JP6399554B2 - MR image super-resolution system - Google Patents

Description

  The present invention connects a client facility equipped with an MRI apparatus and a server facility equipped with MR image super-resolution means so as to be able to transmit and receive data via a cloud network, and captures and generates low-resolution MR image data captured by the client facility. The present invention relates to a system capable of generating super-resolution MR image data by performing high-resolution processing with server equipment. Furthermore, the present invention relates to a system that enables more accurate and precise structural analysis, functional analysis, or feedback using super-resolution MR image data.

  In the present specification, MR refers to “magnetic resonance” and MRI refers to “magnetic resonance imaging”.

  Taking measures against patients with dementia, which increase with aging, is considered a social urgent need. Prevention of dementia, early detection, and early treatment are effective in suppressing the increase in patients with dementia. As one means for that purpose, a diagnosis using a brain MR image captured by an MRI apparatus has been conventionally performed. In this case, the higher the resolution of the MR image, the more accurate and precise diagnosis is possible.

  By the way, advanced medical equipment such as general clinics and research facilities are generally concentrated in urban areas, and there is a problem of regional disparity that sufficient medical services cannot be provided in depopulated areas such as rural areas. There are regional disparities in the above-mentioned measures against dementia. In view of this, it has been studied to construct a medical service network using ICT (Information and Communication Technology), and to use this network to eliminate regional disparities. For example, by building a system that connects a core medical center and multiple regional medical centers installed within the jurisdiction of this core medical center through a network, an environment that can provide uniform medical services regardless of the region is established. It has been tried to do.

  As medical services using networks, medical data and electronic medical records collected at medical centers in various places are shared, or stored and backed up at core medical centers.

  In Patent Document 1, when there is no specialized neurosurgeon in a hospital where an image acquisition device such as an MRI apparatus is installed, a diagnosis support device is installed in a hospital where the neurosurgeon is located, and these hospitals are connected via a network. Thus, it is described that a diagnosis by a neurosurgeon can be made based on a brain MR image taken in a hospital without a neurosurgeon.

Further, in Patent Document 2, in order to improve the diagnosis efficiency of a doctor using an image, an image such as gradation processing for adjusting contrast, frequency enhancement processing for adjusting sharpness, dynamic range compression processing, or the like is applied to medical image data. A medical image processing system that performs processing is described.

JP 2006-43200 A Japanese Patent No. 4385707

  In order to perform highly accurate diagnosis of mild cognitive impairment and dementia diagnosis, it is desirable to use a brain MR image with as high a resolution as possible. However, since the number of high-resolution MRI apparatuses to be installed is limited, the area where a highly accurate diagnosis using a high-resolution MR image can be received is limited. That is, there is a problem that there is a regional disparity in the accuracy and precision of the diagnosis of mild cognitive impairment and dementia required in an aging society. The inventions of Patent Documents 1 and 2 are not techniques for eliminating regional disparities by increasing the resolution.

In order to achieve the above object, the present invention according to claim 1 provides:
Cloud network,
Two or more client facilities each connected to a cloud network so as to be able to transmit and receive data, and each having an MRI apparatus for generating MR image data;
A server facility that is connected to the cloud network so as to be able to transmit and receive data, and that includes super-resolution means for generating high-resolution MR image data by performing high-resolution processing on MR image data,
The server facility has a function of receiving the original MR image data generated by the client facility through the cloud network, and a function of generating super-resolution MR image data by the super-resolution means based on the received original MR image data. This is a super-resolution system for MR images.

  The present invention according to claim 2 is the MR image super-resolution system according to claim 1, wherein the server facility further has a function of transmitting the generated super-resolution MR image data to one or more client facilities. It is.

According to the third aspect of the present invention, the MRI apparatus of the client facility further has a function of generating a plurality of original MR image data in which the imaging positions are shifted by a predetermined distance,
3. The MR image according to claim 1, wherein the super-resolution unit of the server facility generates one super-resolution MR image data from the plurality of original MR image data generated by the client facility. It is a super-resolution system.

According to a fourth aspect of the present invention, the server facility further includes an MR image database in which standard image pattern data having a high resolution corresponding to the original MR image data generated by the client facility is stored in advance,
3. The MR image of claim 1, wherein the super-resolution means generates super-resolution MR image data based on a comparison between the original MR image data and the standard image pattern data corresponding to the original MR image data. It is a super-resolution system.

