DDX-n: Differenital Diagnostic Note Generation Dataset

A differential diagnostic (DDX) note generation dataset. Differential diagnostic note generation is a natural language generation task. The picture shows an example of the input and output of LLMs generating differential diagnoses and relevant evidence. The input includes "Case Description" and "Primary Diagnosis". The output is "Differential Diagnoses & Evidence". "SP", "UP", and "C" represent "Supporting Points", "Unsupporing Points", and "Conclusion", respectively. Models are required to generate the DDX notes with case descriptions and primary diagnoses as premise.

We manually annotate the diagnostic evidence and differential diagnostic notes based on case descriptions and primary diagnoses. Doctors (not annotators) reviewed the annotation to guarantee the quality.