WO2020211076A1 - 一种实现超声穿颅聚焦的方法以及电子设备 - Google Patents

一种实现超声穿颅聚焦的方法以及电子设备 Download PDF

Info

Publication number
WO2020211076A1
WO2020211076A1 PCT/CN2019/083433 CN2019083433W WO2020211076A1 WO 2020211076 A1 WO2020211076 A1 WO 2020211076A1 CN 2019083433 W CN2019083433 W CN 2019083433W WO 2020211076 A1 WO2020211076 A1 WO 2020211076A1
Authority
WO
WIPO (PCT)
Prior art keywords
head
image data
target model
training
dimensional
Prior art date
Application number
PCT/CN2019/083433
Other languages
English (en)
French (fr)
Inventor
王丛知
肖杨
马腾
胡战利
贾富仓
郑海荣
Original Assignee
深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳先进技术研究院 filed Critical 深圳先进技术研究院
Priority to PCT/CN2019/083433 priority Critical patent/WO2020211076A1/zh
Publication of WO2020211076A1 publication Critical patent/WO2020211076A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
    • A61N7/00—Ultrasound therapy

Definitions

  • This application relates to the field of medical technology, in particular to a method and electronic equipment for realizing ultrasound transcranial focusing.
  • HIFU high-intensity focused ultrasound
  • One of the key problems that need to be solved in the use of transcranial focused ultrasound for non-invasive deep brain neuromodulation and intracranial ablation is how to overcome the influence of the skull on ultrasound.
  • the density and speed of sound of the skull are approximately twice that of other human soft tissues, and the sound attenuation coefficient is at least an order of magnitude higher.
  • the skull has a non-uniform and complex structure with multiple layers, fluid-filled and porous, causing ultrasound to pass through the skull. Significant phase distortion and energy attenuation occur, and the ultrasound focal area appears shape distortion and position shift, so that precise and effective nerve stimulation and ablation treatment cannot be performed.
  • the conventional solution to this problem is to first perform a three-dimensional CT scan of the patient’s head, and estimate the density of the skull, sound velocity and other related acoustic parameters from the CT image; then, use a special computer program to calculate the emission of each element of the transducer The phase distortion and other waveform changes in the process of the ultrasonic wave passing through different parts of the skull; finally, according to the calculated waveform changes, the parameters such as the delay of each element of the transducer are corrected to realize the transcranial ultrasonic wave. Focus.
  • this method requires the patient to undergo a three-dimensional CT scan.
  • the embodiments of the present invention provide a method and electronic equipment for realizing ultrasound transcranial focusing, which are used to avoid CT scanning of a subject to be treated in the process of ultrasound deep brain stimulation.
  • the first aspect of the embodiments of the present invention provides an ultrasound deep brain stimulation method, the method including:
  • head three-dimensional magnetic resonance image data is image data obtained by performing a three-dimensional magnetic resonance imaging scan of the head of the subject to be treated
  • the ultrasonic transducer array is controlled to emit ultrasonic waves according to the ultrasonic emission sequence, and the ultrasonic waves are used to achieve transcranial focus on the head of the subject to be treated, and perform ultrasonic deep brain stimulation or ultrasonic thermal ablation treatment.
  • the method further includes:
  • head CT image data used for training the target model is image data obtained by performing CT scanning on a plurality of heads as training samples
  • the head three-dimensional MRI data used for training the target model is obtained by performing three-dimensional magnetic resonance imaging scans on the heads of a plurality of the training samples Image data.
  • the method further includes:
  • the training the head three-dimensional MRI data used for training the target model and the head CT image data used for training the target model to obtain the target model includes:
  • the head three-dimensional magnetic resonance image data used for training the target model and the head CT image data used for training the target model are trained to obtain the target model.
  • the three-dimensional MRI data of the head used for training the target model and the CT image data of the head used for training the target model are trained by a machine learning method to obtain the target model include:
  • the generative confrontation network is trained to obtain the target model.
  • the three-dimensional MRI data of the head used for training the target model and the CT image data of the head used for training the target model are trained by a machine learning method to obtain the target model include:
  • the head three-dimensional nuclear magnetic resonance image data used for training the target model and the head CT image data used for training the target model are trained by a random forest algorithm to obtain the target model.
  • the embodiment of the present invention provides an electronic device, including:
  • An acquiring unit for acquiring head three-dimensional magnetic resonance image data where the head three-dimensional magnetic resonance image data is image data acquired by performing a three-dimensional magnetic resonance imaging scan of the head of the subject to be treated;
  • a establishing unit configured to establish a three-dimensional digital head model based on the three-dimensional MRI image data of the head and the synthesized head CT image data;
  • a generating unit configured to generate an ultrasound transmission sequence according to the three-dimensional digital model of the head and the three-dimensional digital model of the ultrasound transducer array;
  • the control unit is configured to control the ultrasound transducer array to emit ultrasound waves according to the ultrasound emission sequence, and the ultrasound waves are used to achieve transcranial focus on the head of the subject to be treated, and perform ultrasound deep brain stimulation or ultrasound thermal ablation therapy.
  • the acquiring unit is further configured to train the head CT image data of the target model, and the head CT image data used to train the target model is to perform CT scanning on a plurality of heads as training samples The acquired image data;
  • the acquiring unit is further configured to acquire head three-dimensional MRI data used for training the target model, and the head three-dimensional MRI data used for training the target model is performed on the heads of a plurality of training samples.
  • Image data obtained by 3D magnetic resonance imaging scan.
  • the electronic device further includes:
  • the training unit is configured to train the head three-dimensional magnetic resonance image data used for training the target model and the head CT image data used for training the target model to obtain the target model.
  • the training unit is specifically configured to train the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model through a machine learning method To obtain the target model.
  • the training unit is specifically configured to train the generative confrontation network through the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model To obtain the target model.
  • the training unit is specifically configured to train the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model through a random forest algorithm to obtain The target model.
  • the embodiments of the present invention provide a method and electronic equipment for realizing ultrasound transcranial focusing. Without CT imaging of the subject to be treated, only the head 3D MRI image data is input to the target model, and the electronic equipment can obtain To the synthesized head CT image data output by the target model, the electronic device can generate the ultrasonic emission sequence according to the three-dimensional digital model of the head and the three-dimensional digital model of the ultrasonic transducer array, and the electronic device controls the ultrasonic transducer The array emits ultrasound according to the ultrasound emission sequence, through which ultrasound deep brain stimulation can be performed on the head of the subject to be treated.
  • the method shown in this embodiment can eliminate the risk of cancer caused by radiation on the subject to be treated during the CT imaging process, and improve the safety of treatment, while simplifying the treatment steps while ensuring that the positioning accuracy of the ultrasonic transcranial focus focus meets the requirements. , Greatly shorten the treatment time and reduce the treatment cost.
  • Fig. 1 is a schematic diagram of ultrasound deep brain stimulation of the head by a phased array transducer provided in the prior art
  • Figure 2 is a schematic diagram of the method of implanting a hydrophone provided by the prior art
  • FIG. 3 is a flowchart of an embodiment of the ultrasonic deep brain stimulation method provided by the present invention.
  • FIG. 4 is a flowchart of another embodiment of the ultrasound deep brain stimulation method provided by the present invention.
  • FIG. 5 is a schematic diagram of simulation experiment results of an embodiment of realizing transcranial focusing provided by the present invention.
  • FIG. 6 is a schematic diagram of simulation experiment results of another embodiment for realizing transcranial focusing provided by the present invention.
  • FIG. 7 is a schematic structural diagram of an embodiment of an electronic device provided by the present invention.
  • skull 101 has a non-uniform complex structure with multiple layers, liquid-filled and porous, resulting in a transducer
  • the ultrasonic waves emitted by each array element 102 undergo significant phase distortion and energy attenuation after passing through the skull 101, and the ultrasonic focal region appears shape distortion and position shift, so that precise and effective nerve stimulation cannot be performed.
  • the skull 101 may also cause secondary effects such as standing waves, especially when using low-frequency and long-pulse ultrasound, energy accumulation may occur at the "skull-tissue" and "air-tissue” interfaces.
  • low-frequency ultrasound around 250KHz (the wavelength of which is equivalent to the thickness of the skull 101) can reduce the phase distortion to a certain extent.
  • the focal range of low-frequency ultrasound is larger and the cavitation threshold is lower, which increases unnecessary risks. Therefore, clinically, ultrasound with a frequency of 600-1000 MHz is generally used, and at these frequencies, the phase distortion caused by the skull 101 is very obvious.
  • phased array transducer that includes multiple independently driven array elements to correct the emission phase and amplitude of each array element through computer control to achieve compensation for focal distortion and energy attenuation.
  • the key is the method of measuring or estimating the above-mentioned correction value.
  • the time reversal method can perform the above-mentioned phase and amplitude corrections at the same time.
  • the time reversal method first uses an ultrasonic transducer to receive the ultrasonic wave emitted by a strong reflector, and flips the received sound pressure waveform back and forth on a time axis. Then use the reversed signal to excite the transducer to emit ultrasonic waves. Because the propagation of ultrasonic waves is reversible in the time domain, its propagation path will be consistent with the reception, so it will refocus on the position of the strong reflector.
  • This method was originally used for shock wave lithotripsy because the stones in the human body are natural strong reflectors. However, such a natural reflector does not exist in the human brain. Therefore, when this method is applied to transcranial ultrasound focusing, three different methods for realizing time reversal have been gradually developed.
  • the first method of time reversal is the implanted hydrophone method
  • the method of implanting a hydrophone is to place the hydrophone 201 at a desired focus position, and then individually excite each element in the transducer array 202 in turn.
  • the hydrophone 201 can measure the phase shift caused by the presence of the skull, and compensate for these phase shifts on the excitation signal, so that the ultrasound can be focused at the desired focus position.
  • the sound pressure measured by the hydrophone 201 is processed by the amplification 203, display and phase estimation software 204, and finally forms a phase correction sequence, which drives the ultrasound array transducer 202 to achieve transcranial focusing.
