Spaces:
Sleeping
Sleeping
File size: 16,754 Bytes
a297be3 70be421 bba5f79 90f10bf a297be3 bba5f79 90f10bf bba5f79 90f10bf bba5f79 90f10bf a297be3 bba5f79 a297be3 90f10bf bba5f79 90f10bf bba5f79 90f10bf bba5f79 90f10bf bba5f79 a297be3 bba5f79 a297be3 bba5f79 a297be3 90f10bf bba5f79 90f10bf bba5f79 a297be3 70be421 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 |
# import streamlit as st
# import torch
# from transformers import LongformerTokenizer, LongformerForSequenceClassification
# # Load the fine-tuned model and tokenizer
# model_path = "./clinical_longformer"
# tokenizer = LongformerTokenizer.from_pretrained(model_path)
# model = LongformerForSequenceClassification.from_pretrained(model_path)
# model.eval() # Set the model to evaluation mode
# # ICD-9 code columns used during training
# icd9_columns = [
# '038.9', '244.9', '250.00', '272.0', '272.4', '276.1', '276.2', '285.1', '285.9',
# '287.5', '305.1', '311', '36.15', '37.22', '37.23', '38.91', '38.93', '39.61',
# '39.95', '401.9', '403.90', '410.71', '412', '414.01', '424.0', '427.31', '428.0',
# '486', '496', '507.0', '511.9', '518.81', '530.81', '584.9', '585.9', '599.0',
# '88.56', '88.72', '93.90', '96.04', '96.6', '96.71', '96.72', '99.04', '99.15',
# '995.92', 'V15.82', 'V45.81', 'V45.82', 'V58.61'
# ]
# # Function for making predictions
# def predict_icd9(texts, tokenizer, model, threshold=0.5):
# inputs = tokenizer(
# texts,
# padding="max_length",
# truncation=True,
# max_length=512,
# return_tensors="pt"
# )
# with torch.no_grad():
# outputs = model(
# input_ids=inputs["input_ids"],
# attention_mask=inputs["attention_mask"]
# )
# logits = outputs.logits
# probabilities = torch.sigmoid(logits)
# predictions = (probabilities > threshold).int()
# predicted_icd9 = []
# for pred in predictions:
# codes = [icd9_columns[i] for i, val in enumerate(pred) if val == 1]
# predicted_icd9.append(codes)
# return predicted_icd9
# # Streamlit UI
# st.title("ICD-9 Code Prediction")
# st.sidebar.header("Model Options")
# model_option = st.sidebar.selectbox("Select Model", [ "ClinicalLongformer"])
# threshold = st.sidebar.slider("Prediction Threshold", 0.0, 1.0, 0.5, 0.01)
# st.write("### Enter Medical Summary")
# input_text = st.text_area("Medical Summary", placeholder="Enter clinical notes here...")
# if st.button("Predict"):
# if input_text.strip():
# predictions = predict_icd9([input_text], tokenizer, model, threshold)
# st.write("### Predicted ICD-9 Codes")
# for code in predictions[0]:
# st.write(f"- {code}")
# else:
# st.error("Please enter a medical summary.")
# import torch
# import pandas as pd
# import streamlit as st
# from transformers import LongformerTokenizer, LongformerForSequenceClassification
# # Load the fine-tuned model and tokenizer
# model_path = "./clinical_longformer"
# tokenizer = LongformerTokenizer.from_pretrained(model_path)
# model = LongformerForSequenceClassification.from_pretrained(model_path)
# model.eval() # Set the model to evaluation mode
# # Load the ICD-9 descriptions from CSV into a dictionary
# icd9_desc_df = pd.read_csv("D_ICD_DIAGNOSES.csv") # Adjust the path to your CSV file
# icd9_desc_df['ICD9_CODE'] = icd9_desc_df['ICD9_CODE'].astype(str) # Ensure ICD9_CODE is string type for matching
# icd9_descriptions = dict(zip(icd9_desc_df['ICD9_CODE'].str.replace('.', ''), icd9_desc_df['LONG_TITLE'])) # Remove decimals in ICD9 code for matching
# # ICD-9 code columns used during training
# icd9_columns = [
# '038.9', '244.9', '250.00', '272.0', '272.4', '276.1', '276.2', '285.1', '285.9',
# '287.5', '305.1', '311', '36.15', '37.22', '37.23', '38.91', '38.93', '39.61',
# '39.95', '401.9', '403.90', '410.71', '412', '414.01', '424.0', '427.31', '428.0',
# '486', '496', '507.0', '511.9', '518.81', '530.81', '584.9', '585.9', '599.0',
