Method and device for identifying abnormal cell in to-be-detected sample, and storage medium
Methods, apparatus, device, and storage medium for identifying an abnormal cell in a to-be-detected sample are disclosed. The method includes obtaining, by a device, multi-layer images of a to-be-detected sample, the to-be-detected sample comprising a single cell and a cell cluster; obtaining, by the device, multi-layer image blocks of the single cell and multi-layer image blocks of the cell cluster according to the multi-layer images; obtaining, by the device, a first identification result by a first image identification network according to the multi-layer image blocks of the single cell; obtaining, by the device, a second identification result by a second image identification network according to the multi-layer image blocks of the cell cluster; and determining, by the device, whether an abnormal cell exists in the to-be-detected sample according to the first identification result and the second identification result.
1. A method for identifying an abnormal cell in a to-be-detected sample, the method comprising:
obtaining, by a device comprising a memory storing instructions and a processor in communication with the memory, multi-layer images of a to-be-detected sample, the to-be-detected sample comprising a single cell and a cell cluster;
obtaining, by the device, multi-layer image blocks of the single cell and multi-layer image blocks of the cell cluster according to the multi-layer images;
obtaining, by the device, a first identification result by a first image identification network according to the multi-layer image blocks of the single cell;
obtaining, by the device, a second identification result by a second image identification network according to the multi-layer image blocks of the cell cluster, including:
selecting at least one layer of image blocks from the multi-layer image blocks of the cell cluster, and inputting the at least one layer of image blocks into a first part of the second image identification network to obtain a second single-layer identification result corresponding to the multi-layer image blocks of the cell cluster; and
inputting the multi-layer image blocks of the cell cluster into a second part of the second image identification network to obtain a second multi-layer identification result corresponding to the multi-layer image blocks of the cell cluster; and
determining, by the device, whether an abnormal cell exists in the to-be-detected sample according to the first identification result and the second identification result.
2. The method according to claim 1 , wherein the obtaining the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster according to the multi-layer images comprises:
dividing the multi-layer images into a plurality of multi-layer image blocks; and
classifying the plurality of multi-layer image blocks into the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster.
3. The method according to claim 2 , wherein the classifying the plurality of multi-layer image blocks into the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster comprises:
performing the classification by determining a quantity of cell nuclei in each of the plurality of multi-layer image blocks.
4. The method according to claim 1 , wherein the obtaining the first identification result by the first image identification network according to the multi-layer image blocks of the single cell comprises:
selecting at least one layer of image blocks from the multi-layer image blocks of the single cell, and inputting the at least one layer of image blocks into a first part of the first image identification network to obtain a first single-layer identification result corresponding to the multi-layer image blocks of the single cell;
inputting the multi-layer image blocks of the single cell into a second part of the first image identification network to obtain a first multi-layer identification result corresponding to the multi-layer image blocks of the single cell; and
fusing the first single-layer identification result and the first multi-layer identification result to obtain the first identification result, wherein:
the first part of the first image identification network is used for extracting plane information of the single cell, and the second part of the first image identification network is used for extracting three-dimensional structure information of the single cell.
5. The method according to claim 4 , wherein the fusing the first single-layer identification result and the first multi-layer identification result to obtain the first identification result comprises:
performing weighted averaging on the first single-layer identification result and the first multi-layer identification result for a plurality of cell types of the single cell, to obtain the first identification result.
6. The method according to claim 1 , wherein the obtaining the second identification result by the second image identification network according to the multi-layer image blocks of the cell cluster further comprises:
fusing the second single-layer identification result and the second multi-layer identification result to obtain the second identification result, wherein:
the first part of the second image identification network is used for extracting plane information of the cell cluster, and the second part of the second image identification network is used for extracting three-dimensional structure information of the cell cluster.
7. The method according to claim 1 , wherein:
the first image identification network is trained based on a first data set, and the first data set comprises multi-layer images of the single cell and labels of cell types of the multi-layer images of the single cell; and
the second image identification network is trained based on a second data set, and the second data set comprises multi-layer images of the cell cluster and labels of cell types of the multi-layer images of the cell cluster.
8. An apparatus for identifying an abnormal cell in a to-be-detected sample, the apparatus comprising:
a memory storing instructions; and
a processor in communication with the memory, wherein, when the processor executes the instructions, the processor is configured to cause the apparatus to perform:
obtaining multi-layer images of a to-be-detected sample, the to-be-detected sample comprising a single cell and a cell cluster,
obtaining multi-layer image blocks of the single cell and multi-layer image blocks of the cell cluster according to the multi-layer images,
obtaining a first identification result by a first image identification network according to the multi-layer image blocks of the single cell,
obtaining a second identification result by a second image identification network according to the multi-layer image blocks of the cell cluster, including:
selecting at least one layer of image blocks from the multi-layer image blocks of the cell cluster, and inputting the at least one layer of image blocks into a first part of the second image identification network to obtain a second single-layer identification result corresponding to the multi-layer image blocks of the cell cluster; and
inputting the multi-layer image blocks of the cell cluster into a second part of the second image identification network to obtain a second multi-layer identification result corresponding to the multi-layer image blocks of the cell cluster, and
determining whether an abnormal cell exists in the to-be-detected sample according to the first identification result and the second identification result.
