[MA] Software Framework for Hierarchical Encoding and NLP-Driven Standardization of Cytogenetic Reporting

Advancements in machine learning and natural language processing (NLP) have the potential to streamline clinical workflows, especially within cancer research. Cytogenetic reports, essential for identifying chromosomal abnormalities, can be complex and time-consuming to interpret. The International System for Human Cytogenetic Nomenclature (ISCN) provides a standardized reporting format, but real-world reports often contain a mixture of formal ISCN strings and natural language descriptions. This variability hinders efficient analysis and rapid diagnosis.
We introduce a framework for hierarchical coding and NLP-driven standardization of cytogenetic reports. inspired by work by Gumz et al. on automated interpretation, the framework parses ISCN strings and employs NLP techniques to interpret and encode non-ISCN descriptions, ensuring compliance with ISCN for further analysis. A hierarchical encoding scheme allows for dynamic updates without re-encoding the entire dataset. Our framework incorporates a user interaction module for continuous refinement. It streamlines the analysis of cytogenetic reports, facilitates timely diagnosis, and adapts to the evolving landscape of cytogenetic discoveries.

Betreuer: Karsten Wendt