Spaces:
Runtime error
Runtime error
Upload 2 files
Browse files- app.py +653 -0
- requirements.txt +26 -0
app.py
ADDED
@@ -0,0 +1,653 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
import csv
|
4 |
+
import asyncio
|
5 |
+
import xml.etree.ElementTree as ET
|
6 |
+
from typing import Any, Dict, Optional, Tuple, Union, List
|
7 |
+
|
8 |
+
import httpx
|
9 |
+
import gradio as gr
|
10 |
+
import torch
|
11 |
+
from dotenv import load_dotenv
|
12 |
+
from loguru import logger
|
13 |
+
from huggingface_hub import login
|
14 |
+
from openai import OpenAI
|
15 |
+
from reportlab.pdfgen import canvas
|
16 |
+
from transformers import (
|
17 |
+
AutoTokenizer,
|
18 |
+
AutoModelForSequenceClassification,
|
19 |
+
MarianMTModel,
|
20 |
+
MarianTokenizer,
|
21 |
+
)
|
22 |
+
import pandas as pd
|
23 |
+
import altair as alt
|
24 |
+
import spacy
|
25 |
+
import spacy.cli
|
26 |
+
import PyPDF2 # For PDF reading
|
27 |
+
|
28 |
+
# Ensure spaCy model is downloaded
|
29 |
+
try:
|
30 |
+
nlp = spacy.load("en_core_web_sm")
|
31 |
+
except OSError:
|
32 |
+
logger.info("Downloading SpaCy 'en_core_web_sm' model...")
|
33 |
+
spacy.cli.download("en_core_web_sm")
|
34 |
+
nlp = spacy.load("en_core_web_sm")
|
35 |
+
|
36 |
+
# Logging
|
37 |
+
logger.add("error_logs.log", rotation="1 MB", level="ERROR")
|
38 |
+
|
39 |
+
# Load environment variables
|
40 |
+
load_dotenv()
|
41 |
+
HUGGINGFACE_TOKEN = os.getenv("HF_TOKEN")
|
42 |
+
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
|
43 |
+
ENTREZ_EMAIL = os.getenv("ENTREZ_EMAIL")
|
44 |
+
|
45 |
+
# Basic checks
|
46 |
+
if not HUGGINGFACE_TOKEN or not OPENAI_API_KEY:
|
47 |
+
logger.error("Missing Hugging Face or OpenAI credentials.")
|
48 |
+
raise ValueError("Missing credentials for Hugging Face or OpenAI.")
|
49 |
+
|
50 |
+
# API endpoints
|
51 |
+
PUBMED_SEARCH_URL = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi"
|
52 |
+
PUBMED_FETCH_URL = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi"
|
53 |
+
EUROPE_PMC_BASE_URL = "https://www.ebi.ac.uk/europepmc/webservices/rest/search"
|
54 |
+
|
55 |
+
# Hugging Face login
|
56 |
+
login(HUGGINGFACE_TOKEN)
|
57 |
+
|
58 |
+
# Initialize OpenAI
|
59 |
+
client = OpenAI(api_key=OPENAI_API_KEY)
|
60 |
+
|
61 |
+
# Device setting
|
62 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
63 |
+
logger.info(f"Using device: {device}")
|
64 |
+
|
65 |
+
# Model settings
|
66 |
+
MODEL_NAME = "mgbam/bert-base-finetuned-mgbam"
|
67 |
+
try:
|
68 |
+
model = AutoModelForSequenceClassification.from_pretrained(
|
69 |
+
MODEL_NAME, use_auth_token=HUGGINGFACE_TOKEN
|
70 |
+
).to(device)
|
71 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
72 |
+
MODEL_NAME, use_auth_token=HUGGINGFACE_TOKEN
|
73 |
+
)
|
74 |
+
except Exception as e:
|
75 |
+
logger.error(f"Model load error: {e}")
|
76 |
+
raise
|
77 |
+
|
78 |
+
# Translation model settings
|