According to a fifth aspect of the present invention, the server facility is configured to store the individual MR image pattern data stored in advance for each individual at a higher resolution than the original MR image data generated by the client facility. A database,
The super-resolution means generates super-resolution MR image data based on a comparison between the original MR image data generated by the client equipment and the corresponding individual MR image pattern data. Or an MR image super-resolution system according to 2;

The present invention according to claim 6 is an MRI apparatus for client equipment, which is a functional MRI apparatus for generating original functional MR image data reflecting a brain activity state,
The super-resolution unit of the server facility further includes a function of generating super-resolution functional MR image data based on the original functional MR image data generated by the functional MRI apparatus of the client facility. The MR image data super-resolution system according to any one of the above.

In the present invention according to claim 7, the MRI apparatus of the client facility further has a function of generating original functional MR image data continuously or intermittently,
The super-resolution unit of the server facility generates super-resolution functional MR image data continuously or intermittently based on the original functional MR image data generated continuously or intermittently by the MRI apparatus of the client facility. The MR image data super-resolution system according to claim 6, further comprising a function.

According to the first aspect of the present invention, the original MR image data generated by each client facility in two or more client facilities is transmitted to the server facility through the cloud network, and the server facility super-resolution means performs this. Since super-resolution MR image data based on the original MR image data is generated, a super-resolution MR image can be obtained without using a high-resolution MRI apparatus.

  According to the present invention of claim 2, since super-resolution MR image data can be obtained on the client facility side, it is possible to perform more accurate and precise diagnosis in a client facility that does not have a high-resolution MRI apparatus. It is. Therefore, the regional disparity in diagnosis using MR images can be eliminated.

  According to the third aspect of the present invention, the super-resolution means of the server facility is based on a plurality of original MR image data generated by the MRI apparatus of the client facility and whose imaging positions are shifted by a predetermined distance. Since resolving MR image data is generated, super-resolution MR image data with high accuracy can be easily obtained.

  According to the present invention of claim 4, since the super-resolution MR image data is generated using the MR image database in which standard image pattern data having a high resolution corresponding to the original MR image data is stored in advance, the client The number of MR images to be generated on the equipment side can be reduced. A conventionally known apparatus can be used as the MRI apparatus.

  According to the fifth aspect of the present invention, the individual reference MR image database in which the individual MR image pattern data captured for each individual at a higher resolution than the original MR image data generated by the client equipment is stored is used. Since super-resolution MR image data is generated, the number of MR images to be generated on the client equipment side can be reduced. A conventionally known apparatus can be used as the MRI apparatus.

  According to the present invention of claim 6, super-resolution functional MR image data can be generated by the super-resolution means of the server facility from the original functional MR image data generated by the MRI apparatus of the client facility, A highly accurate brain function diagnosis can be performed.

  According to the present invention of claim 7, the super-resolution functional MR image data is generated from the original functional MR image data generated continuously or intermittently by the MRI apparatus of the client facility by the super-resolution means of the server facility. Is generated continuously or intermittently, so that the brain function can be diagnosed more accurately and accurately by real-time scanning of the brain state of the test subject.

It is a block diagram which shows typically an example of the super-resolution system of MR image which concerns on this invention. It is drawing explaining the principle which produces | generates a super-resolution MR image by the super-resolution means based on this invention. It is drawing which shows an example which produces | generates a super-resolution MR image by the super-resolution means based on this invention. It is drawing explaining the principle which produces | generates a super-resolution functional MR image by the super-resolution means based on this invention. It is drawing explaining the effect by producing | generating a super-resolution MR image by the super-resolution means based on this invention.

[System configuration]
FIG. 1 is a block diagram schematically showing the configuration of an MR image super-resolution system (hereinafter referred to as the present system) S according to the present invention. The present invention system S, and the cloud network N, each connected to a cloud network N capable of transmitting and receiving data, the MRI apparatus 21 and the data communication means 22 and 2 or more clients equipment 20 provided respectively in the cloud network N capable of transmitting and receiving data And server equipment 10 that is connected and includes super-resolution means 11 and data transmission / reception means 12.

