INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND PROGRAM
To evaluate medical data having characteristics which a sorter cannot have fully learnt.SOLUTION: An information processing apparatus according to the present invention has a likelihood acquiring unit for acquiring class likelihood with respect to a first medical data given a correct label by using...
Saved in:
Main Author | |
---|---|
Format | Patent |
Language | English Japanese |
Published |
15.02.2021
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | To evaluate medical data having characteristics which a sorter cannot have fully learnt.SOLUTION: An information processing apparatus according to the present invention has a likelihood acquiring unit for acquiring class likelihood with respect to a first medical data given a correct label by using a first classifier for classifying medical data into classes, a first evaluation unit for evaluating a degree of deviation based on the class likelihood acquired by the likelihood acquiring unit and the class corresponding to the correct label, a determination unit for determining whether or not the degree of deviation obtained by the first evaluation unit satisfies a predetermined reference, and a second evaluation unit for, based on a second medical data satisfying the predetermined reference, evaluating similarity of the second medical data and a third medical data to be classified.SELECTED DRAWING: Figure 4
【課題】 分類器が十分に学習できていない特徴を有する医用データに対して評価をすることを目的とする。【解決手段】 本発明に係る情報処理装置は、医用データをクラス分類する第一の分類器を用いて、正解ラベルが付与された第一の医用データに対するクラス尤度を取得する尤度取得部と、尤度取得部により取得したクラス尤度と、正解ラベルに対応するクラスとに基づいて乖離の程度を評価する第一の評価部と、第一の評価部による乖離の程度が所定の基準を満たすか否かを判別する判別部と、所定の基準を満たす第二の医用データに基づいて、第二の医用データと分類対象の第三の医用データとの類似性を評価する第二の評価部と、を備える。【選択図】 図4 |
---|---|
Bibliography: | Application Number: JP20190132264 |