PREDICTION OF THE SKULL'S ACOUSTIC PARAMETERS IN TRANSCRANIAL FOCUSED ULTRASOUND BASED ON NEURAL NETWORK
详细信息 查看官网全文
- 英文论文题名：PREDICTION OF THE SKULL'S ACOUSTIC PARAMETERS IN TRANSCRANIAL FOCUSED ULTRASOUND BASED ON NEURAL NETWORK
- 论文作者：Xiang-da WANG ; Nan-xing LI ; Wei-jun LIN ; Chang SU
- 英文论文作者：Xiang-da WANG ; Nan-xing LI ; Wei-jun LIN ; Chang SU ; State Key Laboratory of Acoustics ; Institute of Acoustics ; Chinese Academy of Sciences ; National Network New Media Engineering Research Center ; Institute of Acoustics ; Chinese Academy of Sciences
- 年：2016
- 作者机构：State Key Laboratory of Acoustics, Institute of Acoustics, Chinese Academy of Sciences;National Network New Media Engineering Research Center,Institute of Acoustics, Chinese Academy of Sciences;
- 英文论文关键词：Transcranial focused ultrasound ; Deeping learning ; Neural network
- 会议召开时间：2016-10-21
- 会议录名称：2016年全国压电和声波理论及器件应用研讨会论文摘要集
- 英文会议录名称：Abstract Book of 2016 Symposium on Piezoelectricity, Acoustic Waves and Device Applications
- 语种：英文
- 分类号：R445.1
- 学会代码：AGLU
- 会议名称：2016年全国压电和声波理论及器件应用研讨会
- 会议地点：中国陕西西安
- 主办单位：中国力学学会、中国声学学会、IEEE UFFC分会
- 学会名称：中国力学学会
- 页数：2
- 文件大小：156k
- 原文格式：D
- 会议级别：全国
摘要
Background, Motivation and Objective The model for calculating the transcranial focused ultrasound(tc FUS) field is the Westervelt equation, in which the acoustic parameters of the skull are necessary. Thus, in order to achieve the precise focusing of transcranial ultrasound, it's of great importance and significance to obtain accurate acoustic parameters of the skull. In the past few decades, deep learning based on the neural network has got a rapid progress and is of much popularity. In this paper, through the use of computed tomography(CT) of a monkey skull from Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, we have proposed a method utilizing feed-forward neural network(FFNN) to precisely predict the acoustic parameters of the skull based on its CT with tiny errors. Moreover, the predicting results of the acoustic parameters are applied to numerically simulate the transcranial focused ultrasound field, so as to further verify the effectiveness of the proposed method. Statement of Contribution/Methods We employ the method put forward by M. Fink et al to generate the learning data set. In their model, the acoustic parameters of the skull including density, sound velocity and acoustic absorption coefficient can be calculated based on the skull's CT data. With the learning data set, our goal is to acquire the mapping relationship between the CT data and the different acoustic parameters through FFNN, namely the appropriate FFNN structure. In the following, focused ultrasonic fields calculated by the numerical algorithm of finite difference time domain(FDTD) with the predicted and the hypothetically real acoustic parameters of the monkey skull are compared and analyzed to further demonstrate the effectiveness of the proposed method. Results Prediction results show that the density, sound velocity and acoustic absorption coefficient are precisely predicted by our method with the accuracy rate of 99.95% at least. Furthermore, comparison of the numerical results of focused ultrasonic fields simulated with the predicted and the hypothetically real acoustic parameters also shows well agreement.Discussion and Conclusions This paper studies the utilization of the neural network to predict acoustic parameters based on CT. The acoustic parameters of the monkey skull predicted through this method are in good agreement with those calculated through the model proposed by M. Fink et al. The maximal relative error is lower than 0.05% and can be further decreased when the learning data set is larger or the learning time is longer. Moreover, each control group of focused ultrasonic fields is demonstrated to confirm the effectiveness of the proposed method. Most importantly, if abundant true experimental data of CTs and acoustic parameters of skulls was gained, we would be freed from the difficult and troublesome measuring experiments to get the skull's acoustic parameters by acquiring the mapping relation through our method.
