Spaces:
Sleeping
Sleeping
File size: 17,581 Bytes
32519eb |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 |
"""
Medical Data Collection Pipeline for Synthex MVP
Collects medical text from free sources for training data
"""
import requests
import pandas as pd
from datasets import load_dataset
import time
import json
from pathlib import Path
from typing import List, Dict, Any
import logging
import sys
from tqdm import tqdm
from bs4 import BeautifulSoup
import re
from datetime import datetime
# Setup logging
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s - %(levelname)s - %(message)s',
handlers=[
logging.StreamHandler(sys.stdout),
logging.FileHandler('data_collection.log')
]
)
logger = logging.getLogger(__name__)
class MedicalDataCollector:
def __init__(self, output_dir: str = "data/raw"):
self.output_dir = Path(output_dir)
self.output_dir.mkdir(parents=True, exist_ok=True)
self.stats = {
"total_samples": 0,
"sources": {},
"errors": [],
"start_time": datetime.now()
}
logger.info(f"Initialized MedicalDataCollector with output directory: {self.output_dir}")
def collect_huggingface_datasets(self) -> Dict[str, List]:
"""Collect medical datasets from Hugging Face Hub"""
# Only include datasets that are known to exist and are medical-related
datasets_to_collect = [
"medical_questions_pairs",
"medalpaca/medical_meadow_medical_flashcards",
"gamino/wiki_medical_terms",
("pubmed_qa", "pqa_artificial") # pubmed_qa requires a config
]
collected_data = {}
for dataset_entry in tqdm(datasets_to_collect, desc="Collecting Hugging Face datasets"):
try:
if isinstance(dataset_entry, tuple):
dataset_name, config = dataset_entry
logger.info(f"Loading dataset: {dataset_name} with config: {config}")
dataset = load_dataset(dataset_name, config, split="train")
dataset_key = f"{dataset_name}_{config}"
else:
dataset_name = dataset_entry
logger.info(f"Loading dataset: {dataset_name}")
dataset = load_dataset(dataset_name, split="train")
dataset_key = dataset_name
# Convert to list of dictionaries
data_list = []
for item in dataset:
processed_item = self._process_dataset_item(item)
if processed_item:
data_list.append(processed_item)
if data_list:
collected_data[dataset_key] = data_list
self.stats["sources"][dataset_key] = len(data_list)
self.stats["total_samples"] += len(data_list)
# Save to file
output_file = self.output_dir / f"{dataset_key.replace('/', '_')}.json"
with open(output_file, 'w', encoding='utf-8') as f:
json.dump(data_list, f, indent=2, ensure_ascii=False)
logger.info(f"Saved {len(data_list)} samples from {dataset_key} to {output_file}")
else:
logger.warning(f"No valid data found in dataset: {dataset_key}")
time.sleep(1) # Be respectful to APIs
except Exception as e:
error_msg = f"Failed to load {dataset_entry}: {str(e)}"
logger.error(error_msg, exc_info=True)
self.stats["errors"].append(error_msg)
continue
return collected_data
def collect_pubmed_abstracts(self, queries: List[str] = None, max_results: int = 1000) -> List[Dict]:
"""Collect PubMed abstracts via API"""
if queries is None:
queries = [
"clinical notes",
"medical case reports",
"patient discharge summaries",
"medical laboratory reports",
"medical imaging reports"
]
all_abstracts = []
for query in tqdm(queries, desc="Collecting PubMed abstracts"):
try:
abstracts = self._collect_pubmed_query(query, max_results)
all_abstracts.extend(abstracts)
self.stats["sources"]["pubmed_" + query.replace(" ", "_")] = len(abstracts)
self.stats["total_samples"] += len(abstracts)
except Exception as e:
error_msg = f"Failed to collect PubMed abstracts for {query}: {str(e)}"
logger.error(error_msg)
self.stats["errors"].append(error_msg)
continue
# Save all abstracts
if all_abstracts:
output_file = self.output_dir / "pubmed_abstracts.json"
with open(output_file, 'w', encoding='utf-8') as f:
json.dump(all_abstracts, f, indent=2, ensure_ascii=False)
return all_abstracts
def _collect_pubmed_query(self, query: str, max_results: int) -> List[Dict]:
"""Collect PubMed abstracts for a specific query"""
base_url = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/"
search_url = f"{base_url}esearch.fcgi"
search_params = {
"db": "pubmed",
"term": query,
"retmax": max_results,