79 |
+
try:
|
80 |
+
translation_model_name = "Helsinki-NLP/opus-mt-en-fr"
|
81 |
+
translation_model = MarianMTModel.from_pretrained(
|
82 |
+
translation_model_name, use_auth_token=HUGGINGFACE_TOKEN
|
83 |
+
).to(device)
|
84 |
+
translation_tokenizer = MarianTokenizer.from_pretrained(
|
85 |
+
translation_model_name, use_auth_token=HUGGINGFACE_TOKEN
|
86 |
+
)
|
87 |
+
except Exception as e:
|
88 |
+
logger.error(f"Translation model load error: {e}")
|
89 |
+
raise
|
90 |
+
|
91 |
+
LANGUAGE_MAP: Dict[str, Tuple[str, str]] = {
|
92 |
+
"English to French": ("en", "fr"),
|
93 |
+
"French to English": ("fr", "en"),
|
94 |
+
}
|
95 |
+
|
96 |
+
### Utility Functions ###
|
97 |
+
def safe_json_parse(text: str) -> Union[Dict, None]:
|
98 |
+
"""Safely parse JSON string into a Python dictionary."""
|
99 |
+
try:
|
100 |
+
return json.loads(text)
|
101 |
+
except json.JSONDecodeError as e:
|
102 |
+
logger.error(f"JSON parsing error: {e}")
|
103 |
+
return None
|
104 |
+
|
105 |
+
def parse_pubmed_xml(xml_data: str) -> List[Dict[str, Any]]:
|
106 |
+
"""Parses PubMed XML data and returns a list of structured articles."""
|
107 |
+
root = ET.fromstring(xml_data)
|
108 |
+
articles = []
|
109 |
+
for article in root.findall(".//PubmedArticle"):
|
110 |
+
pmid = article.findtext(".//PMID")
|
111 |
+
title = article.findtext(".//ArticleTitle")
|
112 |
+
abstract = article.findtext(".//AbstractText")
|
113 |
+
journal = article.findtext(".//Journal/Title")
|
114 |
+
pub_date_elem = article.find(".//JournalIssue/PubDate")
|
115 |
+
pub_date = None
|
116 |
+
if pub_date_elem is not None:
|
117 |
+
year = pub_date_elem.findtext("Year")
|
118 |
+
month = pub_date_elem.findtext("Month")
|
119 |
+
day = pub_date_elem.findtext("Day")
|
120 |
+
if year and month and day:
|
121 |
+
pub_date = f"{year}-{month}-{day}"
|
122 |
+
else:
|
123 |
+
pub_date = year
|
124 |
+
articles.append({
|
125 |
+
"PMID": pmid,
|
126 |
+
"Title": title,
|
127 |
+
"Abstract": abstract,
|
128 |
+
"Journal": journal,
|
129 |
+
"PublicationDate": pub_date,
|
130 |
+
})
|
131 |
+
return articles
|
132 |
+
|
133 |
+
### Async Functions for Europe PMC ###
|
134 |
+
async def fetch_articles_by_nct_id(nct_id: str) -> Dict[str, Any]:
|
135 |
+
params = {"query": nct_id, "format": "json"}
|
136 |
+
async with httpx.AsyncClient() as client_http:
|
137 |
+
try:
|
138 |
+
response = await client_http.get(EUROPE_PMC_BASE_URL, params=params)
|
139 |
+
response.raise_for_status()
|
140 |
+
return response.json()
|
141 |
+
except Exception as e:
|
142 |
+
logger.error(f"Error fetching articles for {nct_id}: {e}")
|
143 |
+
return {"error": str(e)}
|
144 |
+
|
145 |
+
async def fetch_articles_by_query(query_params: str) -> Dict[str, Any]:
|
146 |
+
parsed_params = safe_json_parse(query_params)
|
147 |
+
if not parsed_params or not isinstance(parsed_params, dict):
|
148 |
+
return {"error": "Invalid JSON."}
