Thrombolytic therapy using a therapeutic agent, such as alteplase, is performed for cerebral infarction patients. However, it has been known that the thrombolytic therapy is applied within 4.5 hours from the time when no cerebral infarction has been confirmed and the risk of bleeding after treatment becomes higher as an infarction range becomes wider over time. Therefore, it is necessary to quickly and appropriately discriminate the infarction range using medical images in order to determine whether the thrombolytic therapy is appropriate.

In contrast, it has been known that, in a case in which the infarction region is already wide, the possibility of bleeding is high. However, it is difficult even for a medical specialist to accurately capture the infarction region on the CT image and it is desirable to automatically extract and quantify the infarction region using a computer. For this reason, deep learning which has attracted attention in recent years can be applied as a method for automatically extracting the infarction region. Learning information including a plurality of data sets of CT images and correct infarction regions in the CT images is required for deep learning. However, since the infarction region is not always clear on the CT image, it is difficult to prepare a large amount of data indicating the correct infarction region in the CT image.

The present disclosure has been made in view of the above-mentioned problems and an object of the present disclosure is to provide a technique that discriminates a disease region with high accuracy, using a limited amount of data, even in an image in which it is difficult to prepare a large amount of data indicating a correct disease region.