  • this method is invasive, and the hydrophone 201 needs to be implanted in the brain during clinical application. Second, if a new focal position needs to be generated, the hydrophone needs to be moved and the entire implantation process repeated, which will greatly increase the processing time and the risk of complications.
  • the second time reversal method is the cavitation microbubble method
  • Pernot et al. proposed a method using two different ultrasound array transducers.
  • a high-power ultrasonic transducer is used to perform a high-intensity instantaneous pulse emission to form a cavitation microbubble in the brain's desired focus area.
  • the ultrasonic signal generated by the fragmentation of the microbubble is received by another ultrasonic transducer array and the subsequent time-reversal transmission and focusing are completed. Since only a small cavitation microbubble needs to be generated, this method theoretically does not cause damage to the brain.
  • Aubry et al. proposed to obtain various acoustic parameters of the skull based on CT image data, and then simulate the sound field distribution of sound waves after passing through the skull by the finite time domain difference method (FDTD) to obtain sufficient intensity at the expected position
  • FDTD finite time domain difference method
  • the initial emission sequence of the sound field was used to form cavitation microbubbles in the focal area, and the experimentally measured intensity of the final focal sound pressure reached 97% of that of the implanted hydrophone method.
  • a small liquid droplet that is easy to vaporize is injected into the expected focus area. It instantly vaporizes to form microbubbles, and then completes the time reversal and cranial focusing according to the aforementioned method.
  • the third method of time inversion is virtual sound source simulation
  • Marquet et al. proposed that by realizing an accurate simulation of the ultrasonic propagation process on a computer, a "virtual" time reversal was completed, so as to obtain the ultrasonic emission sequence of each individual element of the transducer required for transcranial focusing.
  • a computerized tomography (CT) scan is performed on multiple isolated skull samples to obtain CT images, and then the hydrophone method is used to measure the phase distortion caused by each sample, and a statistical model is derived based on this to establish skull density, sound velocity, etc. Correspondence between the parameters and the Hounsfield Unit (HU) of the CT image. Then, perform an in-vivo CT scan of the patient's head, and use the above statistical model to obtain its density, sound velocity, etc. from the resulting image, as the input parameters of the FDTD simulation program to solve the linear wave equation.
  • CT computerized tomography
  • the hydrophone implantation method and the cavitation microbubble method have great limitations in clinical application due to their invasiveness and potential safety risks, and they are not suitable for ultrasound deep brain stimulation and intracranial HIFU ablation. treatment.
  • the virtual sound source simulation method facilitates careful planning and repeated optimization before treatment, helps to improve the treatment effect and safety, and is by far the most suitable method for clinical use.
  • this method requires the patient to perform a three-dimensional CT scan, which will not only increase the radiation dose received by the patient, but also increase the risk of cancer (according to statistics, 0.4% of cancer patients in the United States are caused by CT exposure).
  • CT images and MRI images must be accurately registered to ensure the accuracy of the treatment position. The additional steps will increase the cost of treatment, time-consuming, and additional risks caused by registration errors.
  • the ultrasound deep brain stimulation method provided by the present application can perform ultrasound deep brain stimulation or intracranial HIFU ablation treatment without the need for CT scanning of the head of the subject to be treated.
  • Step 301 The CT scanning device performs CT scanning on a plurality of heads as training samples to obtain head CT image data for training the target model.
  • multiple training samples can be collected as needed.
  • the training samples can be human or animal waiting training samples.
  • the CT scanning device performs three-dimensional CT scanning on the heads of all training samples to obtain training targets for each training sample. CT image data of the model's head.
  • Step 302 The electronic device obtains the head CT image data used for training the target model.
  • the CT scanning device may send the head CT image data used for training the target model to the electronic device.
  • This embodiment does not limit the specific process for the electronic device to obtain the CT image data of the head used for training the target model.
  • the CT scanning device may directly use the CT scan of the head for training the target model.
  • the image data is sent to the electronic device, and the head CT image data used for training the target model can also be sent to the electronic device through other devices.
  • Step 303 The nuclear magnetic resonance equipment performs a three-dimensional magnetic resonance imaging scan on the head of the sample to be trained to obtain three-dimensional nuclear magnetic resonance image data of the head for training the target model.
  • Step 304 The electronic device obtains the head three-dimensional nuclear magnetic resonance image data used for training the target model.
  • the nuclear magnetic resonance equipment may send the head three-dimensional nuclear magnetic resonance image data for training the target model to all The electronic device, this embodiment does not limit the specific process for the electronic device to obtain the head three-dimensional nuclear magnetic resonance image data used for training the target model.
  • the nuclear magnetic resonance device may directly use the The head three-dimensional nuclear magnetic resonance image data of the target model is sent to the electronic device, and the head three-dimensional nuclear magnetic resonance image data used for training the target model can also be sent to the electronic device through other devices.
  • Step 305 The electronic device trains the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model to obtain the target model.
  • the target model shown in this embodiment is used to predict the corresponding synthesized head CT image data based on the head three-dimensional magnetic resonance image data.
  • the electronic device shown in this embodiment can train the head three-dimensional MRI data used to train the target model and the head CT image data used to train the target model through a machine learning method To obtain the target model.
  • the machine learning method shown in this embodiment may be a generative confrontation network
  • Generative Adversarial Networks shown in this embodiment includes two parts, a Fully Convolutional Network (FCN, Fully Convolutional Networks) and a Convolutional Neural Network (CNN, Convolutional Neural Networks), The two parts are trained simultaneously.
  • FCN Fully Convolutional Network
  • CNN Convolutional Neural Network
  • the electronic device can train the generative confrontation network through the head three-dimensional MRI data used to train the target model and the head CT image data used to train the target model. Obtain the target model.
  • FCN is trained to use the head CT image data used to train the target model to generate head CT image data that is closer to the real CT image and used to train the target model
  • CNN is trained to distinguish the real ones with little difference.
  • the CT image and the head CT image data used to train the target model, the two networks thus form an adversarial relationship, and the finally trained GAN can also achieve high performance, and the head used to train the target model can be used
  • the three-dimensional MRI image data calculates the head CT image data that is very close to the real CT image and is used to train the target model.
  • the machine learning method shown in this embodiment may also be a random forest algorithm.
  • the random forest algorithm will be described below:
  • Random Forest is an integrated machine learning algorithm. The core idea is to integrate several independent decision trees together. Each decision tree generates classification results independently, and finally obtains the final classification results through voting.
  • the "random" of random forest has two meanings, one is to randomly select data, and the other is to randomly select features or variables.
  • the data on which each decision tree is based uses random resampling (bootstrap) to randomly sample several samples from the original data with replacement to form different self-service sample sets. Then, a random feature selection method is used to generate a decision tree on the self-service sample set.
  • the generated trees are formed into a random forest, and the new data is classified by the random forest.
  • the classification result is determined by the number of votes of the tree classifier.
  • Random forest algorithm has the ability to analyze complex interaction classification features, and has a faster learning speed. Its variable importance measurement method can be used as a feature selection tool for high-dimensional data. Random forest variable importance measurement method: Feature selection refers to the selection of a feature subset or a variable subset that optimizes a certain evaluation criterion from the original feature set. For feature selection, the first is to evaluate the importance of variables. Random forests have two ways to evaluate the importance of variables, one is Mean Decrease Impurity (MDI), and the other is based on out-of-bag data (OOB) The average classification accuracy is reduced (Mean Decrease accuracy, MDA).
  • MDI Mean Decrease Impurity
  • OOB out-of-bag data
  • the electronic device shown in this embodiment can train the head three-dimensional MRI data used to train the target model and the head CT image data used to train the target model through the random forest algorithm to obtain The target model.
  • the electronic device can train the head three-dimensional MRI data used to train the target model and the head CT image data used to train the target model to obtain the target model Based on the target model, the corresponding synthetic head CT image data can be predicted based on the head three-dimensional magnetic resonance image data, so that when the subject’s head is subjected to ultrasound deep brain stimulation, only the head of the subject is required MRI imaging is sufficient, without CT imaging, which can eliminate the risk of cancer caused by radiation and improve the safety of treatment. At the same time, it can simplify the treatment steps, greatly shorten the treatment time, and reduce the treatment cost.
  • Step 401 The nuclear magnetic resonance equipment performs a three-dimensional magnetic resonance imaging scan of the head of the subject to be treated to obtain three-dimensional nuclear magnetic resonance image data of the head.
  • the MRI equipment can perform a three-dimensional magnetic resonance imaging scan of the head of the subject to obtain the head Three-dimensional MRI image data.
  • Step 402 The electronic device acquires the three-dimensional nuclear magnetic resonance image data of the head.
  • the nuclear magnetic resonance equipment may send the head three-dimensional nuclear magnetic resonance image data to the electronic device.
  • the example does not limit the specific process of the electronic device acquiring the three-dimensional nuclear magnetic resonance image data of the head.
  • the nuclear magnetic resonance device may directly send the three-dimensional nuclear magnetic resonance image data of the head to the electronic device.
  • the three-dimensional nuclear magnetic resonance image data of the head can be sent to the electronic device through other devices.
  • Step 403 The electronic device inputs the head three-dimensional nuclear magnetic resonance image data into the target model obtained through pre-training.
  • the head three-dimensional nuclear magnetic resonance image data is input into the pre-trained target model.
  • the target model please refer to the embodiment shown in FIG. 3, which is specifically in this embodiment It is not limited, as long as the target model can predict the corresponding synthetic head computer tomography CT image data based on the head three-dimensional magnetic resonance image data.
  • Step 404 The electronic device obtains the composite head CT image data output by the target model.
