Symbolic coordinate anatomy for neurology (SCAN)

Area of Science:

• Medical Informatics
• Computational Neurology
• Artificial Intelligence in Medicine

Background:

• The CADUCEUS (formerly INTERNIST-1) program provides expert medical diagnostic advice.
• Extending CADUCEUS to neurology requires a robust neuroanatomic reasoning framework.

Purpose of the Study:

• To develop a neuroanatomic knowledge base (SCAN) for advanced reasoning in neurology.
• To facilitate accurate anatomic localization of neurological findings and lesion consequences.

Main Methods:

• Developed SCAN, a hierarchical knowledge base partitioning the nervous system into nested cubes.
• Represented neuroanatomic structures and their spatial relationships within the cube hierarchy.
• Maintained lists of structures, vascular territories, and functional systems within the knowledge base.

Main Results:

• SCAN facilitates computation of anatomic localization from symptoms.
• Enables characterization of consequences from vascular lesions or mass effects.
• Provides a paradigm for neuroanatomic reasoning within a computational diagnostic system.

Conclusions:

• The SCAN knowledge base effectively supports neuroanatomic reasoning for the CADUCEUS diagnostic program.
• This approach enhances the diagnostic capabilities of computer programs in neurology.
• Facilitates understanding of lesion localization and its clinical implications.