IMAGE PROCESSING METHOD, IMAGE PROCESSING DEVICE, AND RECORDING MEDIUM
An image processing method according to the present embodiment is a method for performing an inference task on image data using a trained deep learning model in which a plurality of patch parameters are first set corresponding to the inference task. Next, a plurality of patches are generated for each of the patch parameters, an inference result based on the generated patches is acquired, and an optimized patch parameter and patches corresponding to the optimized patch parameter are determined from among the patch parameters based on the inference result. Next, the inference task is performed by using the deep learning model to acquire an inference result based on the patches corresponding to the determined optimized patch parameter.
1 . An image processing method for performing an inference task on image data using a trained deep learning model, the image processing method comprising:
setting a plurality of patch parameters corresponding to the inference task;
planning by generating a plurality of patches for each of the patch parameters, acquiring an inference result based on the generated patches, and determining an optimized patch parameter and patches corresponding to the optimized patch parameter from among the patch parameters based on the inference result; and
inferring by performing the inference task using the deep learning model to acquire an inference result based on the patches corresponding to the determined optimized patch parameter.
2 . The image processing method according to claim 1 , wherein
the planning further includes:
generating the patches for each of the patch parameters;
adjusting the patches; and
updating a region of interest by subtracting the patches from a region of interest in the image data and updating the region of interest to a new region of interest, and
in the new region of interest, the patches are generated by performing the generating the patches, the adjusting the patches, and the updating the region of interest until a volume of the updated new region of interest becomes zero.
3 . The image processing method according to claim 2 , wherein
at the generating the patches,
each of the patches is generated at a position of a corner point of the region of interest, and
at the adjusting the patches,
a size of each patch is set based on the patch parameter,
the size of the patch is adjusted in accordance with a length of the region of interest, and
a position of the adjusted patch is moved so that a range of the region of interest covered by the patch is increased and a range of a region of non-interest covered by the patch is reduced.
4 . The image processing method according to claim 1 , wherein
at the setting, recommended values for the patch parameters are set based on a parameter associated with the inference task in a history database, and
at the planning, a plurality of patches are generated for each of the patch parameters in the recommended values.
5 . The image processing method according to claim 1 , wherein
at the setting, recommended values for the patch parameters are set based on an input by a user, and
at the planning, a plurality of patches are generated for each of the patch parameters in the recommended values.
6 . The image processing method according to claim 1 , wherein
at the setting, a corresponding input image is generated for each of the patch parameters at regular intervals in a preset range of the patch parameter, an inference result of each input image is acquired, and a recommended value for the patch parameter is determined based on the inference result, and
at the planning, a plurality of patches are generated for each of the patch parameters in the recommended values.
7 . The image processing method according to claim 4 , wherein the patch parameters include a base patch size, a patch size ratio, base resolution, and a resolution ratio.
8 . The image processing method according to claim 5 , wherein the patch parameters include a base patch size, a patch size ratio, base resolution, and a resolution ratio.
9 . The image processing method according to claim 6 , wherein the patch parameters include a base patch size, a patch size ratio, base resolution, and a resolution ratio.
10 . The image processing method according to claim 1 , wherein at the planning, the optimized patch parameter and the patches corresponding to the optimized patch parameter are determined based on inference accuracy and an inference time.
11 . The image processing method according to claim 1 , wherein
the planning further includes editing, and
at the editing, the inference result and the determined patches are displayed on a user interface to be able to be edited by a user, and the patches are adjusted in accordance with editing by the user.
12 . The image processing method according to claim 1 , further comprising:
acquiring the image data;
segmenting the image data to extract a region of interest; and
outputting the inference result.
13 . The image processing method according to claim 1 , further comprising:
analyzing clinical data by extracting critical information related to the inference task from clinical data of a subject, wherein
at the planning, the optimized patch parameter and patches corresponding to the optimized patch parameter are determined based on the critical information.
14 . The image processing method according to claim 1 , further comprising:
updating the deep learning model by storing an editing result by a user, and training the deep learning model based on the editing result.
15 . An image processing device configured to perform an inference task on image data using a trained deep learning model, the image processing device comprising processing circuitry configured to:
set a plurality of patch parameters corresponding to the inference task;
generate a plurality of patches for each of the patch parameters, acquire an inference result based on the generated patches, and determine an optimized patch parameter and patches corresponding to the optimized patch parameter from among the patch parameters based on the inference result; and
perform the inference task using the deep learning model to acquire an inference result based on the patches corresponding to the determined optimized patch parameter.
16 . A non-transitory computer-readable recording medium storing therein a computer program configured to, when executed by a processor, perform an inference task on image data using a trained deep learning model that causes a computer to executed:
setting a plurality of patch parameters corresponding to the inference task;
planning by generating a plurality of patches for each of the patch parameters, acquiring an inference result based on the generated patches, and determining an optimized patch parameter and patches corresponding to the optimized patch parameter from among the patch parameters based on the inference result; and
inferring by performing the inference task using the deep learning model to acquire an inference result based on the patches corresponding to the determined optimized patch parameter.