  • Step 405 The electronic device establishes a three-dimensional digital model of the head according to the three-dimensional nuclear magnetic resonance image data of the head and the synthesized head CT image data.
  • the electronic device performs three-dimensional reconstruction and registration on the head three-dimensional magnetic resonance image data and the synthesized head CT image data to establish the sphincter bone and brain tissue of the head of the subject to be treated Three-dimensional digital model of head with structure, density and acoustic parameters.
  • Step 406 The electronic device generates an ultrasound transmission sequence according to the three-dimensional digital model of the head and the three-dimensional digital model of the ultrasound transducer array.
  • the electronic device can select one or more positions of brain nerve nuclei to be stimulated, and compare the three-dimensional digital model of the head to the Position the brain nuclei to be stimulated.
  • the electronic device can also obtain the structure and density of the skull and brain tissue according to the three-dimensional digital model of the head, and then calculate the acoustic parameters of the head according to the structure and density of the skull and brain tissue.
  • the acoustic parameters include but are not limited to the speed of sound, Attenuation coefficient.
  • the electronic device shown in this embodiment can adjust the virtual spatial position of the three-dimensional digital model of the head and the three-dimensional digital model of the ultrasonic transducer array according to the actual spatial position, and adjust the three-dimensional digital model of the head and the ultrasonic transducer array
  • the virtual space position of the three-dimensional digital model is adjusted in place according to the desired actual space position, and the virtual sound source is placed at one or more positions that need to be focused.
  • the electronic device can simulate the ultrasonic waves emitted by the virtual sound source on the head The state of propagation in a three-dimensional digital model. When the ultrasonic wave propagates to the virtual spatial position where the ultrasonic transducer array is located, the electronic device simulates the voltage signal of the ultrasonic transducer array.
  • the sound intensity signal of the ultrasonic transducer array on the surface of the ultrasonic transducer array is simulated Harmonize the sound pressure signal, simulate the voltage signal of the ultrasonic transducer array based on the sound intensity signal and sound pressure signal and the piezoelectric conversion parameters of the ultrasonic transducer array, that is, simulate the sound intensity signal and sound pressure signal to simulate the ultrasonic transducer passing through The voltage signal after piezoelectric conversion.
  • the electronic device performs time inversion on the voltage signal to generate a time inversion signal as the ultrasonic emission sequence.
  • the time reversal of the voltage signal is a reversal in the sequence of time, and this method can be called a time reversal method.
  • the voltage signal is reversed sequentially in time, it is used to excite the ultrasonic transducer array, and the generated ultrasonic waves will be focused on one or more positions where the virtual sound source has been placed. Therefore, through the ultrasonic time reversal, the present invention can obtain what is needed to stimulate the designated brain nerve nuclei, and can complete the transcranial focused ultrasound transmission sequence.
  • Step 407 The electronic device controls the ultrasound transducer array to emit ultrasound according to the ultrasound emission sequence.
  • the head of the subject to be treated that needs ultrasound deep brain stimulation is fixed at a designated position in the magnetic resonance imaging system using a head fixing and positioning device, and the ultrasound transducer array is also designed in advance The position is fixed.
  • the electronic device issues instructions to control the ultrasound transducer array to perform ultrasound emission according to the ultrasound emission sequence, which is used to achieve transcranial focus on the head of the subject to be treated, and perform ultrasound deep brain stimulation or ultrasound thermal ablation therapy.
  • the above-mentioned methods for performing ultrasound brain stimulation on the brain nerve nucleus to be stimulated include: pulse mode, multi-period mode and coding mode, etc.
  • the present invention is not limited to this.
  • the method shown in this embodiment can be further simplified. For example, omitting the accurate simulation of the whole process of ultrasonic propagation, directly connecting the focal position and the position of each element of the ultrasonic transducer array, assuming that the ultrasonic wave propagates in a straight line between the two, the connection is the ultrasonic propagation path. Ultrasound travels faster on the part of the path that passes through the skull.
  • the influence of this part of the path on the ultrasonic propagation time can be calculated based on the above-mentioned estimated skull sound velocity distribution from the synthetic head CT image data, and then calculated How to correct the transmission delay of each array element of the ultrasonic transducer, so that the ultrasonic waves emitted by each array element can reach the focus position at the same time, and achieve transcranial focusing.
  • the electronic device can obtain the synthetic head output by the target model According to the CT image data, the electronic device can generate the ultrasound transmission sequence according to the three-dimensional digital model of the head and the three-dimensional digital model of the ultrasound transducer array, and the electronic device controls the ultrasound transducer array to emit ultrasound according to the ultrasound transmission sequence.
  • the ultrasound can be used for deep brain stimulation of the head of the subject to be treated.
  • the method shown in this embodiment can eliminate the risk of cancer caused by radiation on the subject to be treated during the CT imaging process, and improve the safety of treatment, while simplifying the treatment steps while ensuring that the positioning accuracy of the ultrasonic transcranial focus focus meets the requirements. , Greatly shorten the treatment time and reduce the treatment cost.
  • MAX represents the maximum value of image gray.
  • MSE is the mean square error.
  • PSNR peak signal-to-noise ratio
  • image quality that is, the ratio of the maximum possible signal power to the destructive noise power that affects its accuracy. Since many signals have a very wide dynamic range, the peak signal-to-noise ratio is usually expressed in logarithmic decibel units. In image processing, it is often necessary to calculate PSNR to objectively evaluate images.
  • PSNR is an objective standard to measure image distortion or noise level. The greater the PSNR value between the two images, the more similar they are. The general standard is 30dB, and the image degradation below 30dB is more obvious.
  • the comparison between the synthetic head CT image data estimated by the target training method shown in this embodiment and the real CT image for the same object to be treated shows that the PSNR value must reach 27.6dB, which is very close to 30dB. It is believed that the synthetic head CT image data obtained by this method can be used to estimate the acoustic parameters of the skull and realize the ultrasound transcranial focusing.
  • the invention uses the ultrasonic time inversion software in the two-dimensional plane to obtain the simulation experiment of the method for realizing the ultrasonic emission sequence required for transcranial focusing.
  • the experimental results are shown in Figure 5 and Figure 6, the small dot at (0,40) in Figure 5 is the initial position of the virtual sound source, and Figure 6 is the ultrasonic focusing effect achieved by simulation using the time inversion method;
  • the upper frame of the picture is a 1024-element linear ultrasonic transducer array.
  • the gray part in Figure 5 and Figure 6 is the skull model reconstructed from CT scan images, and the acoustic parameters of the head are calculated according to the structure and density of the skull and brain tissue, and then imported into the simulation software run by the electronic device to set it as two-dimensional
  • the plane corresponds to the corresponding value of the calculation node. From the simulation results, although the lateral and longitudinal dimensions of the ultrasound focus point have increased compared with the original sound source size after the time-reversal launch, most of the energy is still concentrated in the predicted position, which can satisfy precise point stimulation and guarantee other surrounding areas A dual requirement that is less affected. As long as the method is extended from a two-dimensional plane to a three-dimensional digital model, the needs of the present invention can be met.
  • the mouse after anesthesia, the top of the head is shaved, and it is fixed on the brain stereotaxic device.
  • the ultrasound transducer array is accurately positioned by the brain stereotaxic device, and is close to the skull to radiate pulsed ultrasound to the motor cortex. Ultrasound stimulation can synchronously induce the mice to produce physical response.
  • the electronic device shown in this embodiment includes:
  • the acquiring unit 701 is configured to acquire three-dimensional magnetic resonance image data of the head, where the three-dimensional magnetic resonance image data of the head is image data acquired by performing a three-dimensional magnetic resonance imaging scan of the head of the subject to be treated;
  • the acquiring unit 701 is further configured to train CT image data of the head of the target model, and the CT image data of the head used to train the target model is to perform CT on a plurality of heads as training samples.
  • the acquiring unit 701 is further configured to acquire head three-dimensional magnetic resonance image data used for training the target model, and the head three-dimensional nuclear magnetic resonance image data used for training the target model is the head of a plurality of training samples.
  • the training unit 702 is configured to train the head three-dimensional nuclear magnetic resonance image data used for training the target model and the head CT image data used for training the target model to obtain the target model.
  • the training unit 702 is specifically configured to train the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model through a machine learning method to obtain The target model.
  • the training unit 702 is specifically configured to train the generative confrontation network through the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model To obtain the target model.
  • the training unit 702 is specifically configured to train the head three-dimensional magnetic resonance image data used to train the target model and the head CT image data used to train the target model through the random forest algorithm to obtain The target model.
  • the input unit 703 is used to input the head three-dimensional nuclear magnetic resonance image data into a target model obtained through pre-training, and the target model is used to obtain a corresponding synthesis according to the head three-dimensional nuclear magnetic resonance image data CT image data of head computer tomography;
  • the establishment unit 704 is configured to establish a three-dimensional digital model of the head according to the three-dimensional nuclear magnetic resonance image data of the head and the synthesized head CT image data;
  • a generating unit 705, configured to generate an ultrasound transmission sequence according to the three-dimensional digital model of the head and the three-dimensional digital model of the ultrasound transducer array;
  • the control unit 706 is configured to control the ultrasound transducer array to emit ultrasound waves according to the ultrasound emission sequence, and the ultrasound waves are used to achieve transcranial focus on the head of the subject to be treated, and perform ultrasound deep brain stimulation or ultrasound thermal ablation treatment.
  • the disclosed system, device, and method may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components can be combined or It can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.
  • the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the technical solution of the present invention essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , Including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in each embodiment of the present invention.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code .