  The cloud network N only needs to be a communication network that can transmit data based on a common protocol between the server facility 10 and the client facility 20, and the type is not particularly limited. An Internet line, a WAN (Wide Area Network), a LAN (Local Area Network), a VPN (Virtual Private Network), etc. can be used, and it is not limited to wired or wireless. However, since a large amount of MR image data is transmitted, it is desirable that the transmission speed be as fast as possible.

The server facility 10 is installed in a general hospital, a basic medical center, a university laboratory, or the like. The data transmission / reception means 12 of the server facility 10 has a function of transmitting / receiving data to / from one or a plurality of client facilities 20 through the cloud network N. The super-resolution means 11 has a function of generating super-resolution MR image data with high resolution based on the original MR image data. In order to realize the super-resolution function, the super-resolution means 11 includes a high-speed arithmetic device and a storage device. It is also conceivable to use a massively parallel cluster computer as the computing means and operate it with a parallel program to realize high-speed computation.

  The client facility 20 is installed in a general medical clinic such as a private hospital or clinic, a regional medical center, a local medical examination center, or the like. The data transmission / reception means 22 of the client facility 20 has a function of transmitting / receiving data to / from the server facility 10 through the cloud network N. Further, communication with other client equipment 20 may be possible. The MRI apparatus 21 is an apparatus that captures an examination subject and generates original MR image data, and a known apparatus can be used.

  However, as described later, when one super-resolution MR image data is generated from a plurality of original MR image data, a function capable of simultaneously generating a plurality of original MR image data slightly different in imaging position by one imaging. It is desirable to have Alternatively, it is also possible to generate a plurality of original MR images with slightly different imaging positions by the function of imaging a plurality of times within a short time.

[Super-resolution method]
A method for generating super-resolution MR image data from the original MR image data generated by the client facility 20 by the super-resolution unit 11 of the server facility 10 will be described. The resolution of MR images generally used in clinical practice is about 1 mm square. Therefore, as shown in FIG. 2A, a plurality of original MR image data whose imaging position shift amount is within the resolution (1 mm in this example) is generated, and the resolution is 0.5 mm by a super-resolution algorithm that combines them. One super-resolution MR image is generated. Means for generating a plurality of image data with positional deviation can be realized by adjusting the imaging coordinates of the MRI apparatus, moving the person to be inspected, moving the imaging sequence, and the like. The super-resolution algorithm is disclosed in, for example, International Publication No. 2011/026923, “Super-resolution in magnetic resonance imaging: A review, Electric Van Reth, Ivan W. K. Tham, Cher Heng. Resonance Part A, Volume 40A, Issue 6, pages 306-325, November 2012 ", etc. can be used.

To generate super-resolution MR image data, it is certain to use eight original MR image data. However, by constructing an appropriate algorithm, super-resolution MR image data can be obtained from four original MR image data. It is possible to generate with high accuracy. That is, as illustrated in FIG. 2B, one super-resolution brain slice MR image data is obtained from four brain slice MR image data (original MR image data) whose imaging positions generated by the client equipment 20 are different within the resolution range. It is possible to generate. In the above algorithm, misalignment, blur, resolution difference, noise, and the like between a plurality of original MR images are taken into consideration so that an optimum super-resolution MR image can be generated.

  The super-resolution MR image data generated in this way may be used for diagnosis by a specialist in a general hospital or the like where the server facility 10 is installed, but the client facility 20 that has captured the original MR image data through the cloud network N. May be transmitted. As a result, in the local clinic where the client facility 20 is installed, the doctor in charge of the subject to be inspected uses the super-resolution MR image with higher resolution than that at the time of imaging, and is more accurate and precise than before. It is possible to make a simple diagnosis. Further, even when image analysis is performed by a diagnosis support system using an analysis system or the like, it is possible to obtain an analysis result that is more accurate and precise than before.

  As a different means for super-resolution of the original MR image data, an MR image in which high resolution standard image pattern data corresponding to the original MR image data generated by the client facility 20 is stored in the server facility 10 in advance. A database (not shown) is provided, and the super-resolution means 11 generates super-resolution MR image data based on a comparison between the original MR image data and the corresponding standard image pattern data. It is also possible. In this case, since the client facility 20 does not need to generate a plurality of original MR image data, the burden on the MRI apparatus 21 provided in the client facility 20 can be reduced.

