the approximate error between the approximate curve obtained from the multiple elements included in the feature vector using methods such as spline interpolation and these multiple elements. In this case, the feature vector conversion unit 140 can also output the predicted signal and the calculated reliability to the monitor of the time series signal prediction device 100, etc. In this case, the operator of the treatment device 10 can consider the reliability of the predicted signal and utilize the output predicted signal for the acquisition of X-ray dynamic images or the irradiation of the treatment beam B.

[0045] [Processing flow] Next, refer to Figure 8 The process flow performed by the time series signal prediction device of the first embodiment is described. Figure 8 This is a flowchart illustrating an example of the processing flow performed by the time series signal prediction apparatus of the first embodiment.

[0046] First, the time-series signal acquisition unit 110 acquires a time-series signal related to the patient P's biological information from the sensor S (step S100). Next, the block conversion unit 120 divides the acquired time-series signal into partial signals and acquires them as block signals (step S102). Next, the feature vector acquisition unit 130 acquires feature vectors representing future time-series signals from the acquired block signals (step S104). Next, the feature vector conversion unit converts the acquired feature vectors into a prediction signal related to the patient P's biological information (step S106). Thus, the processing of this flowchart ends.

[0047] According to the first embodiment described above, by converting the input time-series signal into a block signal, the input length is shortened while maintaining the amount of information in the input data. Therefore, compared to the prior art of sequentially inputting time-series signals, the time spent on learning, and calculating, particularly recurrent neural network models, can be reduced. Furthermore, according to the first embodiment, compared to the technique described in Non-Patent Document 1, which only teaches the signal values ​​to be predicted during model learning, the first embodiment increases the amount of information to be learned by learning multiple signal values ​​up to the time to be predicted. Therefore, the accuracy of the feature vector output by the predictor can be improved.

[0048] (Second Implementation) (Overall composition) In the first embodiment, a method for predicting a time-series signal at a more future time based on a time-series signal inputted with biological information about patient P is described. As an example of biological information about patient P, regarding respiratory signals, since the amount of air taken in during inspiration varies in the same patient P, the amplitude of the signal during complete inspiration tends to become irregular. On the other hand, the amplitude during complete exhalation is stable, thus tending to become regular. Taking advantage of this tendency, in radiotherapy, it is preferable to avoid capturing dynamic X-ray images of patient P during inspiration or exhalation and to avoid irradiation by the treatment beam B, and to set the tumor location at the time of complete exhalation as the irradiation point. Against this background, the time-series signal prediction device 200 of the second embodiment can predict the time when the tumor location enters the irradiation point, and avoid capturing dynamic X-ray images and irradiation by the treatment beam B when the patient P is inhaling, exhaling, or experiencing an abnormal respiratory state.

[0049] Figure 9This is a block diagram showing the schematic configuration of the time series signal prediction apparatus 200 according to the second embodiment. The time series signal prediction apparatus 200 includes, for example, a time series signal acquisition unit 210, a block conversion unit 220, a feature vector acquisition unit 230, a feature vector conversion unit 240, and a state estimation unit 250. The configurations of the time series signal acquisition unit 210, block conversion unit 220, feature vector acquisition unit 230, and feature vector conversion unit 240 are the same as those of the time series signal acquisition unit 110, block conversion unit 120, feature vector acquisition unit 130, and feature vector conversion unit 140 of the time series signal prediction apparatus 100 in the first embodiment, therefore, descriptions are omitted. Furthermore, in the second embodiment, Figure 1 The same applies to the configuration of the treatment system 1 shown.

[0050] The state estimation unit 250 estimates the state of patient P based on the prediction signal obtained from the feature vector conversion unit 240 and outputs it to the monitor of the time series signal prediction device 200. Here, the estimated patient state refers, for example, to a state of abnormal breathing due to coughing or other reasons. Before and after coughing or sneezing, breathing becomes disordered and differs from the normal state, thus increasing the difference between the predicted value and the measured value. The state estimation unit 250 may also output the estimated patient state to the monitor of the treatment device 10.

