Medical image processing apparatus, medical image processing method, computer-readable medical-image processing program, moving-object tracking apparatus, and radiation therapy system

A medical image processing apparatus comprising:

• a discriminator taught by machine learning;
• a memory configured to store a parameter of the discriminator by machine learning;
• an image input interface configured to acquire a two-dimensional fluoroscopic image of an object subjected to a radiation therapy by irradiating a target portion with radiation;
• an image determination processor configured to determine whether a feature of the two-dimensional fluoroscopic image included in a normal range or not, by using the normal range indicated by the discriminator of the parameter;
• and
• a determination-signal output interface configured to output a determination signal, the determination signal indicating a result of determination of the image determination processor,
• wherein at least one first image is used for the machine learning of the discriminator, wherein the at least one first image is either entirety of the two-dimensional fluoroscopic image or part of the two-dimensional fluoroscopic image acquired before starting the radiation therapy and the at least one first image is acquired in a normal state in which the object is breathing out,
• the image determination processor is configured to determine whether the feature of a second image is included in the normal range or not, by using the noimal range indicated by the discriminator of the parameter read out from the memory, wherein the second image is the two-dimensional fluoroscopic image acquired by the image input interface during the radiation therapy, and
• the determination-signal output interface is configured to output a determination signal for prohibiting radiation to the target portion when the image determination processor determines that the feature of the second images is not included in the normal range.