Table 2 Descriptive statistics for MIMIC3
From: An explainable CNN approach for medical codes prediction from clinical text
MIMIC-III full | MIMIC-III 50 | |
|---|---|---|
Training documents | 47,724 | 8,067 |
Vocabulary size | 51,917 | 51,917 |
Mean tokens per document | 1,485 | 1,530 |
Mean labels per document | 15.9 | 5.7 |
Total labels | 8922 | 50 |