# '88.56', '88.72', '93.90', '96.04', '96.6', '96.71', '96.72', '99.04', '99.15',
# '995.92', 'V15.82', 'V45.81', 'V45.82', 'V58.61'
# ]
# # Function for making predictions
# def predict_icd9(texts, tokenizer, model, threshold=0.5):
# inputs = tokenizer(
# texts,
# padding="max_length",
# truncation=True,
# max_length=512,
# return_tensors="pt"
# )
# with torch.no_grad():
# outputs = model(
# input_ids=inputs["input_ids"],
# attention_mask=inputs["attention_mask"]
# )
# logits = outputs.logits
# probabilities = torch.sigmoid(logits)
# predictions = (probabilities > threshold).int()
# predicted_icd9 = []
# for pred in predictions:
# codes = [icd9_columns[i] for i, val in enumerate(pred) if val == 1]
# predicted_icd9.append(codes)
# # Fetch descriptions for the predicted ICD-9 codes from the pre-loaded descriptions
# predictions_with_desc = []
# for codes in predicted_icd9:
# code_with_desc = [(code, icd9_descriptions.get(code.replace('.', ''), "Description not found")) for code in codes]
# predictions_with_desc.append(code_with_desc)
# return predictions_with_desc
# # Streamlit UI
# st.title("ICD-9 Code Prediction")
# st.sidebar.header("Model Options")
# threshold = st.sidebar.slider("Prediction Threshold", 0.0, 1.0, 0.5, 0.01)
# st.write("### Enter Medical Summary")
# input_text = st.text_area("Medical Summary", placeholder="Enter clinical notes here...")
# if st.button("Predict"):
# if input_text.strip():
# predictions = predict_icd9([input_text], tokenizer, model, threshold)
# st.write("### Predicted ICD-9 Codes and Descriptions")
# for code, description in predictions[0]:
# st.write(f"- {code}: {description}")
# else:
# st.error("Please enter a medical summary.")
# import torch
# import pandas as pd
# import streamlit as st
# from transformers import LongformerTokenizer, LongformerForSequenceClassification
# # Load the fine-tuned model and tokenizer
# model_path = "./clinical_longformer"
# tokenizer = LongformerTokenizer.from_pretrained(model_path)
# model = LongformerForSequenceClassification.from_pretrained(model_path)
# model.eval() # Set the model to evaluation mode
# # Load the ICD-9 descriptions from CSV into a dictionary
# icd9_desc_df = pd.read_csv("D_ICD_DIAGNOSES.csv") # Adjust the path to your CSV file
# icd9_desc_df['ICD9_CODE'] = icd9_desc_df['ICD9_CODE'].astype(str) # Ensure ICD9_CODE is string type
# icd9_descriptions = dict(zip(icd9_desc_df['ICD9_CODE'].str.replace('.', ''), icd9_desc_df['LONG_TITLE'])) # Remove decimals for matching
# # Load the ICD-9 to ICD-10 mapping
# icd9_to_icd10 = {}
# with open("2015_I9gem.txt", "r") as file:
# for line in file:
# parts = line.strip().split()
# if len(parts) == 3:
# icd9, icd10, _ = parts
# icd9_to_icd10[icd9] = icd10
# # ICD-9 code columns used during training
# icd9_columns = [
# '038.9', '244.9', '250.00', '272.0', '272.4', '276.1', '276.2', '285.1', '285.9',
# '287.5', '305.1', '311', '36.15', '37.22', '37.23', '38.91', '38.93', '39.61',
# '39.95', '401.9', '403.90', '410.71', '412', '414.01', '424.0', '427.31', '428.0',
# '486', '496', '507.0', '511.9', '518.81', '530.81', '584.9', '585.9', '599.0',
# '88.56', '88.72', '93.90', '96.04', '96.6', '96.71', '96.72', '99.04', '99.15',
# '995.92', 'V15.82', 'V45.81', 'V45.82', 'V58.61'
# ]
# # Function for making predictions and mapping to ICD-10
# def predict_icd9(texts, tokenizer, model, threshold=0.5):
# inputs = tokenizer(
# texts,
# padding="max_length",
# truncation=True,
# max_length=512,
# return_tensors="pt"
# )
# with torch.no_grad():
# outputs = model(
# input_ids=inputs["input_ids"],
# attention_mask=inputs["attention_mask"]
# )
# logits = outputs.logits
# probabilities = torch.sigmoid(logits)
# predictions = (probabilities > threshold).int()
# predicted_icd9 = []
# for pred in predictions:
# codes = [icd9_columns[i] for i, val in enumerate(pred) if val == 1]
# predicted_icd9.append(codes)
# # Fetch descriptions and map to ICD-10 codes
# predictions_with_desc = []
# for codes in predicted_icd9:
# code_with_desc = []
# for code in codes:
# icd9_stripped = code.replace('.', '')
# icd10_code = icd9_to_icd10.get(icd9_stripped, "Mapping not found")
# icd9_desc = icd9_descriptions.get(icd9_stripped, "Description not found")
# code_with_desc.append((code, icd9_desc, icd10_code))
# predictions_with_desc.append(code_with_desc)
# return predictions_with_desc
# # Streamlit UI
# st.title("ICD-9 to ICD-10 Code Prediction")
# st.sidebar.header("Model Options")
# threshold = st.sidebar.slider("Prediction Threshold", 0.0, 1.0, 0.5, 0.01)
# st.write("### Enter Medical Summary")
# input_text = st.text_area("Medical Summary", placeholder="Enter clinical notes here...")
# if st.button("Predict"):
# if input_text.strip():
# predictions = predict_icd9([input_text], tokenizer, model, threshold)
# st.write("### Predicted ICD-9 and ICD-10 Codes with Descriptions")
# for icd9_code, description, icd10_code in predictions[0]:
# st.write(f"- ICD-9: {icd9_code} ({description}) -> ICD-10: {icd10_code}")
# else:
# st.error("Please enter a medical summary.")
import os
import torch
import pandas as pd
import streamlit as st
from PIL import Image
from transformers import LongformerTokenizer, LongformerForSequenceClassification
from phi.agent import Agent
from phi.model.google import Gemini
from phi.tools.duckduckgo import DuckDuckGo
# Load the fine-tuned ICD-9 model and tokenizer
model_path = "./clinical_longformer"
tokenizer = LongformerTokenizer.from_pretrained(model_path)
model = LongformerForSequenceClassification.from_pretrained(model_path)
model.eval() # Set the model to evaluation mode
# Load the ICD-9 descriptions from CSV into a dictionary
icd9_desc_df = pd.read_csv("D_ICD_DIAGNOSES.csv") # Adjust the path to your CSV file
icd9_desc_df['ICD9_CODE'] = icd9_desc_df['ICD9_CODE'].astype(str) # Ensure ICD9_CODE is string type for matching
icd9_descriptions = dict(zip(icd9_desc_df['ICD9_CODE'].str.replace('.', ''), icd9_desc_df['LONG_TITLE'])) # Remove decimals in ICD9 code for matching
# ICD-9 code columns used during training
icd9_columns = [
'038.9', '244.9', '250.00', '272.0', '272.4', '276.1', '276.2', '285.1', '285.9',
'287.5', '305.1', '311', '36.15', '37.22', '37.23', '38.91', '38.93', '39.61',
'39.95', '401.9', '403.90', '410.71', '412', '414.01', '424.0', '427.31', '428.0',
'486', '496', '507.0', '511.9', '518.81', '530.81', '584.9', '585.9', '599.0',
'88.56', '88.72', '93.90', '96.04', '96.6', '96.71', '96.72', '99.04', '99.15',
'995.92', 'V15.82', 'V45.81', 'V45.82', 'V58.61'
]
# Function for making ICD-9 predictions
def predict_icd9(texts, tokenizer, model, threshold=0.5):
inputs = tokenizer(
texts,
padding="max_length",
truncation=True,
max_length=512,
return_tensors="pt"
)
with torch.no_grad():
outputs = model(
input_ids=inputs["input_ids"],
attention_mask=inputs["attention_mask"]
)
logits = outputs.logits
probabilities = torch.sigmoid(logits)
predictions = (probabilities > threshold).int()
predicted_icd9 = []
for pred in predictions:
codes = [icd9_columns[i] for i, val in enumerate(pred) if val == 1]
predicted_icd9.append(codes)
predictions_with_desc = []
for codes in predicted_icd9:
code_with_desc = [(code, icd9_descriptions.get(code.replace('.', ''), "Description not found")) for code in codes]
predictions_with_desc.append(code_with_desc)
return predictions_with_desc
# Streamlit UI
st.title("Medical Diagnosis Assistant")
option = st.selectbox(
"Choose Diagnosis Method",
("ICD-9 Code Prediction", "Medical Image Analysis")
)
# ICD-9 Code Prediction
if option == "ICD-9 Code Prediction":
st.write("### Enter Medical Summary")
input_text = st.text_area("Medical Summary", placeholder="Enter clinical notes here...")