|
149 |
+
query_string = " AND ".join(f"{k}:{v}" for k, v in parsed_params.items())
|
150 |
+
params = {"query": query_string, "format": "json"}
|
151 |
+
async with httpx.AsyncClient() as client_http:
|
152 |
+
try:
|
153 |
+
response = await client_http.get(EUROPE_PMC_BASE_URL, params=params)
|
154 |
+
response.raise_for_status()
|
155 |
+
return response.json()
|
156 |
+
except Exception as e:
|
157 |
+
logger.error(f"Error fetching articles: {e}")
|
158 |
+
return {"error": str(e)}
|
159 |
+
|
160 |
+
### PubMed Integration ###
|
161 |
+
async def fetch_pubmed_by_query(query_params: str) -> Dict[str, Any]:
|
162 |
+
parsed_params = safe_json_parse(query_params)
|
163 |
+
if not parsed_params or not isinstance(parsed_params, dict):
|
164 |
+
return {"error": "Invalid JSON for PubMed."}
|
165 |
+
|
166 |
+
search_params = {
|
167 |
+
"db": "pubmed",
|
168 |
+
"retmode": "json",
|
169 |
+
"email": ENTREZ_EMAIL,
|
170 |
+
"retmax": parsed_params.get("retmax", "10"),
|
171 |
+
"term": parsed_params.get("term", ""),
|
172 |
+
}
|
173 |
+
|
174 |
+
async with httpx.AsyncClient() as client_http:
|
175 |
+
try:
|
176 |
+
search_response = await client_http.get(PUBMED_SEARCH_URL, params=search_params)
|
177 |
+
search_response.raise_for_status()
|
178 |
+
search_data = search_response.json()
|
179 |
+
id_list = search_data.get("esearchresult", {}).get("idlist", [])
|
180 |
+
if not id_list:
|
181 |
+
return {"result": ""}
|
182 |
+
|
183 |
+
fetch_params = {
|
184 |
+
"db": "pubmed",
|
185 |
+
"id": ",".join(id_list),
|
186 |
+
"retmode": "xml",
|
187 |
+
"email": ENTREZ_EMAIL,
|
188 |
+
}
|
189 |
+
fetch_response = await client_http.get(PUBMED_FETCH_URL, params=fetch_params)
|
190 |
+
fetch_response.raise_for_status()
|
191 |
+
return {"result": fetch_response.text}
|
192 |
+
except Exception as e:
|
193 |
+
logger.error(f"Error fetching PubMed articles: {e}")
|
194 |
+
return {"error": str(e)}
|
195 |
+
|
196 |
+
### Crossref Integration ###
|
197 |
+
async def fetch_crossref_by_query(query_params: str) -> Dict[str, Any]:
|
198 |
+
parsed_params = safe_json_parse(query_params)
|
199 |
+
if not parsed_params or not isinstance(parsed_params, dict):
|
200 |
+
return {"error": "Invalid JSON for Crossref."}
|
201 |
+
CROSSREF_API_URL = "https://api.crossref.org/works"
|
202 |
+
async with httpx.AsyncClient() as client_http:
|
203 |
+
try:
|
204 |
+
response = await client_http.get(CROSSREF_API_URL, params=parsed_params)
|
205 |
+
response.raise_for_status()
|
206 |
+
return response.json()
|
207 |
+
except Exception as e:
|
208 |
+
logger.error(f"Error fetching Crossref data: {e}")
|
209 |
+
return {"error": str(e)}
|
210 |
+
|
211 |
+
### Core Functions ###
|
212 |
+
def summarize_text(text: str) -> str:
|
213 |
+
"""Summarize text using OpenAI."""
|
214 |
+
if not text.strip():
|
215 |
+
return "No text provided for summarization."