[0051] For example, the state estimation unit 250 can also calculate the absolute value of the difference between the predicted value obtained before a specified time (e.g., 100 milliseconds ago) and the measured value at that time. If the difference is greater than a specified value, it estimates and outputs that the patient's state is abnormal. Furthermore, for example, the state estimation unit 250 can also calculate the absolute value of the difference between the predicted value obtained before a specified time and the measured value at that time. If the state, which is greater than a specified value, persists for a specified period, it estimates and outputs that the patient's state is abnormal. Furthermore, for example, the state estimation unit 250 can also calculate the absolute value of the difference between the predicted value obtained before a specified time and the measured value at that time within multiple time windows. If the number of time windows in which the calculated absolute value is greater than a specified value is more than half, it estimates and outputs that the patient's state is abnormal.

[0052] Furthermore, as another example, the patient's state can be the respiratory phase state near the moment when the patient's breathing changes from exhalation to inhalation, that is, near the moment after complete exhalation. When the amplitude of the respiratory waveform increases during inhalation and decreases during exhalation, the moment when the amplitude value becomes smaller than a certain threshold is considered the aforementioned respiratory phase. This threshold is, for example, preset by the user based on the respiratory waveforms of the same patient over several past respiratory volumes, as monitored and displayed on a monitor such as the time-series signal prediction device 200. For example, the user can set the threshold to a predetermined level (e.g., 20%) of the maximum amplitude in the time-series signal acquired by the time-series signal acquisition unit 110. Previously, this referred to, for example, the period from the end of patient positioning to the start of treatment. Instead, the state estimation unit 250 can automatically determine the maximum amplitude in the time-series signal acquired by the time-series signal acquisition unit 110 without requiring the user to set the threshold, and set the threshold to the predetermined level of the determined maximum amplitude. In this case, the state estimation unit 250 functions as a "parameter setting unit."

[0053] When the predicted signal is the trajectory of a tracking object such as a tumor, diaphragm, or marker within a fluoroscopic dynamic image, the state estimation unit 250 can also estimate and output the state of the tracking object within the fluoroscopic dynamic image. Tracking of tracking objects such as tumors, diaphragms, and markers within the fluoroscopic dynamic image can sometimes become difficult due to increased noise in the fluoroscopic image, leading to image quality deterioration. Furthermore, sometimes the tracking object may exhibit irregular movements due to some abnormality experienced by the patient P (e.g., coughing or sneezing). Therefore, the state estimation unit 250 can also estimate and output that the state of the tracking object is abnormal when the difference between the predicted and measured values ​​of the predicted signal becomes large. Furthermore, for example, similar to the state estimation of respiratory waveforms, the state estimation unit 250 can also, in the same manner as described above, set a threshold when the amplitude of the time series signal, which is planned to be composed of the trajectory of the tracking object, changes from downward to upward, and if the amplitude of the time series signal is below the threshold, estimate and output that the tumor is within the irradiation point.

[0054] Furthermore, while the above description indicates that the state estimation unit 250 is included in the time series signal prediction device 200, the state estimation unit 250 may also be included in the treatment device 10. In this case, the time series signal prediction device 200 may send the prediction signal obtained from the feature vector conversion unit 240 to the treatment device 10, and the state estimation unit 250 included in the treatment device 10 may estimate the state of the patient P based on the received prediction signal and display it on the monitor of the time series signal prediction device 200, etc.

[0055] [Processing flow] Next, refer to Figure 10 The process flow of the time series signal prediction device of the second embodiment will be described. Figure 10 This is a flowchart illustrating an example of the processing flow performed by the time series signal prediction apparatus of the second embodiment.

[0056] First, the time series signal acquisition unit 210 acquires a time series signal related to the patient P's biological information from the sensor S (step S200). Next, the block conversion unit 220 divides the acquired time series signal into partial signals and acquires them as block signals (step S202). Next, the feature vector acquisition unit 230 acquires feature vectors representing future time series signals from the acquired block signals (step S204). Next, the feature vector conversion unit 240 converts the acquired feature vectors into a prediction signal related to the patient P's biological information (step S206). Next, the state estimation unit 250 estimates the patient P's state based on the acquired prediction signal and outputs it to the monitor of the time series signal prediction device 200 (step S208). Thus, the processing of this flowchart ends.

