README.md
5.2 KB · 88 lines · markdown Raw
1 ---
2 language:
3 - en
4 thumbnail: "https://www.onebraveidea.org/wp-content/uploads/2019/07/OBI-Logo-Website.png"
5 tags:
6 - deidentification
7 - medical notes
8 - ehr
9 - phi
10 datasets:
11 - I2B2
12 metrics:
13 - F1
14 - Recall
15 - Precision
16 widget:
17 - text: "Physician Discharge Summary Admit date: 10/12/1982 Discharge date: 10/22/1982 Patient Information Jack Reacher, 54 y.o. male (DOB = 1/21/1928)."
18 - text: "Home Address: 123 Park Drive, San Diego, CA, 03245. Home Phone: 202-555-0199 (home)."
19 - text: "Hospital Care Team Service: Orthopedics Inpatient Attending: Roger C Kelly, MD Attending phys phone: (634)743-5135 Discharge Unit: HCS843 Primary Care Physician: Hassan V Kim, MD 512-832-5025."
20 license: mit
21 ---
22
23 # Model Description
24
25 * A RoBERTa [[Liu et al., 2019]](https://arxiv.org/pdf/1907.11692.pdf) model fine-tuned for de-identification of medical notes.
26 * Sequence Labeling (token classification): The model was trained to predict protected health information (PHI/PII) entities (spans). A list of protected health information categories is given by [HIPAA](https://www.hhs.gov/hipaa/for-professionals/privacy/laws-regulations/index.html).
27 * A token can either be classified as non-PHI or as one of the 11 PHI types. Token predictions are aggregated to spans by making use of BILOU tagging.
28 * The PHI labels that were used for training and other details can be found here: [Annotation Guidelines](https://github.com/obi-ml-public/ehr_deidentification/blob/master/AnnotationGuidelines.md)
29 * More details on how to use this model, the format of data and other useful information is present in the GitHub repo: [Robust DeID](https://github.com/obi-ml-public/ehr_deidentification).
30
31
32 # How to use
33
34 * A demo on how the model works (using model predictions to de-identify a medical note) is on this space: [Medical-Note-Deidentification](https://huggingface.co/spaces/obi/Medical-Note-Deidentification).
35 * Steps on how this model can be used to run a forward pass can be found here: [Forward Pass](https://github.com/obi-ml-public/ehr_deidentification/tree/master/steps/forward_pass)
36 * In brief, the steps are:
37 * Sentencize (the model aggregates the sentences back to the note level) and tokenize the dataset.
38 * Use the predict function of this model to gather the predictions (i.e., predictions for each token).
39 * Additionally, the model predictions can be used to remove PHI from the original note/text.
40
41
42 # Dataset
43
44 * The I2B2 2014 [[Stubbs and Uzuner, 2015]](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4978170/) dataset was used to train this model.
45
46 | | I2B2 | | I2B2 | |
47 | --------- | --------------------- | ---------- | -------------------- | ---------- |
48 | | TRAIN SET - 790 NOTES | | TEST SET - 514 NOTES | |
49 | PHI LABEL | COUNT | PERCENTAGE | COUNT | PERCENTAGE |
50 | DATE | 7502 | 43.69 | 4980 | 44.14 |
51 | STAFF | 3149 | 18.34 | 2004 | 17.76 |
52 | HOSP | 1437 | 8.37 | 875 | 7.76 |
53 | AGE | 1233 | 7.18 | 764 | 6.77 |
54 | LOC | 1206 | 7.02 | 856 | 7.59 |
55 | PATIENT | 1316 | 7.66 | 879 | 7.79 |
56 | PHONE | 317 | 1.85 | 217 | 1.92 |
57 | ID | 881 | 5.13 | 625 | 5.54 |
58 | PATORG | 124 | 0.72 | 82 | 0.73 |
59 | EMAIL | 4 | 0.02 | 1 | 0.01 |
60 | OTHERPHI | 2 | 0.01 | 0 | 0 |
61 | TOTAL | 17171 | 100 | 11283 | 100 |
62
63
64 # Training procedure
65
66 * Steps on how this model was trained can be found here: [Training](https://github.com/obi-ml-public/ehr_deidentification/tree/master/steps/train). The "model_name_or_path" was set to: "roberta-large".
67 * The dataset was sentencized with the en_core_sci_sm sentencizer from spacy.
68 * The dataset was then tokenized with a custom tokenizer built on top of the en_core_sci_sm tokenizer from spacy.
69 * For each sentence we added 32 tokens on the left (from previous sentences) and 32 tokens on the right (from the next sentences).
70 * The added tokens are not used for learning - i.e, the loss is not computed on these tokens - they are used as additional context.
71 * Each sequence contained a maximum of 128 tokens (including the 32 tokens added on). Longer sequences were split.
72 * The sentencized and tokenized dataset with the token level labels based on the BILOU notation was used to train the model.
73 * The model is fine-tuned from a pre-trained RoBERTa model.
74
75 * Training details:
76 * Input sequence length: 128
77 * Batch size: 32 (16 with 2 gradient accumulation steps)
78 * Optimizer: AdamW
79 * Learning rate: 5e-5
80 * Dropout: 0.1
81
82
83 ## Results
84
85 # Questions?
86
87 Post a Github issue on the repo: [Robust DeID](https://github.com/obi-ml-public/ehr_deidentification).
88