"retmode": "json",
"sort": "relevance"
}
try:
response = requests.get(search_url, params=search_params)
response.raise_for_status() # Raise exception for bad status codes
search_results = response.json()
# Check rate limits
rate_limit = int(response.headers.get('X-RateLimit-Limit', '3'))
rate_remaining = int(response.headers.get('X-RateLimit-Remaining', '0'))
logger.info(f"Rate limit: {rate_remaining}/{rate_limit} requests remaining")
if rate_remaining <= 1:
logger.warning("Rate limit nearly reached, waiting 60 seconds")
time.sleep(60)
except requests.exceptions.RequestException as e:
logger.error(f"Failed to fetch PubMed search results for query '{query}': {str(e)}")
return []
except json.JSONDecodeError as e:
logger.error(f"Failed to parse PubMed search results for query '{query}': {str(e)}")
return []
if "esearchresult" not in search_results:
logger.warning(f"No search results found for query '{query}'")
return []
id_list = search_results["esearchresult"]["idlist"]
abstracts = []
batch_size = 100
for i in range(0, len(id_list), batch_size):
batch_ids = id_list[i:i+batch_size]
ids_str = ",".join(batch_ids)
fetch_url = f"{base_url}efetch.fcgi"
fetch_params = {
"db": "pubmed",
"id": ids_str,
"retmode": "xml"
}
try:
response = requests.get(fetch_url, params=fetch_params)
response.raise_for_status()
# Check rate limits
rate_limit = int(response.headers.get('X-RateLimit-Limit', '3'))
rate_remaining = int(response.headers.get('X-RateLimit-Remaining', '0'))
logger.info(f"Rate limit: {rate_remaining}/{rate_limit} requests remaining")
if rate_remaining <= 1:
logger.warning("Rate limit nearly reached, waiting 60 seconds")
time.sleep(60)
# Parse XML with proper features
soup = BeautifulSoup(response.text, 'lxml', features="xml")
except requests.exceptions.RequestException as e:
logger.error(f"Failed to fetch PubMed article batch {i//batch_size + 1}: {str(e)}")
continue
except Exception as e:
logger.error(f"Failed to parse PubMed article batch {i//batch_size + 1}: {str(e)}")
continue
for article in soup.find_all('PubmedArticle'):
try:
abstract = article.find('Abstract')
if abstract:
abstract_text = abstract.get_text().strip()
if len(abstract_text) > 100: # Filter out very short abstracts
title = article.find('ArticleTitle')
if not title:
continue
title_text = title.get_text().strip()
pub_date = article.find('PubDate')
year = "Unknown"
if pub_date and pub_date.find('Year'):
year = pub_date.find('Year').get_text().strip()
abstracts.append({
"title": title_text,
"abstract": abstract_text,
"year": year,
"source": "pubmed",
"query": query
})
except Exception as e:
logger.debug(f"Failed to process article in batch {i//batch_size + 1}: {str(e)}")
continue
# Always wait between batches to respect rate limits
time.sleep(1)
logger.info(f"Collected {len(abstracts)} abstracts for query '{query}'")
return abstracts
def create_training_dataset(self) -> pd.DataFrame:
"""Combine all collected data into training dataset"""
all_texts = []
# Load all collected datasets
for json_file in tqdm(list(self.output_dir.glob("*.json")), desc="Processing collected data"):
try:
with open(json_file, 'r', encoding='utf-8') as f:
data = json.load(f)
# Extract text content
for item in data:
text_content = self._extract_text_content(item)
if text_content:
processed_text = self._clean_text(text_content)
if processed_text:
all_texts.append({
"text": processed_text,
"source": json_file.stem,
"length": len(processed_text),
"type": self._determine_text_type(processed_text)
})
except Exception as e:
error_msg = f"Failed to process {json_file}: {str(e)}"
logger.error(error_msg)
self.stats["errors"].append(error_msg)
continue
# Create DataFrame
df = pd.DataFrame(all_texts)
# Basic filtering
df = df[df['length'] > 100] # Remove very short texts
df = df[df['length'] < 5000] # Remove very long texts
# Remove duplicates
df = df.drop_duplicates(subset=['text'])
# Save processed dataset
output_file = self.output_dir.parent / "processed" / "training_data.csv"
output_file.parent.mkdir(exist_ok=True)
df.to_csv(output_file, index=False, encoding='utf-8')
# Update stats
self.stats["final_samples"] = len(df)
self.stats["text_types"] = df['type'].value_counts().to_dict()
