In a possible implementation, after obtaining the ADC image to be corrected, the user's terminal can input the ADC image to be corrected into the pre-constructed image correction model to correct the grayscale value of tissue position of the ADC image to be corrected and obtain a target image corresponding to the ADC image to be corrected.

S103: outputting a target image corresponding to the ADC image to be corrected.

In the embodiment of the present application, after obtaining the target image corresponding to the ADC image to be corrected, the user's terminal can output the target image corresponding to the ADC image to be corrected.

In the embodiment of the present application, the device for correcting magnetic resonance image first acquires an ADC image to be corrected, which is generated by scanning a target object using an echo planar imaging sequence from a pulse sequence of a portable mobile magnetic resonance apparatus and converting; then the ADC image to be corrected is input into a pre-constructed image correction model to correct a grayscale value of tissue position of the ADC image to be corrected, where the pre-constructed image correction model is generated by function fitting based on magnetic resonance image sequences of different objects; finally, a target image corresponding to the ADC image to be corrected is output. Since the pre-constructed image correction model is generated by function fitting based on magnetic resonance image sequences of different objects in the present application, the constant parameter of the model has a high credibility, so that the problem of low parameter credibility caused by correction based on artificial experience is avoided, thus improving the parameter accuracy of portable mobile magnetic resonance ADC images.

Reference is made to FIG. 2, which is a schematic flowchart of constructing an image correction model provided by an embodiment of the present application. As shown in FIG. 2, the method of the embodiment of the present application may include the following steps S201 to S204.

S201: acquiring a magnetic resonance image sequence of a first object, where the magnetic resonance image sequence includes a first traditional magnetic resonance ADC image, a first mobile magnetic resonance ADC image and a first T1 image, and the image sequence of the first object is the image sequence of any object in an image dataset;

- where the first traditional magnetic resonance ADC image is generated through a traditional fixed magnetic resonance apparatus, the first mobile magnetic resonance ADC image is generated through a portable mobile magnetic resonance apparatus, and the first T1 image is generated by scanning according to a target sequence from the pulse sequence of the traditional fixed magnetic resonance apparatus and converting.

In the embodiment of the present application, when obtaining the image dataset, a single traditional magnetic resonance ADC image, a single mobile magnetic resonance ADC image and a single T1 image of each of n target objects are first collected to construct an image dataset D, where the optimal value of n is 100, but n is not limited to this value.

In a possible implementation, the user's terminal first acquires the image sequence of any object in the image dataset to obtain the magnetic resonance image sequence of the first object.

S202: using the first mobile magnetic resonance ADC image of the first object as an image template to perform image registration on the first traditional magnetic resonance ADC image and the first T1 image respectively so as to obtain a second traditional magnetic resonance ADC image and a second T1 image of the first object;

- where the image registration is a process of matching and overlaying two or more images acquired at different times with different imaging devices or under different conditions.

In the embodiment of the present application, when performing image registration, the first mobile magnetic resonance ADC image of the first object is used as the image template; then first feature points in the first traditional magnetic resonance ADC image and the image template are extracted, and the first feature points are matched to obtain first matching feature points of the image template and the first traditional magnetic resonance ADC image; next, second feature points in the first T1 image and the image template are extracted, and the second feature points are matched to obtain second matching feature points in the image template and the first T1 image; then a second traditional magnetic resonance ADC image of the first object is generated based on the first matching feature points; and finally, a second T1 image of the first object is generated based on the second matching feature points.

Specifically, when generating the second traditional magnetic resonance ADC image of the first object based on the first matching feature points, the first matching feature points are used as a constraint condition to deform the first traditional magnetic resonance ADC image so that the position of the deformed image is consistent with the position of the matching feature points of the image template, thus obtaining the second traditional magnetic resonance ADC image of the first object.

Specifically, when generating the second T1 image of the first object based on the second matching feature points, the second matching feature points are used as a constraint condition to deform the first T1 image so that the position of the deformed image is consistent with the position of the matching feature points of the image template, thus obtaining the second T1 image of the first object.

It should be noted that the optimal setting for the template image is the mobile magnetic resonance ADC image, but it is not limited to this image. The optimal registration method is rigid registration, but it is not limited to this method.

S203: segmenting and labelling structural tissue of the second T1 image of the first object to obtain an anatomical structure labelling map of the first object;

- in the embodiment of the present application, the structural tissue of the second T1 image of the first object is segmented and labeled, where the optimal method for tissue segmentation is the kmeans segmentation method, and the optimal number N_classof anatomical structure segmentation classes is 3 (including cerebrospinal fluid, cerebra gray matter, and cerebra white matter); the anatomical structure labelling map L is obtained (in the anatomical structure labelling map L, the optimal labelling can be set as: background being labelled as 0, the cerebrospinal fluid being labelled as 1, the cerebra gray matter being labelled as 2, and the cerebra white matter being labelled as 3); after the labelling is completed, the anatomical structure labelling map of the first object is obtained.