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

一种实现超声穿颅聚焦的方法以及电子设备，无需对待治疗对象进行头部CT成像的情况下，只需将其头部核磁共振图像数据输入至由训练获得的目标模型，即可获取到所述目标模型输出的合成头部CT图像数据，电子设备即可根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列，电子设备控制所述超声换能器阵列按照所述超声发射序列发射超声波，可对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融等治疗。消除带治疗对象在CT成像过程中受到辐射引发癌症的风险，提高治疗的安全性，同时可以简化治疗步骤，大幅缩短治疗时间，降低治疗成本。

Description

一种实现超声穿颅聚焦的方法以及电子设备 技术领域
本申请涉及医疗技术领域，尤其涉及的是一种实现超声穿颅聚焦的方法以及电子设备。
背景技术
随着社会老龄化和日渐加深的心理压力等因素的影响，世界范围内包括帕金森病、肌张力失常、强迫症、抑郁症、癫痫等神经精神疾病患者的数量急剧增加，目前全球患者已逾五亿六千万。德国的科学家报道了电刺激下犬的大脑皮层可引发特定的躯体反应。这一重大发现，在此后的一个世纪催生了大脑电刺激、磁刺激、神经植入等系列干预技术，极大促进了人们对脑皮层功能定位的认识和脑疾病研究仪器的研发，并开启了情感、记忆、认知等脑功能调控和心理、精神疾病干预治疗的新篇章。利用超声实现非侵入式的深脑神经刺激，不仅安全有效，还可以实现定点特异性神经网络调控、多点网络神经调控等其他方法难以实现的功能,有助于开发中枢神经疾病的潜在疗法，也为探索正常人脑功能，理解认知、决策与思维、精确掌握神经环路活动带来了强有力的新工具。此外，利用高强度聚焦超声(HIFU)对颅内的脑肿瘤或其他病灶组织进行消融治疗，可以有效治疗颅内恶性实体肿瘤、神经性震颤等疾病，也是极具发展潜力的前沿治疗技术。
利用经颅聚焦超声进行非侵入式深脑神经调节以及颅内消融治疗需要解决的关键问题之一，就是如何克服颅骨对超声的影响。颅骨的密度和声速都大约是其他人体软组织的两倍，声衰减系数则至少高出一个数量级，再加上颅骨具有多层、充液和多孔的非均匀性复杂结构，造成超声穿过颅骨后发生显著的相位畸变和能量衰减，超声焦域出现形状扭曲和位置偏移，以至无法进行精确有效的神经刺激和消融治疗。这一问题的常规解决方法是，首先，对患者进行头部三维CT扫描，通过CT图像估算颅骨的密度、声速等相关声学参数；然后，利用专门的计算机程序计算换能器各个阵元所发射的超声波穿过颅骨不同部位的过程中出现的相位畸变等波形变化情况；最后，根据所计算的波形变化情况对换能器各个阵元发射超声波的延时等参数进行修正，实现超声波的经颅聚焦。但这种方法需要患者进行三维CT扫描，一方面会增加患者所受到的辐射剂量，增加患者罹患癌症的风险(据统计，美国癌症患者中有0.4％的比例是由于照射CT引起的)；另一方面，无论是进行深脑神经刺激还是颅内HIFU治疗，都是在核磁共振(MRI)的引导下进行的，因此必须将CT图像和MRI图像进行精确配准才能保证治疗位置的准确性，这些额外步骤会增加治疗的成本、耗时和配准误差引起的附加风险。
发明内容
本发明实施例提供了一种实现超声穿颅聚焦的方法以及电子设备，其用于在超声深部脑刺激的过程中，避免对待治疗对象进行CT扫描。
本发明实施例第一方面提供了一种超声深部脑刺激方法，所述方法包括：
获取头部三维核磁共振图像数据，所述头部三维核磁共振图像数据为对待治疗对象的头部进行三维磁共振成像扫描所获取到的图像数据；
将所述头部三维核磁共振图像数据作为输入，输入至通过预先训练获得的目标模型中，所述目标模型用于根据所述头部三维核磁共振图像数据获得对应的合成头部电子计算机断层扫描CT图像数据；
根据所述头部三维核磁共振图像数据以及所述合成头部CT图像数据建立头部三维数字模型；
根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列；
控制所述超声换能器阵列按照所述超声发射序列发射超声波，所述超声波用于对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗。
可选的，所述方法还包括：
获取用于训练目标模型的头部CT图像数据，所述用于训练目标模型的头部CT图像数据为对多个作为训练样本的头部进行CT扫描所获取到的图像数据；
获取用于训练目标模型的头部三维核磁共振图像数据，所述用于训练目标模型的头部三维核磁共振图像数据为对多个所述训练样本的头部进行三维磁共振成像扫描所获取到的图像数据。
可选的，所述方法还包括：
对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
可选的，所述对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型包括：
通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
可选的，所述通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型包括：
通过所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据，对生成式对抗网络进行训练以获取所述目标模型。
可选的，所述通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型包括：
通过随机森林算法对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
本发明实施例提供了一种电子设备，包括：
获取单元，用于获取头部三维核磁共振图像数据，所述头部三维核磁共振图像数据为对待治疗对象的头部进行三维磁共振成像扫描所获取到的图像数据；
输入单元，用于将所述头部三维核磁共振图像数据作为输入，输入至通过预先训练获得的目标模型中，所述目标模型用于根据所述头部三维核磁共振图像数据获得对应的合成头部电子计算机断层扫描CT图像数据；
建立单元，用于根据所述头部三维核磁共振图像数据以及所述合成头部CT图像数据建立头部三维数字模型；
生成单元，用于根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列；
控制单元，用于控制所述超声换能器阵列按照所述超声发射序列发射超声波，所述超声波用于对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗。
可选的，所述获取单元还用于，用于训练目标模型的头部CT图像数据，所述用于训练目标模型的头部CT图像数据为对多个作为训练样本的头部进行CT扫描所获取到的图像数据；
所述获取单元还用于，获取用于训练目标模型的头部三维核磁共振图像数据，所述用于训练目标模型的头部三维核磁共振图像数据为对多个所述训练样本的头部进行三维磁共振成像扫描所获取到的图像数据。
可选的，所述电子设备还包括：
训练单元，用于对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
可选的，所述训练单元具体用于，通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
可选的，所述训练单元具体用于，通过所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据，对生成式对抗网络进行训练以获取所述目标模型。
可选的，所述训练单元具体用于，通过随机森林算法对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
本发明实施例提供了一种实现超声穿颅聚焦的方法以及电子设备，无需对待治疗对象进行CT拍摄的情况下，只需将头部三维核磁共振图像数据输入至目标模型，电子设备即可获取到所述目标模型输出的合成头部CT图像数据，电子设备即可根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列，电子设备控制所述超声换能器阵列按照所述超声发射序列发射超声波，通过该超声波可对待治疗对象的头部进行超声深部脑刺激。本实施例所示的方法可在保证超声穿颅聚焦焦点定位精度满足需求的前提下，消除带治疗对象在CT成像过程中受到辐射引发癌症的风险，提高治疗的安全性，同时可以简化治疗步骤，大幅缩短治疗时间，降低治疗成本。
附图说明
图1为现有技术所提供的相控阵换能器对头部进行超声深部脑刺激的示意图；
图2为现有技术所提供的植入水听器法的示意图；
图3为本发明所提供的超声深部脑刺激方法的一种实施例步骤流程图；
图4为本发明所提供的超声深部脑刺激方法的另一种实施例步骤流程图；
图5为本发明所提供的实现穿颅聚焦的一种实施例仿真实验结果示意图；
图6为本发明所提供的实现穿颅聚焦的另一种实施例仿真实验结果示意图；
图7为本发明所提供的电子设备的一种实施例结构示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本申请中出现的术语“和/或”，可以是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本申请中字符“/”，一般表示前后关联对象是一种“或”的关系。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的实施例能够以除了在这里图示或描述的内容以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或模块的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或模块，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或模块。
如图1所示，利用经颅聚焦超声进行非侵入式深脑神经调节以及颅内消融治疗需要解决的关键问题之一，就是如何克服颅骨101对超声的影响。颅骨101的密度和声速都大约是其他人体软组织的两倍，声衰减系数则至少高出一个数量级，再加上颅骨101具有多层、充液和多孔的非均匀性复杂结构，造成换能器102各个阵元所发射的超声波穿过颅骨101后发生显著的相位畸变和能量衰减，超声焦域出现形状扭曲和位置偏移，以至无法进行精确有效的神经刺激。此外，颅骨101还有可能引起驻波等次生效应，特别是当使用低频和长脉冲超声时，可能在“头骨-组织”及“空气-组织”界面形成能量累积现象。虽然使用250KHz左右的低频超声(其波长与颅骨101厚度相当)可以在一定程度上减少相位畸变。但低频超声的焦域更大，空化阈值更低，增加了不必要的风险。因此，临床上一般使用频率为600-1000MHz的超声，而在这些频率上，颅骨101引起的相位畸变非常明显。要解决这一问题，必须利用包括多个可独立驱动阵元的相控阵换能器，通过计算机控制校正各阵元的发射相位和幅度，实现对焦点失真和能量衰减的补偿。而其中的关键，就是测量或估算上述校正数值的方法。
