Image registration method and system
An image registration method is read by a processing device to perform: obtaining a first medical image and a second medical image generated by different imaging devices, with the first medical image including soft and hard tissue image; segmenting the first medical image and the second medical image to obtain a first hard tissue image and a second hard tissue image, respectively; aligning a coordinate axis of the first hard tissue image and a coordinate axis of the second hard tissue image, and obtaining a registration field indicating a corresponding relationship between the first hard tissue image and the second hard tissue image; obtaining a scale ratio between the first hard tissue image and the second hard tissue image according to the registration field; and generating a target soft and hard tissue image according to the scale ratio, the soft and hard tissue image and the second hard tissue image.
1 . An image registration method, read by a processing device to perform:
obtaining a first medical image and a second medical image, wherein the first medical image and the second medical image are generated by different imaging devices, and the first medical image comprises soft and hard tissue image;
segmenting the first medical image and the second medical image to obtain a first hard tissue image and a second hard tissue image, respectively;
aligning a coordinate axis of the first hard tissue image and a coordinate axis of the second hard tissue image, and obtaining a registration field indicating a corresponding relationship between the first hard tissue image and the second hard tissue image;
obtaining a scale ratio between the first hard tissue image and the second hard tissue image according to the registration field; and
generating a target soft and hard tissue image according to the scale ratio, the soft and hard tissue image and the second hard tissue image.
2 . The image registration method according to claim 1 , wherein the first hard tissue image is a plurality of first two dimensional images, the second hard tissue image is a plurality of second two dimensional images, and aligning the coordinate axis of the first hard tissue image and the coordinate axis of the second hard tissue image comprises:
performing an interpolation reconstruction calculation on the plurality of first two dimensional images to generate a first three dimensional image;
performing the interpolation reconstruction calculation on the plurality of second two dimensional images to generate a second three dimensional image; and
using a coordinate axis of the first three dimensional image as the coordinate axis of the first hard tissue image and using a coordinate axis of the second three dimensional image as the coordinate axis of the second hard tissue image.
3 . The image registration method according to claim 1 , wherein obtaining the registration field indicating the corresponding relationship between the first hard tissue image and the second hard tissue image comprises:
inputting the first hard tissue image and the second hard tissue image into a registration model to obtain a third hard tissue image and the registration field.
4 . The image registration method according to claim 3 , wherein obtaining the scale ratio between the first hard tissue image and the second hard tissue image according to the registration field comprises:
obtaining first displacement information between two different hard tissue blocks of a same piece of bone in the third hard tissue image;
obtaining second displacement information corresponding to the two different hard tissue blocks in the first hard tissue image from the registration field; and
performing an interpolation calculation on the first displacement information and the second displacement information to obtain the scale ratio.
5 . The image registration method according to claim 3 , further comprising, by the processing device or another processing device, performing:
performing training by using a plurality of first training images and a plurality of second training images to obtain the registration model,
wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image, and
wherein the plurality of first training images correspond to one of a supine posture and a prone posture, and the plurality of second training images correspond to one of the supine posture and the prone posture.
6 . The image registration method according to claim 1 , wherein generating the target soft and hard tissue image according to the scale ratio, the soft and hard tissue image and the second hard tissue image comprises:
adjusting the soft and hard tissue image according to the scale ratio to generate an intermediate image; and
inputting the second hard tissue image and the intermediate image into a soft and hard tissue generation model to generate the target soft and hard tissue image.
7 . The image registration method according to claim 6 , wherein inputting the second hard tissue image and the intermediate image into the soft and hard tissue generation model to generate the target soft and hard tissue image comprises:
converting the intermediate image into a first latent space;
converting the second hard tissue image into a second latent space; and
inputting latent space information of the first latent space and latent space information of the second latent space into the soft and hard tissue generation model.
8 . The image registration method according to claim 6 , further comprising, by the processing device or another processing device, performing:
performing training by using a plurality of first training images and a plurality of second training images to obtain the soft and hard tissue generation model,
wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image, and
wherein the plurality of first training images correspond to one of a supine posture and a prone posture, and the plurality of second training images correspond to one of the supine posture and the prone posture.
9 . The image registration method according to claim 1 , wherein segmenting the first medical image and the second medical image to obtain the first hard tissue image and the second hard tissue image, respectively comprises:
inputting the first medical image into a first segmentation model to obtain the first hard tissue image; and
inputting the second medical image into a second segmentation model to obtain the second hard tissue image.
10 . The image registration method according to claim 9 , further comprising, by the processing device or another processing device, performing:
labeling a plurality of first hard tissue labels in a plurality of first training images, and labeling a plurality of second hard tissue labels in a plurality of second training images, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image; and
training a first initial model by using the plurality of labeled first training images to generate the first segmentation model, and training a second initial model by using the plurality of labeled second training images to generate the second segmentation model.