[0057] According to the second embodiment described above, based on the prediction signal obtained from the biological information of the patient P, the state of the patient P is estimated and output to the monitor of the time series signal prediction device 200, etc. For example, while the operator of the treatment device 10 confirms the state of the patient P output to the monitor, etc., the operator determines the timing of X-ray dynamic image acquisition and the timing of treatment beam irradiation, and performs X-ray dynamic image acquisition and treatment beam irradiation. As a result, radiation accompanying unnecessary X-ray dynamic image acquisition and treatment beam irradiation can be reduced.

[0058] (Third Implementation) In the second embodiment, a method is described for estimating the state of patient P based on a predicted signal obtained from biological information about patient P and outputting it to a monitor of the time-series signal prediction device 200. In this case, the operator of the treatment device 10 manually performs the acquisition of X-ray dynamic images and the irradiation of the treatment beam B while confirming the state of patient P output to the monitor, but human error or timing deviation may occur. Against this background, the time-series signal prediction device 300 of the third embodiment automatically controls the treatment device 10 based on the estimated state of patient P.

[0059] Figure 11This is a block diagram showing the schematic configuration of the time series signal prediction device 300 according to the third embodiment. The time series signal prediction device 300 includes, for example, a time series signal acquisition unit 310, a block conversion unit 320, a feature vector acquisition unit 330, a feature vector conversion unit 340, a state estimation unit 350, and a treatment device control unit 360. The configuration of the time series signal acquisition unit 310, block conversion unit 320, feature vector acquisition unit 330, feature vector conversion unit 340, and state estimation unit 350 is the same as that of the time series signal acquisition unit 210, block conversion unit 220, feature vector acquisition unit 230, feature vector conversion unit 240, and state estimation unit 250 of the time series signal prediction device 200 according to the second embodiment, therefore, description is omitted. Furthermore, in the third embodiment, Figure 1 The same applies to the configuration of the treatment system 1 shown.

[0060] The treatment device control unit 360 controls the treatment device 10 based on the state of the patient P estimated by the state estimation unit 350. More specifically, the treatment device control unit 360, through the state estimation unit 350, initiates the acquisition of X-ray dynamic images of the patient P and / or the irradiation preparation of the treatment beam B from the treatment device 10 when the patient P's state is near the respiratory phase after complete exhalation (in other words, when the amplitude value becomes smaller than a certain threshold). Furthermore, for example, the treatment device control unit 360 may also shorten the acquisition interval of the X-ray dynamic images and initiate the irradiation preparation of the treatment beam B from the treatment device 10 when the state estimation unit 350 estimates that the tumor in the fluoroscopic dynamic image is within the irradiation point.

[0061] On the other hand, if the state estimation unit 350 determines that the patient P's state is abnormal, the treatment device control unit 360 stops the acquisition of X-ray dynamic images of the patient P and / or the irradiation from the treatment beam B of the treatment device 10, or shortens the interval. Furthermore, for example, the treatment device control unit 360 may also stop the acquisition of X-ray dynamic images of the patient P and / or the irradiation from the treatment beam B of the treatment device 10 when the state estimation unit 350 determines that the patient P has completely exhaled and then resumed inhalation. Furthermore, for example, the treatment device control unit 360 may also stop the acquisition of X-ray dynamic images of the patient P and / or the irradiation from the treatment beam B of the treatment device 10 when the state estimation unit 350 determines that the trajectory of the tumor, diaphragm, or other tracked objects in the fluoroscopic dynamic image is abnormal.

[0062] Furthermore, while the above description indicates that the treatment device control unit 360 is included in the time series signal prediction device 300, the treatment device control unit 360 may also be included in the treatment device 10. In this case, the time series signal prediction device 300 may send the state of the patient P estimated by the state estimation unit 350 to the treatment device 10, and the treatment device control unit 360 included in the treatment device 10 may control the treatment device 10 based on the received state of the patient P.

[0063] [Processing flow] Next, refer to Figure 12 The process flow performed by the time series signal prediction device of the third embodiment is explained. Figure 12 This is a flowchart illustrating an example of the processing flow performed by the time series signal prediction apparatus of the third embodiment.