  As a different means for super-resolution of the original MR image data, the individual MR image pattern captured for each individual in the server facility 10 at a higher resolution than the original MR image data generated by the client facility 20 in advance. A personal reference MR image database (not shown) in which data is stored is provided, and the super-resolution means 11 performs super-resolution based on the comparison between the original MR image data and the corresponding individual MR image pattern data. A configuration for generating resolved MR image data is also conceivable. In this case, since the client facility 20 does not need to generate a plurality of original MR image data, the burden on the MRI apparatus 21 provided in the client facility 20 can be reduced.

[Super-resolution of functional MR images]
The present invention can also super-resolution a functional MR image captured by a functional MRI apparatus (fMRI apparatus). A functional MR image showing a blood flow state of the brain is a dynamic stereoscopic image, and generally has a temporal resolution of about 2-3 seconds per scan, but has a low spatial resolution of about 3 mm. Therefore, if the super-resolution means of the present invention is applied to obtain high-resolution functional MR image data, highly accurate diagnosis can be performed. As shown in FIG. 3, in the super-resolution, one super-resolution MR image data is generated from a plurality of original functional MR image data having a positional deviation within a resolution range (within 1 voxel). The super-resolution functional MR image data has sub-voxels (1/2 voxels in this example) obtained by dividing the voxel as a unit, and therefore has a resolution eight times that of the original functional MR image data.

Since the super-resolution functional MR image data obtained in this manner can express the pattern in detail, the recognition performance of the pattern that is difficult to identify with the low-resolution original functional MR image data is improved. The type of improvement in discrimination is shown below with reference to FIG.
(1) Another pattern, which is displayed in the same pattern at low resolution and cannot be classified, can be recognized as a different pattern.
(2) At low resolution, a pattern that has been recognized as a heterogeneous pattern due to subtle differences in imaging conditions and environments can be recognized as a common pattern by increasing the resolution.
(3) At low resolution, a similar pattern having a slight difference that must be recognized as the same pattern can be recognized as a different pattern.

[Application to pattern recognition means]
It is conceivable to construct a machine learning algorithm and a pattern analysis algorithm using super-resolution functional MR image data as input data, and to create a diagnosis support means including a learning means and a pattern recognition means using these algorithms. . As described above, according to the present invention, it is possible to express a fine pattern difference by super-resolution. As a result, the characteristics of pattern changes that have been difficult to grasp due to the minuteness can be clearly captured, so the recognition accuracy when performing pattern analysis by machine learning is improved and diagnostic accuracy is improved. Brought about. Further, since it becomes possible to grasp a slight difference in patterns, an increase in learning data is brought about, and an effect of improving the generalization performance of the pattern recognition means can be obtained.

[Application to neurofeedback]
The super-resolution technology of the present invention is used when applying neurofeedback training for training to control brain activity state to be appropriate while measuring brain activity state in real time. It is possible to apply. In order to perform the neurofeedback, first, the functional MRI apparatus 21 of the client facility 20 performs a functional localizer scan of the subject, and the subject's brain hippocampus or the training target for the recovery and improvement of the cognitive function. Identify the hippocampal gyrus as a region of interest. Next, the region of interest of the subject identified by the localizer scan is scanned in real time by the functional MRI apparatus 21, and the strength of the brain activity in the region of interest is presented to the subject as visual feedback. The visual feedback can be displayed by, for example, a circle that expands as the brain activity increases. The subject can perform independent training to activate the activity of the hippocampus or parahippocampus by being conscious of keeping the size of the displayed circle as large as possible.

  In the above-described neurofeedback training, the original functional MR image generated by the functional localization scan of the subject in the client facility 20 is converted into super-resolution functional MR image data by the server facility 10. By using this, it is possible to accurately identify the region of interest and to provide detailed and highly accurate feedback by super-resolving the real-time scan results of the state of brain activity. Become. As a result, the training effect is improved.