|
216 |
+
try:
|
217 |
+
response = client.chat.completions.create(
|
218 |
+
model="gpt-3.5-turbo",
|
219 |
+
messages=[{"role": "user", "content": f"Summarize the following clinical data:\n{text}"}],
|
220 |
+
max_tokens=200,
|
221 |
+
temperature=0.7,
|
222 |
+
)
|
223 |
+
return response.choices[0].message.content.strip()
|
224 |
+
except Exception as e:
|
225 |
+
logger.error(f"Summarization Error: {e}")
|
226 |
+
return "Summarization failed."
|
227 |
+
|
228 |
+
def predict_outcome(text: str) -> Union[Dict[str, float], str]:
|
229 |
+
"""Predict outcomes (classification) using a fine-tuned model."""
|
230 |
+
if not text.strip():
|
231 |
+
return "No text provided for prediction."
|
232 |
+
try:
|
233 |
+
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
|
234 |
+
inputs = {k: v.to(device) for k, v in inputs.items()}
|
235 |
+
with torch.no_grad():
|
236 |
+
outputs = model(**inputs)
|
237 |
+
probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)[0]
|
238 |
+
return {f"Label {i+1}": float(prob.item()) for i, prob in enumerate(probabilities)}
|
239 |
+
except Exception as e:
|
240 |
+
logger.error(f"Prediction Error: {e}")
|
241 |
+
return "Prediction failed."
|
242 |
+
|
243 |
+
def generate_report(text: str, filename: str = "clinical_report.pdf") -> Optional[str]:
|
244 |
+
"""Generate a PDF report from the given text."""
|
245 |
+
try:
|
246 |
+
if not text.strip():
|
247 |
+
logger.warning("No text provided for the report.")
|
248 |
+
c = canvas.Canvas(filename)
|
249 |
+
c.drawString(100, 750, "Clinical Research Report")
|
250 |
+
lines = text.split("\n")
|
251 |
+
y = 730
|
252 |
+
for line in lines:
|
253 |
+
if y < 50:
|
254 |
+
c.showPage()
|
255 |
+
y = 750
|
256 |
+
c.drawString(100, y, line)
|
257 |
+
y -= 15
|
258 |
+
c.save()
|
259 |
+
logger.info(f"Report generated: {filename}")
|
260 |
+
return filename
|
261 |
+
except Exception as e:
|
262 |
+
logger.error(f"Report Generation Error: {e}")
|
263 |
+
return None
|
264 |
+
|
265 |
+
def visualize_predictions(predictions: Dict[str, float]) -> Optional[alt.Chart]:
|
266 |
+
"""Visualize model prediction probabilities using Altair."""
|
267 |
+
try:
|
268 |
+
data = pd.DataFrame(list(predictions.items()), columns=["Label", "Probability"])
|
269 |
+
chart = (
|
270 |
+
alt.Chart(data)
|
271 |
+
.mark_bar()
|
272 |
+
.encode(
|
273 |
+
x=alt.X("Label:N", sort=None),
|
274 |
+
y="Probability:Q",
|
275 |
+
tooltip=["Label", "Probability"],
|
276 |
+
)
|
277 |
+
.properties(title="Prediction Probabilities", width=500, height=300)
|
278 |
+
)
|
279 |
+
return chart
|
280 |
+
except Exception as e:
|
281 |
+
logger.error(f"Visualization Error: {e}")
|
282 |
+
return None
|
283 |
+
|
284 |
+
def translate_text(text: str, translation_option: str) -> str:
|
285 |
+
"""Translate text between English and French."""
|
286 |
+
if not text.strip():
|
287 |
+
return "No text provided for translation."
|
288 |
+
try:
|
289 |
+
if translation_option not in LANGUAGE_MAP:
|
290 |
+
return "Unsupported translation option."
