Method and system for filtering obstacle data in machine learning of medical images
The present disclosure relates to a method for filtering selectively obstacle to be an obstacle to machine learning according to a learning purpose and a system thereof. A system for filtering obstacle data in machine learning of medical images may include an obstacle data definition unit configured to receive definitions of obstacle data according to a machine learning purpose; a filter generation unit configured to generate a filter for filtering the obstacle data; and a filtering unit configured to remove obstacle data in machine learning using the generated filter.
1. A method for filtering obstacle data in machine learning of medical images using a system including an obstacle data definition unit, a filter generation unit, and a filtering unit, the method comprising:
receiving, by the obstacle data definition unit, definitions of obstacle data according to a machine learning purpose;
generating, by the filter generation unit, a filter for filtering the obstacle data; and
removing, by the filtering unit, the obstacle data using the generated filter,
wherein the obstacle data are classified into absolute obstacle data and relative obstacle data,
wherein the absolute obstacle data is data being an obstacle to learning regardless of the machine learning purpose, and the relative obstacle data is data being an obstacle to learning according to a type of the machine learning purpose,
wherein a type of an image classified as the obstacle data varies according to the machine learning purpose, and
wherein the image classified as the obstacle data is filtered or is not filtered by using the filter according to the machine learning purpose.
2. The method of claim 1 , wherein in the generating of the filter, the filter generation unit generates an obstacle data filter using at least one of an image processing-based filter and a learning-based filter.
3. The method of claim 1 , further comprising:
performing, by a labeling unit, labeling based on the machine learning for the medical image data for the machine learning,
wherein the labeling is to classify types of the obstacle data.
4. The method of claim 1 , further comprising:
classifying the image into a non-obstacle and an obstacle image,
excluding an absolute obstacle image corresponding to the absolute obstacle data from the obstacle image, and
filtering the obstacle image from which the absolute obstacle image has been excluded according to the machine learning purpose.
5. The method of claim 4 , wherein the image is a capsule endoscopic image, and
wherein the machine learning purpose is a purpose of learning to detect a lesion or a purpose of learning to determine a position of a capsule endoscope.
6. A system for filtering obstacle data in machine learning of medical images, the system comprising:
a processor; and
a memory storing instructions executable by the processor,
wherein the processor is configured to:
receive definitions of obstacle data according to a machine learning purpose;
generate a filter for filtering the obstacle data; and
remove the obstacle data using the generated filter,
wherein the obstacle data are classified into absolute obstacle data and relative obstacle data,
wherein the absolute obstacle data is data being an obstacle to learning regardless of the machine learning purpose and the relative obstacle data is data being an obstacle to learning according to a type of the machine learning purpose,
wherein a type of an image classified as the obstacle data varies according to the machine learning purpose, and
wherein the image classified as the obstacle data is filtered or is not filtered by using the filter according to the machine learning purpose.
7. The system of claim 6 , wherein the processor is further configured to generate an obstacle data filter using at least any one of an image processing-based filter and a learning-based filter.
8. The system of claim 6 ,
wherein the processor is further configured to perform labeling based on the machine learning for the medical image data for the machine learning,
wherein the labeling is to classify types of the obstacle data.
9. The system of claim 6 , wherein the processor is further configured to:
classify the image into a non-obstacle and an obstacle image,
exclude an absolute obstacle image corresponding to the absolute obstacle data from the obstacle image, and
filter the obstacle image from which the absolute obstacle image has been excluded according to the machine learning purpose.
10. The system of claim 9 , wherein the image is a capsule endoscopic image, and
wherein the machine learning purpose is a purpose of learning to detect a lesion or a purpose of learning to determine a position of a capsule endoscope.