  The MR image super-resolution system according to the present invention can be used not only for diagnosis of brain structure and function but also for diagnosis of circulatory organs and digestive organs. In any case, even if only a low-resolution MRI apparatus is installed in the local client equipment, the super-resolution MR image data super-resolved by the server equipment 10 such as the core center is stored in the client. Since it can be used for diagnosis in facilities, it is possible to eliminate regional disparities in medical services.

The present invention can also be applied to image analysis by PET (Positron Emission Tomography). That is, by connecting a client facility equipped with a relatively low-resolution PET apparatus and a server facility equipped with a super-resolution means via a cloud network, the server facility can be operated from the original PET image data generated by the client facility. By constructing a super-resolution system for generating super-resolution PET image data by the resolution means, it becomes possible to eliminate regional disparities in PET image diagnosis.

S system N of the present invention N cloud network 10 server equipment 11 super-resolution means 12 data transmission / reception means 20 client equipment 21 MRI apparatus 22 data transmission / reception means

Claims (6)

Cloud network,
Two or more client facilities each connected to a cloud network so as to be able to transmit and receive data, and each having an MRI apparatus for generating MR image data;
A server facility that is connected to the cloud network so as to be able to transmit and receive data, and that includes super-resolution means for generating high-resolution MR image data by performing high-resolution processing of MR image data;
The server facility has a function of receiving the original MR image data generated by the client facility through the cloud network, and a function of generating super-resolution MR image data by the super-resolution means based on the received original MR image data. and,
The server facility further includes an MR image database in which standard image pattern data having a high resolution corresponding to the original MR image data generated by the client facility is stored in advance, and the super-resolution means includes the original MR image data and the MR image database. 3. The MR image super-resolution system according to claim 1, wherein the MR image data is generated based on comparison with corresponding standard image pattern data. Cloud network,
Two or more client facilities each connected to a cloud network so as to be able to transmit and receive data, and each having an MRI apparatus for generating MR image data;
A server facility that is connected to the cloud network so as to be able to transmit and receive data, and that includes super-resolution means for generating high-resolution MR image data by performing high-resolution processing of MR image data;
The server facility has a function of receiving the original MR image data generated by the client facility through the cloud network, and a function of generating super-resolution MR image data by the super-resolution means based on the received original MR image data. and,
The server facility further includes an individual reference MR image database in which individual MR image pattern data captured for each individual at a higher resolution than the original MR image data generated by the client facility is stored.
The super-resolution means generates super-resolution MR image data based on a comparison between the original MR image data generated by the client equipment and the corresponding individual MR image pattern data. Or the MR image super-resolution system according to 2; Cloud network,
Two or more client facilities each connected to a cloud network so as to be able to transmit and receive data, and each having an MRI apparatus for generating MR image data;
A server facility that is connected to the cloud network so as to be able to transmit and receive data, and that includes super-resolution means for generating high-resolution MR image data by performing high-resolution processing of MR image data;
The server facility has a function of receiving the original MR image data generated by the client facility through the cloud network, and a function of generating super-resolution MR image data by the super-resolution means based on the received original MR image data. and,
The MRI apparatus of the client facility further has a function of generating a plurality of original MR image data whose imaging positions are shifted by a predetermined distance,
The server facility super-resolution means is generated by the MRI apparatus of the client facility, the data of the means for generating a plurality of image data having a positional deviation such as the deviation amount data of the imaging position, and the MRI apparatus of the client facility. A super-resolution system for MR images , wherein one super-resolution MR image data for each client is generated based on the plurality of original MR image data. Server equipment, the super-resolution system in MR images as claimed in any one of claims 1-3, further having a function of transmitting the generated super-resolution MR image data one or more of the client equipment. The MRI apparatus of the client facility is a functional MRI apparatus that generates original functional MR image data reflecting the brain activity state,
Super-resolution means of the server equipment is based on the original functional MR image data is functional MRI device of the client equipment to produce, according to claim 1 4, further comprising a function of generating a super-resolution functional MR image data The super-resolution system for MR image data according to any one of the above. The MRI apparatus of the client facility further has a function of generating original functional MR image data continuously or intermittently,
The super-resolution unit of the server facility generates super-resolution functional MR image data continuously or intermittently based on the original functional MR image data generated continuously or intermittently by the MRI apparatus of the client facility. The MR image data super-resolution system according to claim 5 , further comprising a function.

Images