[0064] First, the time-series signal acquisition unit 310 acquires a time-series signal related to the patient P's biological information from the sensor S (step S300). Next, the block conversion unit 320 divides the acquired time-series signal into partial signals and acquires them as block signals (step S302). Next, the feature vector acquisition unit 330 acquires feature vectors representing future time-series signals from the acquired block signals (step S304). Next, the feature vector conversion unit 340 converts the acquired feature vectors into a prediction signal related to the patient P's biological information (step S306). Next, the state estimation unit 350 estimates the patient P's state based on the acquired prediction signal (step S308). Next, the treatment device control unit 360 controls the treatment device 10 based on the estimated patient P's state. Thus, the processing of this flowchart ends.

[0065] According to the third embodiment described above, based on the estimated state of the patient P, the treatment device 10 is controlled to perform or stop the acquisition of X-ray dynamic images of the patient P and / or the irradiation of the treatment beam B from the treatment device 10. This further reduces radiation associated with the acquisition of unwanted X-ray dynamic images and the irradiation of the treatment beam.

[0066] Several embodiments of the present invention have been described above, but these embodiments are provided as examples and are not intended to limit the scope of the invention. These new embodiments can be implemented in various other ways, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments or their variations are included in the scope or spirit of the invention, and are included within the scope of the invention as described in the claims and its equivalents.

[0067] The time series signal prediction device, radiotherapy device, time series signal prediction method, program, and storage medium of the present invention adopt the following configuration.

[0068] (1): A time series signal prediction apparatus according to one aspect of the present invention comprises: a time series signal acquisition unit that acquires a time series signal related to the patient's biological information; a block conversion unit that divides the time series signal into partial signals at predetermined intervals and converts them into block signals of time series shorter than the time series of the time series signal; a feature vector acquisition unit that acquires feature vectors from the block signals, the feature vectors representing a time series signal at a more future time than when the time series signal was acquired; and a feature vector conversion unit that converts the feature vectors into a prediction signal related to the patient's biological information.

[0069] (2): In the above (1) scheme, the time series signal prediction device further includes a state estimation unit that predicts the patient's state based on the prediction signal.

[0070] (3): In the above (2) scheme, the time series signal prediction device further includes a parameter setting unit, which adjusts the parameters used by the state estimation unit in the prediction according to the time series signal and the prediction signal.

[0071] (4): In the above schemes (1) to (3), the multiple block signals repeatedly include the signal at a certain moment.

[0072] (5): In the above schemes (1) to (4), the feature vector conversion unit calculates the reliability of the predicted signal based on the feature vector.

[0073] (6): In the above schemes (1) to (5), the time series signal is a signal that changes according to the patient’s breathing, and the state is an abnormal state of the patient’s breathing action.

[0074] (7): In the above schemes (1) to (5), the time series signal is a signal that changes according to the patient’s breathing, and the state is the timing of the patient’s breathing action changing from exhalation to inhalation.

[0075] (8): In the above schemes (1) to (5), the time series signal is a signal that changes according to the location of the patient's tumor, and the prediction signal is the predicted location of the patient's tumor.

[0076] (9): In one aspect of the present invention, a radiotherapy device controls the timing of capturing fluoroscopic images of the patient based on the state predicted by the time-series signal prediction device of the above (2) aspect.

[0077] (10): In one aspect of the present invention, a radiotherapy device controls the timing of the irradiation of the therapeutic beam onto the patient based on the state predicted by the time-series signal prediction device of the above (2) aspect.

[0078] (11): One aspect of the present invention is a time series signal prediction method in which a computer performs the following processing: obtaining a time series signal related to the patient's biological information, dividing the time series signal into partial signals at each predetermined interval, converting them into block signals of a time series shorter than the time series signal, obtaining a feature vector from the block signals, the feature vector representing a time series signal at a more future time than when the time series signal was obtained, and converting the feature vector into a prediction signal related to the patient's biological information.

[0079] (12): A procedure of one aspect of the present invention causes a computer to perform the following processing: acquiring a time-series signal related to the patient's biological information, dividing the time-series signal into partial signals at predetermined intervals, converting it into a block signal of a time series shorter than the time series signal, acquiring a feature vector from the block signal, the feature vector representing a time-series signal at a more future time than when the time-series signal was acquired, and converting the feature vector into a prediction signal related to the patient's biological information.