时间反演法可以同时进行上述的相位和幅度校正，时间反演法首先使用超声换能器接收某个强反射子发出的超声波，并将所接收的声压波形在一段时间轴上前后翻转，再用翻转后的信号激励换能器发射超声波，因为超声波的传播在时域可逆，其传播路径会与接收时保持一致，因此会重新聚焦于强反射子的位置。该方法最初被用于冲击波碎石，因为人体内的结石是天然的强反射体。但是，人脑中并不存在这样的天然反射体。因此，在将这种方法应用于经颅超声聚焦时，逐渐发展出三种不同的实现时间反演的方法。
第一种时间反演的方法是植入水听器法
如图2所示，植入水听器法是将水听器201放置在希望聚焦的位置，然后对换能器阵列202中的每个阵元依次单独激励。这时可以用水听器201测量由于颅骨存在所引起的相位偏移，在激励信号上对这些相位偏移进行补偿，就可以实现超声在期望焦点位置的聚焦。具体的，水听器201测量的声压经过放大203、显示和相位估计软件204处理，最后形成相位修正序列，驱动超声阵列换能器202实现穿颅聚焦。虽然这种方法所得到的结果目前被认为是同类方法中的“金标准”，但它的应用也受到很大的限制。首先，该方法是有创的，临床应用时需要在大脑植入水听器201。第二，如果需要产生一个新的焦点位置，水听器就需要被移动并重复整个植入过程，这将极大地增加处理时间和出现并发症的风险。
第二种时间反演的方法是空化微泡法
为了解决时间反演法需要大脑中存在主动或被动声源的问题，Pernot等提出一种使用两个不同的超声阵列换能器的方法。首先，使用其中高功率的超声换能器进行一次高强度的瞬时脉冲发射，以在大脑期望聚焦的区域中形成一个空化微泡。微泡破碎所产生的超声信号被另一个超声换能器阵列接收并完成后续的时间反演发射和聚焦。由于只需要产生一个很小的空化微泡，这种方法理论上不会对大脑产生伤害。但是，由于颅骨的存在，使得第一次发射时很难得到足够的声压幅度以在预期位置产生空化效应。为了解决这个问题，Aubry等提出，基于CT图像数据获得颅骨的各种声学参数，再通过有限时域差分法(FDTD)仿真声波穿过颅骨后的声场分布情况，以获得在预期位置形成足够强度声场的初始发射序列。这个初始发射序列被用来在聚焦区域形成空化微泡，实验测得最终获得的焦点声压强度达到了植入水听器法的97％。为避免诱导空化所需要的高声压对大脑造成伤害，Haworth等对上述方法进行了改进，先将某种易于气化的微小液滴注射到预期聚焦区域，在利用高频高功率超声使其瞬间气化形成微泡，然后再按照前述方法完成时间反演和穿颅聚焦。
第三种时间反演的方法是虚拟声源仿真法
Marquet等提出，通过在计算机上实现超声波传播过程的精确仿真，完成一次“虚拟”的时间反演，从而获得穿颅聚焦所需的换能器各个独立阵元的超声发射序列。
首先，对多个离体颅骨样本进行计算机断层扫描(CT)扫描获得其CT图像，再利用水听器法测量各个样本所造成的相位畸变，据此推导统计模型，建立起颅骨密度、声速等参数与CT图像的Hounsfield单位(HU)之间的对应关系。然后，对患者头部进行在体CT扫描，并利用上述统计模型从所得图像中获得其密度、声速等，作为FDTD仿真程序求解线性波动方程的输入参数。在仿真程序中，放置一个虚拟声源在预期聚焦位置，仿真其所发射声波传播的整个过程，就可以获得换能器阵元表面空间位置上的声压波形，进而实现时间反演和穿颅聚焦。经过实验验证，由这种方法实现的焦点位置误差为0.7mm，聚焦能量可以达到植入水听器法的90％。
近期很多研究都对该方法进行了尝试和改进，Pinton等先后采用三维FDTD方法实现了虚拟声源发射的线性和非线性声场仿真。由于FDTD法计算时间过长，一些替代算法被相继提出，包括混合有限差分/相位投影算法，基于k空间的声波传播模型数值算法等。Leduc等在最近的一项研究中，采用该方法不仅实现了穿颅聚焦，还通过迭代放置额外点生源的方式，实现了对不需要的多余聚焦区域(比如由驻波引起的额外焦域)的消除。
综上所述，植入水听器法和空化微泡法由于其侵入性和潜在的安全风险，在临床应用中有很大的局限性，不适用于超声深脑刺激和颅内HIFU消融治疗。而虚拟声源仿真法便于在治疗前制订周密计划并反复优化，有助于提高治疗效果和安全性，是目前为止最适合在临床上使用的方法。但是，这种方法需要患者先进行三维CT扫描，不仅会增加患者所受到的辐射剂量，增加患者罹患癌症的风险(据统计，美国癌症患者中有0.4％的比例是由于照射CT引起的)，而且无论是进行深脑神经刺激还是颅内HIFU消融治疗，都是在核磁共振(MRI)的引导下进行的，因此必须将CT图像和MRI图像进行精确配准才能保证治疗位置的准确性，这些额外步骤会增加治疗的成本、耗时和配准误差引起的附加风险。
本申请所提供的超声深部脑刺激方法，可无需对待治疗对象的头部进行CT扫描的情况下，即可进行超声深脑刺激或颅内HIFU消融治疗。
首先，请参见图3所示对本申请所示的对目标模型的具体训练过程进行说明：
步骤301、CT扫描设备对多个作为训练样本的头部进行CT扫描以获取用于训练目标模型的头部CT图像数据。
本实施例中，可根据需要采集多个训练样本，该训练样本可为人或动物等待训练样本，CT扫描设备对所有训练样本的头部进行三维的CT扫描以获取各个训练样本的用于训练目标模型的头部CT图像数据。
步骤302、电子设备获取所述用于训练目标模型的头部CT图像数据。
在CT扫描设备获取到所述用于训练目标模型的头部CT图像数据的情况下，所述CT扫描设备可将所述用于训练目标模型的头部CT图像数据发送给所述电子设备，本实施例对所述电子设备获取所述用于训练目标模型的头部CT图像数据的具体过程不做限定，例如，所述CT扫描设备可直接将所述用于训练目标模型的头部CT图像数据发送给所述电子设备，也可经过其他设备将所述用于训练目标模型的头部CT图像数据发送给所述电子设备。
步骤303、核磁共振设备对所述待训练样本的头部进行三维磁共振成像扫描以获取用于训练目标模型的头部三维核磁共振图像数据。
对已经标定好的所有待训练样本的头部的相同成像截面进行三维磁共振成像扫描以获取所述用于训练目标模型的头部三维核磁共振图像数据。
步骤304、电子设备获取所述用于训练目标模型的头部三维核磁共振图像数据。
在核磁共振设备获取到所述用于训练目标模型的头部三维核磁共振图像数据的情况下，所述核磁共振设备可将所述用于训练目标模型的头部三维核磁共振图像数据发送给所述电子设备，本实施例对所述电子设备获取所述用于训练目标模型的头部三维核磁共振图像数据的具体过程不做限定，例如，所述核磁共振设备可直接将所述用于训练目标模型的头部三维核磁共振图像数据发送给所述电子设备，也可经过其他设备将所述用于训练目标模型的头部三维核磁共振图像数据发送给所述电子设备。
步骤305、电子设备对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取目标模型。
本实施例所示的目标模型用于根据头部三维核磁共振图像数据预测出对应的合成头部CT图像数据。
具体的，本实施例所示的电子设备可通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
可选的，本实施例所示的所述机器学习的方法可为生成式对抗网络；
本实施例所示的所述生成式对抗网络(GAN，Generative Adversarial Networks)包括两个部分，一个全卷积网络(FCN，Fully convolutional Networks)和一个卷积神经网络(CNN，Convolutional Neural Networks)，两个部分被同步训练。
在本实施例中，所述电子设备可通过所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据，对生成式对抗网络进行训练以获取所述目标模型。
其中，FCN被训练来利用所述用于训练目标模型的头部CT图像数据生成更加接近真实CT图像的用于训练目标模型的头部CT图像数据，CNN被训练来区分差别已经很小的真实CT图像和用于训练目标模型的头部CT图像数据，两个网络因此形成一种对抗性关系，最终训练所得的GAN也因此可以达到很高的性能，可以利用用于训练目标模型的头部三维核磁共振图像数据计算出非常接近真实CT图像的用于训练目标模型的头部CT图像数据。
可选的，本实施例所示的机器学习的方法还可为随机森林算法，下面对随机森林算法进行说明：
随机森林(Random Forest,RF)是一种集成机器学习算法，核心思想就是将若干棵独立的决策树集成在一起，每棵决策树独立地产生分类结果，最后通过投票得到最终分类结果。随机森林的‘随机’有两层含义，一是随机筛选数据，一是随机筛选特征或者说变量。每棵决策树所依据的数据，都是利用随机重采样方法(bootstrap)有放回地随机抽取原始数据中的若干个样本,构成不同的自助样本集。之后在自助样本集上利用随机特征选取方法生成决策树。
生成随机森林的具体步骤如下:
(1)从原始的训练数据集中,应用自助(bootstrap)方法，有放回地随机抽取K个新的自助样本集，并由此构建K棵分类回归树,每次未被抽到的样本组成了K个袋外数据(out-of-bag,OOB)。
(2)设有数据集有n个特征,则在每一棵树的每个节点处随机抽取m个特征(m≤n)，在实际运用中，对于分类任务一般采用原特征个数的平方根，对于回归任务一般采用原特征个数的1/3。通过计算每个特征蕴含的信息量,在m个特征中选择一个最具有分类能力的特征进行节点分裂。
(3)每棵树最大限度地生长,不做任何剪裁。
(4)将生成的多棵树组成随机森林,用随机森林对新的数据进行分类,分类结果按树分类器的投票多少而定。
随机森林算法具有分析复杂相互作用分类特征的能力，并且具有较快的学习速度。其变量重要性度量方法可以作为高维数据的特征选择工具。随机森林的变量重要性度量方法：特征选择是指从原始特征集中选择使某种评估标准最优的特征子集或者说变量子集。对于 特征选择，首要是评估变量的重要性，随机森林评估变量重要性有两种方式，一种是平均不纯度减少(Mean decrease impurity，MDI)，一种是可以采用基于袋外数据(OOB)的平均分类准确率减少(Mean decreaseaccuracy，MDA)。
具体的，本实施例所示的电子设备可通过随机森林算法对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
采用本实施例所示的方法，电子设备可对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型，基于所述目标模型即可根据头部三维核磁共振图像数据预测出对应的合成头部CT图像数据，从而在对待治疗对象的头部进行超声深部脑刺激时，仅需要对待治疗对象的头部进行核磁共振的拍摄即可，无需进行CT拍摄，可以消除患者受到辐射引发癌症的风险，提高治疗的安全性，同时可以简化治疗步骤，大幅缩短治疗时间，降低治疗成本。
