Methods and systems for medical data processing

1 . A method for generating a medical data processing network, wherein the method is performed by a medical imaging system including a medical imaging device, a processing device, a storage device, and one or more terminals, and the method comprises:

in response to receiving a training instruction, obtaining, by the processing device, first medical data in real time in case of changes in a clinical application environment, the training instruction being triggered by a user on an application interface of the processing device or the one or more terminals of the medical imaging system;

obtaining, by the processing device, a second network by training a first network based on the first medical data, wherein the first network and the second network are machine learning models, the training includes an unsupervised training, and the training the first network based on the first medical data includes:

training, by the processing device, the first network at a preset training frequency, wherein the preset training frequency is correlated to a daily scanning volume of the medical imaging device;

generating, by the processing device, a test result by inputting second medical data into the second network, the second medical data including a test sample set;

determining, by the processing device, an evaluation result of a performance of the second network based on the test result; and

in response to the evaluation result indicating that the performance of the second network satisfies a requirement, storing, by the processing device, a parameter of the second network in the storage device.

2 . The method of claim 1 , wherein the obtaining a second network by training a first network based on the first medical data includes:

automatically generating training data and a label based on the first medical data; and

obtaining the second network by training the first network based on the training data and the label.

3 . The method of claim 1 , wherein the first network is used for a reconstruction of the first medical data, the first network is implemented based on a diffusion model, and the training the first network based on the first medical data includes:

generating an intermediate image by reconstructing the first medical data using a preset algorithm; and

training the first network based on the intermediate image.

4 . The method of claim 3 , wherein the training the first network based on the intermediate image includes:

training, by adding image quality interference information to the intermediate image, the first network based on the intermediate image.

5 . The method of claim 4 , wherein the training, by adding the image quality interference information to the intermediate image, the first network based on the intermediate image includes:

iteratively updating the first network until a preset termination condition is satisfied, wherein each round of the iterative update includes:

generating a noise sample;

generating a sample intermediate image by adding the noise sample to the intermediate image;

obtaining a noise prediction value based on the sample intermediate image through the first network;

determining a first parameter based on the noise sample and the noise prediction value;

determining a second parameter based on the intermediate image; and

updating the parameter of the first network based on the first parameter and the second parameter.

6 . The method of claim 5 , wherein the generating a noise sample includes selecting a noise distribution from a noise distribution set as the noise sample, wherein

the noise distribution is a Gaussian distribution or a Poisson distribution, and

the noise distribution set is determined by:

determining a generation parameter of the noise distribution set based on a function and a network complexity of the first network, the generation parameter at least including a size of the noise distribution set; and

forming the noise distribution set by obtaining a certain count of noise distributions according to the generation parameter.

7 . The method of claim 4 , wherein the training, by adding the image quality interference information to the intermediate image, the first network based on the intermediate image includes:

obtaining a denoised image by performing a denoising on the intermediate image; and

training the first network based on the denoised image.

8 . The method of claim 1 , wherein the method further includes:

presenting the test result on the application interface.

9 . The method of claim 8 , wherein

in response to receiving a confirmation instruction, determining that the evaluation result indicates that the performance of the second network satisfies the requirement, the confirmation instruction being triggered by the user on the application interface; and

in response to receiving a re-training instruction, determining that the evaluation result indicates that the performance of the second network does not satisfy the requirement, the re-training instruction being triggered by the user on the application interface.

10 . The method of claim 1 , wherein the method further includes:

in response to that the evaluation result indicates that the performance of the second network does not satisfy the requirement, re-obtaining new first medical data;

obtaining a third network based on the new first medical data by training the second network; and

generating a test result by inputting the second medical data into the third network; and

determining an evaluation result of a performance of the third network based on the test result of the third network until the evaluation result of the third network indicates that the performance of the third network satisfies the requirement.

11 . The method of claim 1 , wherein the obtaining first medical data includes:

obtaining initial medical data;

determining whether the initial medical data satisfies a preset training condition, the preset training condition including at least one of whether a count of the initial medical data reaches a preset count threshold, whether the initial medical data is accurate, or whether a size of the initial medical data satisfies a preset size threshold;

in response to that the initial medical data satisfies the preset training condition, designating the initial medical data as the first medical data; and

in response to that the initial medical data does not satisfy the preset training condition, obtaining new initial medical data and repeatedly determining whether the new initial medical data satisfies the preset training condition until the new initial medical data that satisfies the preset training condition is obtained.

