In the above-described embodiment, the convolutional neural network is used as each CNN. However, the technology according to the present disclosure is not limited thereto. For example, neural networks including a plurality of processing layers, such as a deep neural network (DNN) and a recurrent neural network (RNN) may be used. In addition, all neural networks may not be the same neural network. For example, the first CNN 31 may be a convolutional neural network and the other CNNs may be recurrent neural networks. The type of CNN may be appropriately changed.

In the above-described embodiment, the CNNs other than the first CNN 31 which is the common learning unit according to the present disclosure are not connected to each other. However, in the technology according to the present disclosure, the CNNs other than the first CNN 31 may be connected to each other.

In the above-described embodiment, the non-contrast-enhanced CT images are used as the CT images Bc1 and Bc0. However, both the contrast-enhanced CT image and the non-contrast-enhanced CT image may be used to train the discriminator 23. As such, the use of the trained discriminator 23 makes it possible to discriminate a disease region even in a case in which the CT image which is a discrimination target is either a contrast-enhanced CT image or a non-contrast-enhanced CT image.