Information processing apparatus, method, and program
A processor uses a trained neural network to derive a property score for each of a plurality of predetermined property items regarding a structure of interest included in an image; corrects a property score for at least one property item among the plurality of property items by referring to information indicating a relationship between the plurality of property items, the information being derived by analyzing a co-occurrence relationship between descriptions of properties included in a plurality of sentences; and derives a discrimination result for the plurality of property items regarding the structure of interest based on the corrected property score.
1 . An information processing apparatus comprising at least one processor,
wherein the processor is configured to:
use a trained neural network to derive a property score for each of a plurality of predetermined property items regarding a structure of interest included in an image;
correct a property score for at least one property item among the plurality of property items by referring to information indicating a relationship between the plurality of property items, the information being derived by analyzing a co-occurrence relationship between descriptions of properties included in a plurality of sentences; and
derive a discrimination result for the plurality of property items regarding the structure of interest based on the corrected property score.
2 . The information processing apparatus according to claim 1 ,
wherein the processor is configured to:
further train the neural network using supervised training data in which a structure of interest included in a medical image and the plurality of property items regarding the structure of interest are specified; and
update the information indicating the relationship based on a result of the training.
3 . The information processing apparatus according to claim 1 ,
wherein the information indicating the relationship is a relationship matrix in which a weight, which is larger as a co-occurrence relationship between the plurality of property items is stronger, is defined as an element.
4 . The information processing apparatus according to claim 2 ,
wherein the information indicating the relationship is a relationship matrix in which a weight, which is larger as a co-occurrence relationship between the plurality of property items is stronger, is defined as an element.
5 . The information processing apparatus according to claim 3 ,
wherein the weight is scaled within a predetermined range.
6 . The information processing apparatus according to claim 4 ,
wherein the weight is scaled within a predetermined range.
7 . The information processing apparatus according to claim 3 ,
wherein the processor is configured to:
present the relationship matrix; and
correct the relationship matrix by receiving a correction of the weight in the presented relationship matrix.
8 . The information processing apparatus according to claim 4 ,
wherein the processor is configured to:
present the relationship matrix; and
correct the relationship matrix by receiving a correction of the weight in the presented relationship matrix.
9 . The information processing apparatus according to claim 5 ,
wherein the processor is configured to:
present the relationship matrix; and
correct the relationship matrix by receiving a correction of the weight in the presented relationship matrix.
10 . The information processing apparatus according to claim 6 ,
wherein the processor is configured to:
present the relationship matrix; and
correct the relationship matrix by receiving a correction of the weight in the presented relationship matrix.
11 . The information processing apparatus according to claim 1 ,
wherein the trained neural network is constructed by machine-learning a convolutional neural network, and
the processor is configured to correct the property score using one fully-connected layer to which an output of the convolutional neural network is input and the information indicating the relationship is applied.
12 . An information processing method comprising:
using a trained neural network to derive a property score for each of a plurality of predetermined property items regarding a structure of interest included in an image;
correcting a property score for at least one property item among the plurality of property items by referring to information indicating a relationship between the plurality of property items, the information being derived by analyzing a co-occurrence relationship between descriptions of properties included in a plurality of sentences; and
deriving a discrimination result for the plurality of property items regarding the structure of interest based on the corrected property score.
13 . A non-transitory computer-readable storage medium that stores an information processing program for causing a computer execute:
a procedure of using a trained neural network to derive a property score for each of a plurality of predetermined property items regarding a structure of interest included in an image;
a procedure of correcting a property score for at least one property item among the plurality of property items by referring to information indicating a relationship between the plurality of property items, the information being derived by analyzing a co-occurrence relationship between descriptions of properties included in a plurality of sentences; and
a procedure of deriving a discrimination result for the plurality of property items regarding the structure of interest based on the corrected property score.