以下结合图4所示对基于所述目标模型如何进行超声深部脑刺激的具体过程进行说明：
步骤401、核磁共振设备对待治疗对象的头部进行三维磁共振成像扫描以获取头部三维核磁共振图像数据。
在需要对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗的情况下，核磁共振设备可对待治疗对象的头部进行三维磁共振成像扫描以获取到所述头部三维核磁共振图像数据。
步骤402、电子设备获取所述头部三维核磁共振图像数据。
在核磁共振设备获取到所述待治疗对象的所述头部三维核磁共振图像数据的情况下，所述核磁共振设备可将所述头部三维核磁共振图像数据发送给所述电子设备，本实施例对所述电子设备获取所述头部三维核磁共振图像数据的具体过程不做限定，例如，所述核磁共振设备可直接将所述头部三维核磁共振图像数据发送给所述电子设备，也可经过其他设备将所述头部三维核磁共振图像数据发送给所述电子设备。
步骤403、电子设备将所述头部三维核磁共振图像数据输入至通过预先训练获得的目标模型。
具体的，所述头部三维核磁共振图像数据作为输入，输入至预先训练的所述目标模型中，所述目标模型的具体说明请详见图3所示的实施例，具体在本实施例中不做限定，只要所述目标模型能够根据所述头部三维核磁共振图像数据预测出对应的合成头部电子计算机断层扫描CT图像数据即可。
步骤404、电子设备获取所述目标模型输出的所述合成头部CT图像数据。
步骤405、电子设备根据所述头部三维核磁共振图像数据以及所述合成头部CT图像数据建立头部三维数字模型。
具体的，所述电子设备对所述头部三维核磁共振图像数据以及所述合成头部CT图像数据进行三维重建和配准，建立起所述待治疗对象的头部的括颅骨和脑部组织结构、密度及 声学参数的头部三维数字模型。
步骤406、电子设备根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列。
具体的，在电子设备获取到所述头部三维数字模型的情况下，电子设备可选择一个或多个待刺激的脑部神经核团的位置，并在所述头部三维数字模型对所述待刺激的脑部神经核团的位置进行定位。
所述电子设备还可根据所述头部三维数字模型得到颅骨和脑组织的结构及密度，然后根据颅骨和脑组织的结构及密度推算头部的声学参数，该声学参数包括但不限于声速、衰减系数。
本实施例所示的电子设备可根据实际空间位置对头部三维数字模型及所述超声换能器阵列的三维数字模型的虚拟空间位置进行调整，将头部三维数字模型及超声换能器阵列的三维数字模型的虚拟空间位置按照所希望采用的实际空间位置调整到位，在需要聚焦的一个或多个位置放置虚拟声源，电子设备可仿真所述虚拟声源所发出的超声波在所述头部三维数字模型中的传播状态。当所述超声波传播到所述超声换能器阵列所处的虚拟空间位置时，电子设备仿真所述超声换能器阵列的电压信号。
具体地，当超声波传播到超声换能器阵列所处的虚拟空间位置(按照所希望采用的实际空间位置调整到位后的虚拟空间位置)时，仿真超声波在超声换能器阵列表面的声强信号和声压信号，根据声强信号和声压信号以及超声换能器阵列的压电转换参数仿真出超声换能器阵列的电压信号，即将声强信号和声压信号仿真出超声换能器经过压电转换后的电压信号。电子设备对所述电压信号进行时间反演，生成时间反演信号，作为所述超声发射序列。
具体地，该电压信号进行的时间反演为按时间前后顺序的翻转，该方法可以称为时间反演方法。该电压信号按时间前后顺序翻转后，再用来激励超声换能器阵列，所产生的超声波会在已放置虚拟声源的一个或多个位置聚焦。因此，通过该超声时间反演，本发明可以得到刺激指定的脑部神经核团所需要的，能够完成穿颅聚焦的超声发射序列。
步骤407、电子设备控制所述超声换能器阵列按照所述超声发射序列发射超声波。
本实施例中，将需要进行超声深脑刺激的待治疗对象的头部，利用头部固定和定位装置固定在磁共振成像系统中的指定位置，同时将超声换能器阵列也按照预先设计好的位置固定好。电子设备发出指令以控制超声换能器阵列，按照超声发射序列进行超声发射，所述超声波用于对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗。
上述对待刺激的脑部神经核团实施超声神脑刺激的方式包括：脉冲方式、多周期方式及编码方式等，本发明不以此为限。
可选的，本实施例所示的方法也可以进一步简化。例如，省略对超声波传播全过程的精确仿真，直接将焦点位置与超声换能器阵列各阵元位置进行连线，假设超声波在两者间完全直线传播，则连线即为超声波的传播路径，超声波在穿过颅骨的部分路径上的传播速度较快，可以根据上述从合成头部CT图像数据中估算的颅骨声速分布等参数，计算这部分路径对超声波传播时间的影响，再据此计算出需要如何修正超声换能器各个阵元的发射延时，才能使各个阵元发射的超声波同时到达聚焦位置，实现穿颅聚焦。利用上述简化方法， 同样可以得到能够完成穿颅聚焦的超声发射序列，可以大幅减少计算量，缩短计算时间。
采用本实施例所示的方法，可无需对待治疗对象进行CT拍摄的情况下，只需将头部三维核磁共振图像数据输入至目标模型，电子设备即可获取到所述目标模型输出的合成头部CT图像数据，电子设备即可根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列，电子设备控制所述超声换能器阵列按照所述超声发射序列发射超声波，通过该超声波可对待治疗对象的头部进行超声深部脑刺激。本实施例所示的方法可在保证超声穿颅聚焦焦点定位精度满足需求的前提下，消除带治疗对象在CT成像过程中受到辐射引发癌症的风险，提高治疗的安全性，同时可以简化治疗步骤，大幅缩短治疗时间，降低治疗成本。
以下对本实施例所示的方法的精确性进行分析：
为实现对带诊断对象进行精确的超声深部脑刺激，则根据目标模型所估计出来的合成头部CT图像数据和针对的CT图像数据越一致，则超声深部脑刺激会越精确。
具体还可参见下述公式所示：
Figure PCTCN2019083433-appb-000001
其中，MAX表示图像灰度的最大数值。MSE为均方差。
峰值信噪比(PSNR)是一个衡量信号或图像质量的指标，即信号最大可能功率和影响其精度的破坏性噪声功率的比值。由于许多信号都有非常宽的动态范围，峰值信噪比常用对数分贝单位来表示。在图像处理中，要对图像进行客观的评价，常常需要计算PSNR。PSNR是衡量图像失真或是噪声水平的客观标准。2个图像之间PSNR值越大，则越相似。普遍基准为30dB，30dB以下的图像劣化较为明显。
采用本实施例所示的目标训练方法所估计出来的所述合成头部CT图像数据和针对同一待治疗对象的真实CT图像之间比较，其PSNR值一定达到27.6dB，非常接近30dB，因此可以认为该方法得到的合成头部CT图像数据完全可以用来进行颅骨声学参数估算和实现超声穿颅聚焦。
本发明在二维平面中利用超声时间反演软件获得了实现穿颅聚焦所需超声发射序列的方法的仿真实验。实验结果如图5及图6所示，图5中(0，40)处的小圆点为虚拟声源的初始位置，图6为利用时间反演方法仿真实现的超声聚焦效果；
图上边框为1024阵元线阵超声换能器阵列。图5及图6中灰色部分为由CT扫描图像重建出的颅骨模型，并根据颅骨和脑组织的结构及密度推算头部的声学参数，然后导入电子设备所运行的仿真软件中设置为二维平面对应计算节点的相应数值。从仿真结果看，尽管经过时间反演发射后，超声聚焦点的横向和纵向尺寸都比原声源尺寸有所增加，但能量大部分依然集中在预计位置，可以满足精确定点刺激和保证周围其他区域受到较小影响的双重要求。只要将该方法从二维平面扩展到三维空间数字模型中，即可满足本发明的需要。
在超声刺激小鼠颅脑诱导动作响应的实验方面：麻醉后的小鼠，头颅顶部去毛，固定于脑立体定位仪。超声换能器阵列由脑立体定位仪精确定位，并贴近颅骨向运动皮层辐射脉冲超声。超声的刺激，能同步诱发小鼠产生肢体动作响应。
以下结合图7所示对本实施例所提供的能够实现上述实施例所示的超声深部脑刺激方法的电子设备的具体结构进行说明：
如图7所示，本实施例所示的电子设备包括：
获取单元701，用于获取头部三维核磁共振图像数据，所述头部三维核磁共振图像数据为对待治疗对象的头部进行三维磁共振成像扫描所获取到的图像数据；
可选的，所述获取单元701还用于，用于训练目标模型的头部CT图像数据，所述用于训练目标模型的头部CT图像数据为对多个作为训练样本的头部进行CT扫描所获取到的图像数据；
所述获取单元701还用于，获取用于训练目标模型的头部三维核磁共振图像数据，所述用于训练目标模型的头部三维核磁共振图像数据为对多个所述训练样本的头部进行三维磁共振成像扫描所获取到的图像数据；
训练单元702，用于对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
所述训练单元702具体用于，通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
具体的，所述训练单元702具体用于，通过所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据，对生成式对抗网络进行训练以获取所述目标模型。
具体的，所述训练单元702具体用于，通过随机森林算法对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
输入单元703，用于将所述头部三维核磁共振图像数据作为输入，输入至通过预先训练获得的目标模型中，所述目标模型用于根据所述头部三维核磁共振图像数据获得对应的合成头部电子计算机断层扫描CT图像数据；
建立单元704，用于根据所述头部三维核磁共振图像数据以及所述合成头部CT图像数据建立头部三维数字模型；
生成单元705，用于根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列；
控制单元706，用于控制所述超声换能器阵列按照所述超声发射序列发射超声波，所述超声波用于对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗。
所述电子设备执行超声深部脑刺激方法的具体过程，请详见上述实施例所示，具体在本实施例中不做赘述。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统，装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统，装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的 划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本发明各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims (12)

