Spaces:
Sleeping
Sleeping
File size: 3,717 Bytes
bf72463 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 |
import torch
import json
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from sklearn.preprocessing import MultiLabelBinarizer
import numpy as np
import re
# ----------------------------------------------------------------------
# Text Preprocessing (same as during training)
# ----------------------------------------------------------------------
def preprocess_text(text: str) -> str:
"""
Perform advanced text cleaning:
- Convert to lowercase
- Remove bracketed deidentifications [**...**]
- Remove excessive punctuation
- Convert multiple spaces/newlines to single space
- Strip whitespace
"""
text = text.lower()
text = re.sub(r"\[\*\*.*?\*\*\]", " ", text) # remove deidentified brackets
text = re.sub(r"([!?.,])\1+", r"\1", text) # collapse repeated punctuation
text = re.sub(r"[\r\n\t]+", " ", text) # collapse lines/tabs to space
text = re.sub(r"\s+", " ", text) # multiple spaces -> single
text = text.strip()
return text
# ----------------------------------------------------------------------
# Load Trained Model and Artifacts
# ----------------------------------------------------------------------
def load_trained_model(model_dir: str):
"""
Load the trained model, tokenizer, and MultiLabelBinarizer.
"""
# Load the model
model = AutoModelForSequenceClassification.from_pretrained(model_dir)
model.eval() # Set to evaluation mode
# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_dir)
# Load the MultiLabelBinarizer
with open(f"{model_dir}/mlb_classes.json", "r") as f:
top_codes_list = json.load(f)
mlb = MultiLabelBinarizer(classes=top_codes_list)
mlb.fit([[]]) # Initialize the binarizer
return model, tokenizer, mlb
# ----------------------------------------------------------------------
# Predict ICD-9 Codes
# ----------------------------------------------------------------------
def predict_icd9(input_text: str, model, tokenizer, mlb, max_length=512, threshold=0.5):
"""
Predict ICD-9 codes for a given clinical text.
"""
# Preprocess the input text
processed_text = preprocess_text(input_text)
# Tokenize the input
inputs = tokenizer(
processed_text,
return_tensors="pt",
truncation=True,
max_length=max_length,
padding="max_length"
)
# Perform inference
with torch.no_grad():
logits = model(**inputs).logits
# Apply sigmoid to convert logits to probabilities
probs = torch.sigmoid(logits).squeeze().cpu().numpy()
# Apply threshold to get predicted labels
y_pred = (probs > threshold).astype(int)
# Decode the predicted labels back to ICD-9 codes
predicted_codes = mlb.inverse_transform(np.array([y_pred])) # Ensure 2D array
return predicted_codes[0] # Return as a list of codes
# ----------------------------------------------------------------------
# Inference Example
# ----------------------------------------------------------------------
if __name__ == "__main__":
# Directory where the model and artifacts are saved
model_dir = "./"
# Load the model and related artifacts
model, tokenizer, mlb = load_trained_model(model_dir)
# Example input text
input_text = """Acute nasopharyngitis , true acute abnormality """
# Predict ICD-9 codes
predicted_codes = predict_icd9(input_text, model, tokenizer, mlb, threshold= 0.2)
# Print the predicted ICD-9 codes
print("Predicted ICD-9 Codes:", predicted_codes)
|