threshold = st.slider("Prediction Threshold", 0.0, 1.0, 0.5, 0.01)
if st.button("Predict ICD-9 Codes"):
if input_text.strip():
predictions = predict_icd9([input_text], tokenizer, model, threshold)
st.write("### Predicted ICD-9 Codes and Descriptions")
for code, description in predictions[0]:
st.write(f"- {code}: {description}")
else:
st.error("Please enter a medical summary.")
# Medical Image Analysis
elif option == "Medical Image Analysis":
if "GOOGLE_API_KEY" not in st.session_state:
st.warning("Please enter your Google API Key in the sidebar to continue")
else:
medical_agent = Agent(
model=Gemini(
api_key=st.session_state.GOOGLE_API_KEY,
id="gemini-2.0-flash-exp"
),
tools=[DuckDuckGo()],
markdown=True
)
query = """
You are a highly skilled medical imaging expert with extensive knowledge in radiology and diagnostic imaging. Analyze the patient's medical image and structure your response as follows:
### 1. Image Type & Region
- Specify imaging modality (X-ray/MRI/CT/Ultrasound/etc.)
- Identify the patient's anatomical region and positioning
- Comment on image quality and technical adequacy
### 2. Key Findings
- List primary observations systematically
- Note any abnormalities in the patient's imaging with precise descriptions
- Include measurements and densities where relevant
- Describe location, size, shape, and characteristics
- Rate severity: Normal/Mild/Moderate/Severe
### 3. Diagnostic Assessment
- Provide primary diagnosis with confidence level
- List differential diagnoses in order of likelihood
- Support each diagnosis with observed evidence from the patient's imaging
- Note any critical or urgent findings
### 4. Patient-Friendly Explanation
- Explain the findings in simple, clear language that the patient can understand
- Avoid medical jargon or provide clear definitions
- Include visual analogies if helpful
- Address common patient concerns related to these findings
### 5. Research Context
- Use the DuckDuckGo search tool to find recent medical literature about similar cases
- Provide a list of relevant medical links
- Include key references to support your analysis
"""
upload_container = st.container()
image_container = st.container()
analysis_container = st.container()
with upload_container:
uploaded_file = st.file_uploader(
"Upload Medical Image",
type=["jpg", "jpeg", "png", "dicom"],
help="Supported formats: JPG, JPEG, PNG, DICOM"
)
if uploaded_file is not None:
with image_container:
col1, col2, col3 = st.columns([1, 2, 1])
with col2:
image = Image.open(uploaded_file)
width, height = image.size
aspect_ratio = width / height
new_width = 500
new_height = int(new_width / aspect_ratio)
resized_image = image.resize((new_width, new_height))
st.image(resized_image, caption="Uploaded Medical Image", use_container_width=True)
analyze_button = st.button("🔍 Analyze Image")
with analysis_container:
if analyze_button:
image_path = "temp_medical_image.png"
with open(image_path, "wb") as f:
f.write(uploaded_file.getbuffer())
with st.spinner("🔄 Analyzing image... Please wait."):
try:
response = medical_agent.run(query, images=[image_path])
st.markdown("### 📋 Analysis Results")
st.markdown(response.content)
except Exception as e:
st.error(f"Analysis error: {e}")
finally:
if os.path.exists(image_path):
os.remove(image_path)
else:
st.info("👆 Please upload a medical image to begin analysis")
|