12 . The method of claim 11 , wherein the preset count threshold is correlated with at least one of function of the first network or a training progress of the first network.

13 . The method of claim 1 , wherein the storing a parameter of the second network includes:

obtaining a labeled parameter by labeling the parameter of the second network, the labeling being related to at least one of a function and a type of the second network; and

storing the labeled parameter.

14 . The method of claim 1 , further comprising:

syncing or sharing the preset training frequency across a plurality of medical centers associated with the medical data processing network to make training frequencies of the plurality of medical centers identical.

15 . The method of claim 14 , wherein the method is performed during a spare time when the medical imaging device is not performing a scan.

16 . The method of claim 15 , wherein the preset training frequency is determined by using a vector database based on imaging data of the medical imaging device, and the vector database includes reference vectors of the imaging data corresponding to a plurality of daily scanning volumes and reference training frequencies corresponding to the reference vectors.

17 . The method of claim 16 , wherein the preset training frequency is determined by:

constructing, by the processing device, a target feature vector based on the daily scanning volume of the medical imaging device;

matching, by the processing device, at least one reference vector in the vector database that satisfies a preset condition based on the target feature vector, wherein the preset condition includes a distance between the target feature vector and each of the at least one reference vector is less than a distance threshold; and

taking, by the processing device, an average value of reference training frequencies corresponding to the at least one reference vector as the preset training frequency.

18 . A medical imaging system including a medical imaging device, a processing device, a storage device, one or more terminals, and a network; wherein the processing device is configured to:

in response to receiving a training instruction, obtain first medical data in real time in case of changes in a clinical application environment, the training instruction being triggered by a user on an application interface of the processing device or the one or more terminals of the medical imaging system;

obtain a second network by training a first network based on the first medical data, wherein the first network and the second network are machine learning models, the training includes an unsupervised training, and the training the first network based on the first medical data includes:

training the first network at a preset training frequency, wherein the preset training frequency is correlated to a daily scanning volume of the medical imaging device;

generate a test result by inputting second medical data into the second network, the second medical data including a test sample set;

determine an evaluation result of a performance of the second network based on the test result; and

in response to the evaluation result indicating that the performance of the second network satisfies a requirement, store a parameter of the second network in the storage device.

19 . A method for processing medical data, wherein the method is performed on a medical imaging system including a medical imaging device, a processing device, a storage device, and one or more terminals, and the method comprises:

obtaining, by the processing device, medical data to be processed from the medical imaging device;

in response to receiving a selection instruction, obtaining, by the processing device, a basic network model and a target parameter, wherein the selection instruction is triggered by a user on an application interface of the processing device or the one or more terminals, and the selection instruction is configured to select the target parameter from a plurality of parameters stored in the storage device and the basic network model from a plurality of network models stored in the storage device;

generating a target network model based on the target parameter and the basic network model, wherein the target network model and the basic network model are machine learning models; and

obtaining a processing result by inputting the medical data to be processed into the target network model, wherein the processing result includes a processed medical data.

20 . The method of claim 19 , wherein the plurality of parameters are generated in the medical imaging system by:

in response to receiving a training instruction, obtaining, by the processing device, first medical data in real time in case of changes in a clinical application environment, the training instruction being triggered by the user on the application interface of the processing device or the one or more terminals of the medical imaging system;

obtaining, by the processing device, a second network by training a first network based on the first medical data, wherein the training includes an unsupervised training, and the training the first network based on the first medical data includes:

training the first network at a preset training frequency, wherein the preset training frequency is correlated to a daily scanning volume of the medical imaging device;

generating, by the processing device, a test result by inputting second medical data into the second network, the second medical data including a test sample set;

determining, by the processing device, an evaluation result of a performance of the second network based on the test result; and

in response to the evaluation result indicating that the performance of the second network satisfies a requirement, designating, by the processing device, a parameter of the second network as one of the plurality of parameters.