Medical Code Prediction

The task of medical code prediction aims to automatically annotate multiple medical codes from clinical notes, which are selected from a subset of the nearly 70,000 codes in the existing ICD-10 system. This task involves multi-label classification, and clinical notes are typically lengthy, loosely structured, and contain specialized medical terminology, sometimes accompanied by spelling errors. Automating this process can not only significantly save human time but also improve the accuracy and efficiency of medical coding, driving the development of autonomous medical coding systems that perform on par with human coders.