Combining domain knowledge and foundation models for one-shot medical image feature localization

A system, comprising:

• at least one memory that stores computer-executable components;
• and
• at least one processor that executes the computer-executable components stored in the at least one memory, wherein the computer-executable components comprise:
• a feature localization component that sequentially determines positions of target features within a target medical image of an anatomical region of a subject, wherein for each current target feature, the feature localization component:
• defines a search region within the target medical image corresponding to a portion of the target medical image associated with the current target feature based on one or more previously determined positions, within the target medical image, of one or more preceding target features;
• and
• determines a position of the target feature based on a measure of similarity between a subset of pixel features of the target image included in the search region and a reference subset of reference pixel features associated with the current target feature;
• and
• a labeling component that generates label information for the target features identifying the target features and their positions and associates the label information with the target medical image.

A non-transitory machine-readable storage medium, comprising executable instructions that, when executed by a processor, facilitate performance of operations, comprising:

• determining positions of target features within a target medical image of an anatomical region of a subject, wherein the determining comprises:
• determining a first position of a first target feature within the target medical image;
• defining a search region within the target medical image corresponding to a portion of the target medical image associated with a second target feature based on a reference spatial relationship between the first target feature and the second target feature;
• and
• determining a second position of the second target feature based on a measure of similarity between a pixel feature subset included in the search region and a reference pixel feature subset associated with the second target feature;
• generating label information for the target features identifying the target features and their positions;
• and
• associating the label information with the target medical image.