LEARNING DEVICE, METHOD, AND PROGRAM OF DISCRIMINATOR FOR DISCRIMINATING DISEASE REGION, DISCRIMINATOR FOR DISCRIMINATING DISEASE REGION, AND DISEASE REGION DISCRIMINATION DEVICE AND PROGRAM
PROBLEM TO BE SOLVED: To accurately discriminate a disease region by using limited data even in an image in which it is difficult to prepare many pieces of data indicating a correct region of a disease in a learning device, a method, and a program of a discriminator for discriminating a disease region, a discriminator for discriminating a disease region, and a disease region discrimination device and program.SOLUTION: A discriminator includes a common learning unit and a plurality of learning units connected to an output part of the common learning unit. When a first image is input to the common learning unit by using the first image acquired by imaging a subject developing a disease and a plurality of data sets of a correct mask of a disease region in the first image, the discriminator is caused to learn so that information indicating the disease region is output from a first learning unit. When the first image is input to an input part of the common learning unit by using a plurality of data sets of sets of images aligning the first image and the second image of a kind different from that of the first image, the discriminator is caused to learn so that an estimation image of the second image is output from an output part of a second learning unit.SELECTED DRAWING: Figure 2COPYRIGHT: (C)2020,JPO&INPIT【課題】疾患領域を判別する判別器の学習装置、方法及びプログラム、疾患領域を判別する判別器、並びに疾患領域判別装置及びプログラムにおいて、疾患の正解領域を示すデータを多く用意することが困難な画像であっても、限られたデータを用いて疾患領域を精度よく判別する。【解決手段】判別器は、共通学習部と、共通学習部の出力部に接続された複数の学習部と、を備える。疾患を発症している被検体を撮影することにより得られた第１の画像及び第１の画像における疾患領域の正解マスクのデータセットを複数用いて、第１の画像を共通学習部へ入力した場合に、第１の学習部から疾患領域を示す情報が出力されるように学習させる。第１の画像及び第１の画像とは異なる種類の第２の画像とを位置合わせした画像の組のデータセットを複数用いて、第１の画像を共通学習部の入力部へ入力した場合に、第２の学習部の出力部から第２の画像の推定画像が出力されるように学習させる。【選択図】図２