  1. 一种超声深部脑刺激方法，其特征在于，所述方法包括：
    获取头部三维核磁共振图像数据，所述头部三维核磁共振图像数据为对待治疗对象的头部进行三维磁共振成像扫描所获取到的图像数据；
    将所述头部三维核磁共振图像数据作为输入，输入至通过预先训练获得的目标模型中，所述目标模型用于根据所述头部三维核磁共振图像数据获得对应的合成头部电子计算机断层扫描CT图像数据；
    根据所述头部三维核磁共振图像数据以及所述合成头部CT图像数据建立头部三维数字模型；
    根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列；
    控制所述超声换能器阵列按照所述超声发射序列发射超声波，所述超声波用于对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗。
  2. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    获取用于训练目标模型的头部CT图像数据，所述用于训练目标模型的头部CT图像数据为对多个作为训练样本的头部进行CT扫描所获取到的图像数据；
    获取用于训练目标模型的头部三维核磁共振图像数据，所述用于训练目标模型的头部三维核磁共振图像数据为对多个所述训练样本的头部进行三维磁共振成像扫描所获取到的图像数据。
  3. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
  4. 根据权利要求3所述的方法，其特征在于，所述对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型包括：
    通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
  5. 根据权利要求4所述的方法，其特征在于，所述通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型包括：
    通过所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据，对生成式对抗网络进行训练以获取所述目标模型。
  6. 根据权利要求4所述的方法，其特征在于，所述通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型包括：
    通过随机森林算法对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
  7. 一种电子设备，其特征在于，包括：
    获取单元，用于获取头部三维核磁共振图像数据，所述头部三维核磁共振图像数据为对待治疗对象的头部进行三维磁共振成像扫描所获取到的图像数据；
    输入单元，用于将所述头部三维核磁共振图像数据作为输入，输入至通过预先训练获得的目标模型中，所述目标模型用于根据所述头部三维核磁共振图像数据获得对应的合成头部电子计算机断层扫描CT图像数据；
    建立单元，用于根据所述头部三维核磁共振图像数据以及所述合成头部CT图像数据建立头部三维数字模型；
    生成单元，用于根据所述头部三维数字模型和超声换能器阵列的三维数字模型生成超声发射序列；
    控制单元，用于控制所述超声换能器阵列按照所述超声发射序列发射超声波，所述超声波用于对待治疗对象的头部实现穿颅聚焦，进行超声深部脑刺激或超声热消融治疗。
  8. 根据权利要求7所述的电子设备，其特征在于，所述获取单元还用于，用于训练目标模型的头部CT图像数据，所述用于训练目标模型的头部CT图像数据为对多个作为训练样本的头部进行CT扫描所获取到的图像数据；
    所述获取单元还用于，获取用于训练目标模型的头部三维核磁共振图像数据，所述用于训练目标模型的头部三维核磁共振图像数据为对多个所述训练样本的头部进行三维磁共振成像扫描所获取到的图像数据。
  9. 根据权利要求8所述的电子设备，其特征在于，所述电子设备还包括：
    训练单元，用于对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
  10. 根据权利要求9所述的电子设备，其特征在于，
    所述训练单元具体用于，通过机器学习的方法，对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
  11. 根据权利要求10所述的电子设备，其特征在于，
    所述训练单元具体用于，通过所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据，对生成式对抗网络进行训练以获取所述目标模型。
  12. 据权利要求10所述的电子设备，其特征在于，
    所述训练单元具体用于，通过随机森林算法对所述用于训练目标模型的头部三维核磁共振图像数据以及所述用于训练目标模型的头部CT图像数据进行训练以获取所述目标模型。
PCT/CN2019/083433 2019-04-19 2019-04-19 一种实现超声穿颅聚焦的方法以及电子设备 WO2020211076A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/083433 WO2020211076A1 (zh) 2019-04-19 2019-04-19 一种实现超声穿颅聚焦的方法以及电子设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/083433 WO2020211076A1 (zh) 2019-04-19 2019-04-19 一种实现超声穿颅聚焦的方法以及电子设备