logger.info(f"Created training dataset with {len(df)} samples")
return df
def _process_dataset_item(self, item: Dict) -> Dict:
"""Process and validate a dataset item"""
try:
# Extract text content
text = self._extract_text_content(item)
if not text or len(text) < 100:
return None
# Clean text
cleaned_text = self._clean_text(text)
if not cleaned_text:
return None
# Create processed item
processed = {
"text": cleaned_text,
"source": "huggingface",
"type": self._determine_text_type(cleaned_text)
}
# Add metadata if available
for key in ['title', 'question', 'answer', 'instruction']:
if key in item:
processed[key] = str(item[key])
return processed
except Exception:
return None
def _extract_text_content(self, item: Dict) -> str:
"""Extract relevant text content from dataset item"""
# Common text fields in medical datasets
text_fields = ['text', 'content', 'abstract', 'question', 'answer',
'instruction', 'output', 'input', 'context']
for field in text_fields:
if field in item and item[field]:
return str(item[field])
# Fallback: combine multiple fields
combined_text = ""
for key, value in item.items():
if isinstance(value, str) and len(value) > 20:
combined_text += f"{value} "
return combined_text.strip()
def _clean_text(self, text: str) -> str:
"""Clean and normalize text"""
if not text:
return ""
# Remove special characters and normalize whitespace
text = re.sub(r'[^\w\s.,;:!?()-]', ' ', text)
text = re.sub(r'\s+', ' ', text)
# Remove common noise
text = re.sub(r'http\S+', '', text)
text = re.sub(r'www\S+', '', text)
text = re.sub(r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b', '', text)
return text.strip()
def _determine_text_type(self, text: str) -> str:
"""Determine the type of medical text"""
text = text.lower()
if any(term in text for term in ['discharge', 'summary', 'discharge summary']):
return 'discharge_summary'
elif any(term in text for term in ['lab', 'laboratory', 'test results']):
return 'lab_report'
elif any(term in text for term in ['prescription', 'medication', 'drug']):
return 'prescription'
elif any(term in text for term in ['question', 'answer', 'qa']):
return 'medical_qa'
else:
return 'clinical_note'
def generate_report(self) -> Dict:
"""Generate a report of the data collection process"""
# Convert all datetime objects to strings
for k, v in self.stats.items():
if isinstance(v, datetime):
self.stats[k] = str(v)
self.stats["end_time"] = str(datetime.now())
if isinstance(self.stats["start_time"], datetime):
self.stats["start_time"] = str(self.stats["start_time"])
# Calculate duration as string
try:
start_dt = datetime.fromisoformat(self.stats["start_time"])
end_dt = datetime.fromisoformat(self.stats["end_time"])
self.stats["duration"] = str(end_dt - start_dt)
except Exception:
self.stats["duration"] = "unknown"
report_file = self.output_dir.parent / "reports" / "collection_report.json"
report_file.parent.mkdir(exist_ok=True)
with open(report_file, 'w', encoding='utf-8') as f:
json.dump(self.stats, f, indent=2, ensure_ascii=False)
return self.stats
def main():
"""Run data collection pipeline"""
try:
collector = MedicalDataCollector()
# Collect from Hugging Face
logger.info("Starting Hugging Face dataset collection...")
hf_data = collector.collect_huggingface_datasets()
# Collect from PubMed
logger.info("Starting PubMed collection...")
pubmed_data = collector.collect_pubmed_abstracts()
# Create training dataset
logger.info("Creating training dataset...")
training_df = collector.create_training_dataset()
# Generate report
report = collector.generate_report()
# Print summary
logger.info("\nData Collection Summary:")
logger.info(f"Total samples collected: {report['total_samples']}")
logger.info(f"Final training samples: {report['final_samples']}")
logger.info(f"Duration: {report['duration']}")
logger.info("\nText types distribution:")
for type_, count in report['text_types'].items():
logger.info(f"- {type_}: {count}")
if report['errors']:
logger.warning(f"\nEncountered {len(report['errors'])} errors during collection")
except Exception as e:
logger.error(f"Data collection failed: {str(e)}", exc_info=True)
sys.exit(1)
if __name__ == "__main__":
main() |