Deep learning-based medical image processing method, equipment and medium
A technology of medical images and processing methods, applied in the field of medical image processing based on deep learning, to achieve the effect of reducing burden, saving costs and reducing harm
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0062] This embodiment provides a medical image processing method based on deep learning, such as figure 1 As shown, the method specifically includes the following steps:
[0063]S11, acquiring an initial medical image.
[0064] In this embodiment, the initial medical image refers to an image scanned according to medical imaging technology, that is, a medical image obtained by scanning a scanning part through a medical imaging device, such as an MR image (such as a DWI image), a CT image, a T2W image, etc. Wait. Wherein, the scanning site may include tissues, organs, etc. of a living body.
[0065] S12. Acquire a region of interest that affects diagnosis in the initial medical image.
[0066] In this embodiment, a pre-trained multi-task learning model may be used to acquire the region of interest affecting diagnosis in the initial medical image. Wherein, the region of interest that affects the diagnosis may be an artifact or noise region in the image, or other regions in t...
Embodiment 2
[0110] In the actual medical imaging process, different hospitals may set different scanning parameters for different parts or different diseases. For example, for DWI images, the setting of B value is particularly important, and a reasonable setting of B value can improve the ability of lesion detection. However, for specific patients, specific parts and specific diseases, the setting of B value is not unique, and doctors with many years of experience in film reading experience are required to give good advice, otherwise it may be necessary to repeat filming or multiple filmings to obtain the desired DWI image.
[0111] In this regard, this embodiment is further improved on the basis of embodiment 1. Specifically, when the target scanning parameters input by the user (referred to as the first target scanning parameters in this embodiment) are received before step S13, this embodiment may optimize the region of interest in step S13, and may also The first target scan paramete...
Embodiment 3
[0125] This embodiment is a further improvement on Embodiment 1. Such as Figure 7 As shown, in this embodiment, when the multi-task learning model judges that the initial medical image does not affect the diagnosis, that is, it does not need to be optimized, and receives the target scan parameters input by the user (this embodiment is marked as the second target scan parameters), the medical image processing method may further include: S14, directly converting the scan parameters of the initial medical image according to the input second target scan parameters.
[0126] Such as Figure 8 As shown, the specific implementation process of step S14 is as follows:
[0127] S141. Acquire first image features based on the initial medical image.
[0128] In this embodiment, the high-dimensional features of the initial medical image may also be extracted through the encoding module of the pre-trained identity mapping model, which is recorded as the first image feature.
[0129] S1...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More