Background, Motivation and Objective The model for calculating the transcranial focused ultrasound(tc FUS) field is the Westervelt equation, in which the acoustic parameters of the skull are necessary. Thus, in order to achieve the precise focusing of transcranial ultrasound, it's of great importance and significance to obtain accurate acoustic parameters of the skull. In the past few decades, deep learning based on the neural network has got a rapid progress and is of much popularity. In this paper, through the use of computed tomography(CT) of a monkey skull from Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, we have proposed a method utilizing feed-forward neural network(FFNN) to precisely predict the acoustic parameters of the skull based on its CT with tiny errors. Moreover, the predicting results of the acoustic parameters are applied to numerically simulate the transcranial focused ultrasound field, so as to further verify the effectiveness of the proposed method. Statement of Contribution/Methods We employ the method put forward by M. Fink et al to generate the learning data set. In their model, the acoustic parameters of the skull including density, sound velocity and acoustic absorption coefficient can be calculated based on the skull's CT data. With the learning data set, our goal is to acquire the mapping relationship between the CT data and the different acoustic parameters through FFNN, namely the appropriate FFNN structure. In the following, focused ultrasonic fields calculated by the numerical algorithm of finite difference time domain(FDTD) with the predicted and the hypothetically real acoustic parameters of the monkey skull are compared and analyzed to further demonstrate the effectiveness of the proposed method. Results Prediction results show that the density, sound velocity and acoustic absorption coefficient are precisely predicted by our method with the accuracy rate of 99.95% at least. Furthermore, comparison of the numerical results of focused ultrasonic fields simulated with the predicted and the hypothetically real acoustic parameters also shows well agreement.Discussion and Conclusions This paper studies the utilization of the neural network to predict acoustic parameters based on CT. The acoustic parameters of the monkey skull predicted through this method are in good agreement with those calculated through the model proposed by M. Fink et al. The maximal relative error is lower than 0.05% and can be further decreased when the learning data set is larger or the learning time is longer. Moreover, each control group of focused ultrasonic fields is demonstrated to confirm the effectiveness of the proposed method. Most importantly, if abundant true experimental data of CTs and acoustic parameters of skulls was gained, we would be freed from the difficult and troublesome measuring experiments to get the skull's acoustic parameters by acquiring the mapping relation through our method.
Background, Motivation and Objective The model for calculating the transcranial focused ultrasound(tc FUS) field is the Westervelt equation, in which the acoustic parameters of the skull are necessary. Thus, in order to achieve the precise focusing of transcranial ultrasound, it's of great importance and significance to obtain accurate acoustic parameters of the skull. In the past few decades, deep learning based on the neural network has got a rapid progress and is of much popularity. In this paper, through the use of computed tomography(CT) of a monkey skull from Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, we have proposed a method utilizing feed-forward neural network(FFNN) to precisely predict the acoustic parameters of the skull based on its CT with tiny errors. Moreover, the predicting results of the acoustic parameters are applied to numerically simulate the transcranial focused ultrasound field, so as to further verify the effectiveness of the proposed method. Statement of Contribution/Methods We employ the method put forward by M. Fink et al to generate the learning data set. In their model, the acoustic parameters of the skull including density, sound velocity and acoustic absorption coefficient can be calculated based on the skull's CT data. With the learning data set, our goal is to acquire the mapping relationship between the CT data and the different acoustic parameters through FFNN, namely the appropriate FFNN structure. In the following, focused ultrasonic fields calculated by the numerical algorithm of finite difference time domain(FDTD) with the predicted and the hypothetically real acoustic parameters of the monkey skull are compared and analyzed to further demonstrate the effectiveness of the proposed method. Results Prediction results show that the density, sound velocity and acoustic absorption coefficient are precisely predicted by our method with the accuracy rate of 99.95% at least. Furthermore, comparison of the numerical results of focused ultrasonic fields simulated with the predicted and the hypothetically real acoustic parameters also shows well agreement.Discussion and Conclusions This paper studies the utilization of the neural network to predict acoustic parameters based on CT. The acoustic parameters of the monkey skull predicted through this method are in good agreement with those calculated through the model proposed by M. Fink et al. The maximal relative error is lower than 0.05% and can be further decreased when the learning data set is larger or the learning time is longer. Moreover, each control group of focused ultrasonic fields is demonstrated to confirm the effectiveness of the proposed method. Most importantly, if abundant true experimental data of CTs and acoustic parameters of skulls was gained, we would be freed from the difficult and troublesome measuring experiments to get the skull's acoustic parameters by acquiring the mapping relation through our method.
引文