S204: obtaining a pre-constructed image correction model based on the second traditional magnetic resonance ADC image, the first mobile magnetic resonance ADC image and the anatomical structure labelling map of the first object.

In the embodiment of the present application, when obtaining the pre-constructed image correction model based on the second traditional magnetic resonance ADC image, the first mobile magnetic resonance ADC image and the anatomical structure labelling map of the first object, a target parameter correction function predefined for the ADC image scanned and converted by the portable mobile magnetic resonance apparatus is loaded first; then image parameters of the second traditional magnetic resonance ADC image of the first object are used as a first variable of the target parameter correction function, image parameters of the first mobile magnetic resonance ADC image are used as a second variable of the target parameter correction function, and image parameters of the anatomical structure labelling map are used as a third variable of the target parameter correction function to obtain an image parameter expression of the first object; where the first variable is a corrected pADC image IC_pADC, the second variable is a pADC image to be corrected I_pADC, and the third variable is an anatomical structure segmentation class il; next, function fitting is performed based on the image parameter expression of the first object and the least square method to obtain an image correction function; and finally, modeling is performed based on the image correction function to obtain a pre-constructed image correction model.

Specifically, the target parameter correction function is:

f
⁡
(
x
)

=

∑

i
=
0

N
f

⁢

a
i

⁢

x
i

=

F
⁡
(

I
pADC

(

il
,
ix
,
iy

)

)

;

IC
pADC

(

il
,
ix
,
iy

)

=

F
⁡
(

I
pADC

(

il
,
ix
,
iy

)

)

,

il
≤

N
class

,

0
≤
ix
≤
W

,

0
≤
iy
≤
H

;

where IC_pADCis the corrected pADC image, I_pADCis the pADC image to be corrected, and il is the anatomical structure segmentation class; ix and iy are horizontal and vertical coordinates of each pixel position of the target tissue in the image respectively, N_classis the number of anatomical structure segmentation class, and W, H are the width and height of the image respectively; x is the independent variable, i is the order of the polynomial function, a; is the coefficient of each order, and N_fis the number of orders of the polynomial function.

For example, after the target parameter correction function is obtained, the image parameters of the second traditional magnetic resonance ADC image of the first object are first used as the corrected pADC image IC_pADCin this function; then, the image parameters of the first mobile magnetic resonance ADC image are used as the pADC image to be corrected I_pADCin this function; next, the image parameters of the anatomical structure labelling map L are used as the structure segmentation class in this function; and finally, the least square method is used to fit the function f(x) to solve for the constant parameter a_iin this function to finally obtain the fitted correction function F_fit(x).

After the correction function F_fit(x) is obtained, the user's terminal can correct any ADC image to be corrected in combination with the correction function I_fit(x), so as to obtain the corrected image IC_pADC.

In the embodiment of the present application, the device for correcting magnetic resonance image first acquires an ADC image to be corrected, which is generated by scanning a target object using an echo planar imaging sequence from a pulse sequence of a portable mobile magnetic resonance apparatus and converting; then the ADC image to be corrected is input into a pre-constructed image correction model to correct a grayscale value of tissue position of the ADC image to be corrected, where the pre-constructed image correction model is generated by function fitting based on magnetic resonance image sequences of different objects; finally, a target image corresponding to the ADC image to be corrected is output. Since the pre-constructed image correction model is generated by function fitting based on magnetic resonance image sequences of different objects in the present application, the constant parameter of the model has a high credibility, so that the problem of low parameter credibility caused by correction based on artificial experience is avoided, thus improving the parameter accuracy of portable mobile magnetic resonance ADC images.

The following is a device embodiment of the present disclosure, which can be used to carry out the method embodiment of the present disclosure. For details not disclosed in the device embodiment of the present disclosure, reference may be made to the method embodiment of the present disclosure.

Reference is made to FIG. 3, which is a schematic diagram of the structure of a device for correcting magnetic resonance image provided by an illustrative embodiment of the present disclosure. The device for correcting magnetic resonance image can be realized as the entirety or part of a terminal through software, hardware, or a combination of both software and hardware. The device 1 includes an image acquisition module 10, an image parameter correction module 20, and an image output module 30.

The image acquisition module 10 is configured to acquire an ADC image to be corrected, which is generated by scanning using an echo planar imaging sequence from a pulse sequence of a portable mobile magnetic resonance apparatus and converting;

- the image parameter correction module 20 is configured to input the ADC image to be corrected into a pre-constructed image correction model to correct a grayscale value of tissue position of the ADC image to be corrected; where the pre-constructed image correction model is generated by function fitting based on magnetic resonance image sequences of different objects; and