9. The apparatus according to claim 8 , wherein the obtaining the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster according to the multi-layer images comprises:
dividing the multi-layer images into a plurality of multi-layer image blocks; and
classifying the plurality of multi-layer image blocks into the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster.
10. The apparatus according to claim 9 , wherein the classifying the plurality of multi-layer image blocks into the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster comprises:
performing the classification by determining a quantity of cell nuclei in each of the plurality of multi-layer image blocks.
11. The apparatus according to claim 8 , wherein the obtaining the first identification result by the first image identification network according to the multi-layer image blocks of the single cell comprises:
selecting at least one layer of image blocks from the multi-layer image blocks of the single cell, and inputting the at least one layer of image blocks into a first part of the first image identification network to obtain a first single-layer identification result corresponding to the multi-layer image blocks of the single cell;
inputting the multi-layer image blocks of the single cell into a second part of the first image identification network to obtain a first multi-layer identification result corresponding to the multi-layer image blocks of the single cell; and
fusing the first single-layer identification result and the first multi-layer identification result to obtain the first identification result, wherein:
the first part of the first image identification network is used for extracting plane information of the single cell, and the second part of the first image identification network is used for extracting three-dimensional structure information of the single cell.
12. The apparatus according to claim 11 , wherein the fusing the first single-layer identification result and the first multi-layer identification result to obtain the first identification result comprises:
performing weighted averaging on the first single-layer identification result and the first multi-layer identification result for a plurality of cell types of the single cell, to obtain the first identification result.
13. The apparatus according to claim 8 , wherein the obtaining the second identification result by the second image identification network according to the multi-layer image blocks of the cell cluster further comprises:
fusing the second single-layer identification result and the second multi-layer identification result to obtain the second identification result, wherein:
the first part of the second image identification network is used for extracting plane information of the cell cluster, and the second part of the second image identification network is used for extracting three-dimensional structure information of the cell cluster.
14. The apparatus according to claim 8 , wherein:
the first image identification network is trained based on a first data set, and the first data set comprises multi-layer images of the single cell and labels of cell types of the multi-layer images of the single cell; and
the second image identification network is trained based on a second data set, and the second data set comprises multi-layer images of the cell cluster and labels of cell types of the multi-layer images of the cell cluster.
15. A non-transitory computer-readable storage medium, storing computer-readable instructions, wherein, the computer-readable instructions, when executed by a processor, are configured to cause the processor to perform:
obtaining multi-layer images of a to-be-detected sample, the to-be-detected sample comprising a single cell and a cell cluster;
obtaining multi-layer image blocks of the single cell and multi-layer image blocks of the cell cluster according to the multi-layer images;
obtaining a first identification result by a first image identification network according to the multi-layer image blocks of the single cell;
obtaining a second identification result by a second image identification network according to the multi-layer image blocks of the cell cluster, including:
selecting at least one layer of image blocks from the multi-layer image blocks of the cell cluster, and inputting the at least one layer of image blocks into a first part of the second image identification network to obtain a second single-layer identification result corresponding to the multi-layer image blocks of the cell cluster; and
inputting the multi-layer image blocks of the cell cluster into a second part of the second image identification network to obtain a second multi-layer identification result corresponding to the multi-layer image blocks of the cell cluster; and
determining whether an abnormal cell exists in the to-be-detected sample according to the first identification result and the second identification result.
16. The non-transitory computer-readable storage medium according to claim 15 , wherein the obtaining the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster according to the multi-layer images comprises:
dividing the multi-layer images into a plurality of multi-layer image blocks; and
classifying the plurality of multi-layer image blocks into the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster.
17. The non-transitory computer-readable storage medium according to claim 16 , wherein the classifying the plurality of multi-layer image blocks into the multi-layer image blocks of the single cell and the multi-layer image blocks of the cell cluster comprises:
performing the classification by determining a quantity of cell nuclei in each of the plurality of multi-layer image blocks.
18. The non-transitory computer-readable storage medium according to claim 15 , wherein the obtaining the first identification result by the first image identification network according to the multi-layer image blocks of the single cell comprises:
selecting at least one layer of image blocks from the multi-layer image blocks of the single cell, and inputting the at least one layer of image blocks into a first part of the first image identification network to obtain a first single-layer identification result corresponding to the multi-layer image blocks of the single cell;
inputting the multi-layer image blocks of the single cell into a second part of the first image identification network to obtain a first multi-layer identification result corresponding to the multi-layer image blocks of the single cell; and
fusing the first single-layer identification result and the first multi-layer identification result to obtain the first identification result, wherein:
the first part of the first image identification network is used for extracting plane information of the single cell, and the second part of the first image identification network is used for extracting three-dimensional structure information of the single cell.
19. The non-transitory computer-readable storage medium according to claim 18 , wherein the fusing the first single-layer identification result and the first multi-layer identification result to obtain the first identification result comprises:
performing weighted averaging on the first single-layer identification result and the first multi-layer identification result for a plurality of cell types of the single cell, to obtain the first identification result.
20. The non-transitory computer-readable storage medium according to claim 15 , wherein the obtaining the second identification result by the second image identification network according to the multi-layer image blocks of the cell cluster further comprises:
fusing the second single-layer identification result and the second multi-layer identification result to obtain the second identification result, wherein:
the first part of the second image identification network is used for extracting plane information of the cell cluster, and the second part of the second image identification network is used for extracting three-dimensional structure information of the cell cluster.