Publications (1)

Publication Number Publication Date
WO2020211076A1 true WO2020211076A1 (zh) 2020-10-22

Family

ID=72837980

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/083433 WO2020211076A1 (zh) 2019-04-19 2019-04-19 一种实现超声穿颅聚焦的方法以及电子设备

Country Status (1)

Country Link
WO (1) WO2020211076A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101541378A (zh) * 2006-11-22 2009-09-23 韩国标准科学研究院 通过控制电子信号来使用聚焦超声波的设备及其使用方法
US20110092800A1 (en) * 2002-04-30 2011-04-21 Seung-Schik Yoo Methods for modifying electrical currents in neuronal circuits
CN104548390A (zh) * 2014-12-26 2015-04-29 中国科学院深圳先进技术研究院 一种超声深部脑刺激方法及系统
CN104545919A (zh) * 2014-12-31 2015-04-29 中国科学院深圳先进技术研究院 一种超声穿颅聚焦的方法
CN108430579A (zh) * 2015-11-11 2018-08-21 浩宇生医股份有限公司 通过使用超声系统治疗脑肿瘤的方法和试剂盒

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110092800A1 (en) * 2002-04-30 2011-04-21 Seung-Schik Yoo Methods for modifying electrical currents in neuronal circuits
CN101541378A (zh) * 2006-11-22 2009-09-23 韩国标准科学研究院 通过控制电子信号来使用聚焦超声波的设备及其使用方法
CN104548390A (zh) * 2014-12-26 2015-04-29 中国科学院深圳先进技术研究院 一种超声深部脑刺激方法及系统
CN104545919A (zh) * 2014-12-31 2015-04-29 中国科学院深圳先进技术研究院 一种超声穿颅聚焦的方法
CN108430579A (zh) * 2015-11-11 2018-08-21 浩宇生医股份有限公司 通过使用超声系统治疗脑肿瘤的方法和试剂盒

Similar Documents

Publication Publication Date Title
Constans et al. A 200–1380-kHz quadrifrequency focused ultrasound transducer for neurostimulation in rodents and primates: transcranial in vitro calibration and numerical study of the influence of skull cavity
US11826585B2 (en) Adaptive refocusing of ultrasound transducer arrays using image data
Younan et al. Influence of the pressure field distribution in transcranial ultrasonic neurostimulation
Tanter et al. Compensating for bone interfaces and respiratory motion in high-intensity focused ultrasound
Clement et al. A non-invasive method for focusing ultrasound through the human skull
CA2898503C (en) System and method for measuring and correcting ultrasound phase distortions induced by aberrating media
JP4558504B2 (ja) 超音波治療における組織異常の修正
US20160038770A1 (en) Focused transcranial ultrasound systems and methods for using them
CA3046392A1 (en) Systems and methods for performing transcranial ultrasound therapeutic and imaging procedures
Yu et al. Design of a volumetric imaging sequence using a vantage-256 ultrasound research platform multiplexed with a 1024-element fully sampled matrix array
CN104545919B (zh) 一种超声穿颅聚焦的方法
CN105536156A (zh) 一种基于大规模面阵元的超声脑刺激或调控方法及装置
CN109893784A (zh) 一种实现超声穿颅聚焦的方法以及电子设备
Maimbourg et al. Computationally efficient transcranial ultrasonic focusing: Taking advantage of the high correlation length of the human skull
Estrada et al. Spherical array system for high-precision transcranial ultrasound stimulation and optoacoustic imaging in rodents
CN111494816A (zh) 一种超声精准自适应聚焦系统和方法
Pichardo BabelBrain: An Open-Source Application for Prospective Modeling of Transcranial Focused Ultrasound for Neuromodulation Applications
JP2023549792A (ja) 超音波手技のためのマルチパラメトリック最適化
KR20140102994A (ko) 관심 영역 내에 다중 초점을 형성하는 초음파를 생성하는 방법, 장치 및 ｈｉｆｕ 시스템
Thies et al. Real-time visualization of a focused ultrasound beam using ultrasonic backscatter
Liu et al. Artifact suppression for passive cavitation imaging using U-Net CNNs with uncertainty quantification
WO2020211076A1 (zh) 一种实现超声穿颅聚焦的方法以及电子设备
Zhang et al. Research on the modulation method of transcranial focused ultrasound based on multi-array transducer
Andrés et al. Multifocal acoustic holograms for deep-brain neuromodulation and BBB opening
WO2019191863A1 (zh) 一种超声成像系统、方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19925271

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19925271

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 18.03.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19925271

Country of ref document: EP

Kind code of ref document: A1