[0080] (13): A storage medium of one aspect of the present invention stores a program that enables a computer to perform the following processing: acquiring a time-series signal related to the patient's biological information, dividing the time-series signal into partial signals at predetermined intervals, converting it into a block signal of a time series shorter than the time series signal, acquiring a feature vector from the block signal, the feature vector representing a time-series signal at a more future time than when the time series signal was acquired, and converting the feature vector into a prediction signal related to the patient's biological information.

[0081] Explanation of reference numerals in the attached figures 1…treatment system, 10…treatment device, 12…examination bed, 14…CT imaging device, 16…treatment beam irradiation gate, 100, 200, 300…time series signal prediction device, 110, 210, 310…time series signal acquisition unit, 120, 220, 320…block conversion unit, 130, 230, 330…feature vector acquisition unit, 140, 240, 340…feature vector conversion unit, 250, 350…state estimation unit, 360…treatment device control unit.

Claims

1. A time series signal prediction device, wherein, Possessing: a time series signal acquisition section that acquires a time series signal relating to biological information of a patient; a block conversion section that divides the time series signal into partial signals every prescribed interval, and converts into a block signal of a time series shorter than that of the time series signal; a feature vector acquisition section that acquires a feature vector from the block signal, the feature vector indicating a time series signal of a time later than the time of acquisition of the time series signal; and a feature vector conversion section that converts the feature vector into a prediction signal relating to the biological information of the patient.

2. The time series signal prediction device according to claim 1, wherein a state estimation section that predicts a state of the patient from the prediction signal is further possessed.

3. The time series signal prediction device according to claim 2, wherein a parameter setting section that adjusts a parameter used in prediction by the state estimation section from the time series signal and the prediction signal is further possessed.

4. The time series signal prediction device according to claim 1, wherein a plurality of the block signals repeatedly include a signal of a certain time.

5. The time series signal prediction device according to claim 1, wherein the feature vector conversion section calculates a reliability of the prediction signal from the feature vector.

6. The time series signal prediction device according to claim 2 or 3, wherein the time series signal is a signal that varies according to respiration of the patient, and the state is an abnormal state of a respiratory action of the patient.

7. The time series signal prediction device according to claim 2 or 3, wherein the time series signal is a signal that varies according to respiration of the patient, and the state is a timing at which the respiratory action of the patient changes from expiration to inspiration.

8. The time series signal prediction device according to any one of claims 1 to 5, wherein the time series signal is a signal that varies according to a tumor position of the patient, and the prediction signal is a predicted position of the tumor of the patient.

9. A radiation therapy device, wherein a timing of photographing of a fluoroscopic image of the patient is controlled based on the state predicted by the time series signal prediction device according to claim 2.

10. A radiation therapy device, wherein a timing of irradiation of a therapy beam to the patient is controlled based on the state predicted by the time series signal prediction device according to claim 2.

11. A time series signal prediction method, wherein a computer performs the following processing: acquires a time series signal relating to biological information of a patient, divides the time series signal into partial signals every prescribed interval, and converts into a block signal of a time series shorter than that of the time series signal, acquires a feature vector from the block signal, the feature vector indicating a time series signal of a time later than the time of acquisition of the time series signal, converts the feature vector into a prediction signal relating to the biological information of the patient.

12. A program, wherein causes a computer to perform the following processing: acquires a time series signal relating to biological information of a patient, dividing the time-series signal into partial signals for each prescribed interval, converting into a block signal of a time series shorter than that of the time-series signal, acquiring a feature vector from the block signal, the feature vector indicating a time-series signal for a time later than the time of acquisition of the time-series signal, converting the feature vector into a prediction signal relating to the patient's biological information.

13. A storage medium in which a program is stored, the program causing a computer to perform the following processing: acquiring a time-series signal relating to a patient's biological information, dividing the time-series signal into partial signals for each prescribed interval, converting into a block signal of a time series shorter than that of the time-series signal, acquiring a feature vector from the block signal, the feature vector indicating a time-series signal for a time later than the time of acquisition of the time-series signal, converting the feature vector into a prediction signal relating to the patient's biological information.

Images