11 . An image registration system, comprising:
an imaging device configured to generate a first medical image, wherein the first medical image comprises soft and hard tissue image; and
a processing device connected to the imaging device, and configured to perform:
obtaining the first medical image and a second medical image, wherein the second medical image is generated by another imaging device;
segmenting the first medical image and the second medical image to obtain a first hard tissue image and a second hard tissue image, respectively;
aligning a coordinate axis of the first hard tissue image and a coordinate axis of the second hard tissue image, and obtaining a registration field indicating a corresponding relationship between the first hard tissue image and the second hard tissue image;
obtaining a scale ratio between the first hard tissue image and the second hard tissue image according to the registration field; and
generating a target soft and hard tissue image according to the scale ratio, the soft and hard tissue image and the second hard tissue image.
12 . The image registration system according to claim 11 , wherein the first hard tissue image is a plurality of first two dimensional images, the second hard tissue image is a plurality of second two dimensional images, and the processing device is configured to perform an interpolation reconstruction calculation on the plurality of first two dimensional images to generate a first three dimensional image; perform the interpolation reconstruction calculation on the plurality of second two dimensional images to generate a second three dimensional image; and use a coordinate axis of the first three dimensional image as the coordinate axis of the first hard tissue image and using a coordinate axis of the second three dimensional image as the coordinate axis of the second hard tissue image.
13 . The image registration system according to claim 11 , wherein the processing device is configured to input the first hard tissue image and the second hard tissue image into a registration model to obtain a third hard tissue image and the registration field.
14 . The image registration system according to claim 13 , wherein the processing device is configured to obtain first displacement information between two different hard tissue blocks of a same piece of bone in the third hard tissue image; obtain second displacement information corresponding to the two different hard tissue blocks in the first hard tissue image from the registration field; and perform an interpolation calculation on the first displacement information and the second displacement information to obtain the scale ratio.
15 . The image registration system according to claim 13 , wherein the processing device is further configured to perform training by using a plurality of first training images and a plurality of second training images to obtain the registration model, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image, and wherein the plurality of first training images correspond to one of a supine posture and a prone posture, and the plurality of second training images correspond to one of the supine posture and the prone posture.
16 . The image registration system according to claim 13 , further comprising:
another processing device connected to the processing device, wherein the another processing device is configured to perform training by using a plurality of first training images and a plurality of second training images to obtain the registration model, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image, and wherein the plurality of first training images correspond to one of a supine posture and a prone posture, and the plurality of second training images correspond to one of the supine posture and the prone posture.
17 . The image registration system according to claim 11 , wherein the processing device is further configured to adjust the soft and hard tissue image according to the scale ratio to generate an intermediate image; and input the second hard tissue image and the intermediate image into a soft and hard tissue generation model to generate the target soft and hard tissue image.
18 . The image registration system according to claim 17 , wherein the processing device is configured to convert the intermediate image into a first latent space; convert the second hard tissue image into a second latent space; and input latent space information of the first latent space and latent space information of the second latent space into the soft and hard tissue generation model.
19 . The image registration system according to claim 17 , wherein the processing device is further configured to perform training by using a plurality of first training images and a plurality of second training images to obtain the soft and hard tissue generation model, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image, and wherein the plurality of first training images correspond to one of a supine posture and a prone posture, and the plurality of second training images correspond to one of the supine posture and the prone posture.
20 . The image registration system according to claim 17 , further comprising:
another processing device connected to the processing device, wherein the another processing device is configured to perform training by using a plurality of first training images and a plurality of second training images to obtain the soft and hard tissue generation model, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image, and wherein the plurality of first training images correspond to one of a supine posture and a prone posture, and the plurality of second training images correspond to one of the supine posture and the prone posture.
21 . The image registration system according to claim 11 , wherein the processing device is configured to input the first medical image into a first segmentation model to obtain the first hard tissue image; and input the second medical image into a second segmentation model to obtain the second hard tissue image.
22 . The image registration system according to claim 21 , further comprising:
another processing device connected to the processing device, wherein the another processing device is configured to perform:
labeling a plurality of first hard tissue labels in a plurality of first training images, and labeling a plurality of second hard tissue labels in a plurality of second training images, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image; and
training a first initial model by using the plurality of labeled first training images to generate the first segmentation model, and training a second initial model by using the plurality of labeled second training images to generate the second segmentation model.
23 . The image registration system according to claim 21 , wherein the processing device is further configured to perform:
labeling a plurality of first hard tissue labels in a plurality of first training images, and labeling a plurality of second hard tissue labels in a plurality of second training images, wherein the plurality of first training images are generated by a same imaging device as the first medical image, and the plurality of second training images are generated by a same imaging device as the second medical image; and
training a first initial model by using the plurality of labeled first training images to generate the first segmentation model, and training a second initial model by using the plurality of labeled second training images to generate the second segmentation model.