|
291 |
+
inputs = translation_tokenizer(text, return_tensors="pt", padding=True).to(device)
|
292 |
+
translated_tokens = translation_model.generate(**inputs)
|
293 |
+
return translation_tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
|
294 |
+
except Exception as e:
|
295 |
+
logger.error(f"Translation Error: {e}")
|
296 |
+
return "Translation failed."
|
297 |
+
|
298 |
+
def perform_named_entity_recognition(text: str) -> str:
|
299 |
+
"""Perform Named Entity Recognition (NER) using spaCy."""
|
300 |
+
if not text.strip():
|
301 |
+
return "No text provided for NER."
|
302 |
+
try:
|
303 |
+
doc = nlp(text)
|
304 |
+
entities = [(ent.text, ent.label_) for ent in doc.ents]
|
305 |
+
if not entities:
|
306 |
+
return "No named entities found."
|
307 |
+
return "\n".join(f"{ent_text} -> {ent_label}" for ent_text, ent_label in entities)
|
308 |
+
except Exception as e:
|
309 |
+
logger.error(f"NER Error: {e}")
|
310 |
+
return "Named Entity Recognition failed."
|
311 |
+
|
312 |
+
### Enhanced EDA ###
|
313 |
+
def perform_enhanced_eda(df: pd.DataFrame) -> Tuple[str, Optional[alt.Chart], Optional[alt.Chart]]:
|
314 |
+
"""
|
315 |
+
Perform a more advanced EDA given a DataFrame:
|
316 |
+
- Show dataset info (columns, shape, numeric summary).
|
317 |
+
- Generate a correlation heatmap (for numeric columns).
|
318 |
+
- Generate distribution plots (histograms) for numeric columns.
|
319 |
+
Returns (text_summary, correlation_chart, distribution_chart).
|
320 |
+
"""
|
321 |
+
try:
|
322 |
+
# Basic info
|
323 |
+
columns_info = f"Columns: {list(df.columns)}"
|
324 |
+
shape_info = f"Shape: {df.shape[0]} rows x {df.shape[1]} columns"
|
325 |
+
|
326 |
+
# Use describe with "include='all'" to show all columns summary
|
327 |
+
with pd.option_context("display.max_colwidth", 200, "display.max_rows", None):
|
328 |
+
describe_info = df.describe(include="all").to_string()
|
329 |
+
|
330 |
+
summary_text = (
|
331 |
+
f"--- Enhanced EDA Summary ---\n"
|
332 |
+
f"{columns_info}\n{shape_info}\n\n"
|
333 |
+
f"Summary Statistics:\n{describe_info}\n"
|
334 |
+
)
|
335 |
+
|
336 |
+
# Correlation heatmap
|
337 |
+
numeric_cols = df.select_dtypes(include="number")
|
338 |
+
corr_chart = None
|
339 |
+
if numeric_cols.shape[1] >= 2:
|
340 |
+
corr = numeric_cols.corr()
|
341 |
+
corr_melted = corr.reset_index().melt(id_vars="index")
|
342 |
+
corr_melted.columns = ["Feature1", "Feature2", "Correlation"]
|
343 |
+
corr_chart = (
|
344 |
+
alt.Chart(corr_melted)
|
345 |
+
.mark_rect()
|
346 |
+
.encode(
|
347 |
+
x="Feature1:O",
|
348 |
+
y="Feature2:O",
|
349 |
+
color="Correlation:Q",
|
350 |
+
tooltip=["Feature1", "Feature2", "Correlation"]
|
351 |
+
)
|
352 |
+
.properties(width=400, height=400, title="Correlation Heatmap")
|
353 |
+
)
|
354 |
+
|
355 |
+
# Distribution plots (histograms) for numeric columns
|
356 |
+
distribution_chart = None
|
357 |
+
if numeric_cols.shape[1] >= 1:
|
358 |
+
df_long = numeric_cols.melt(var_name='Column', value_name='Value')
|
359 |
+
distribution_chart = (
|
360 |
+
alt.Chart(df_long)
|
361 |
+
.mark_bar()
|
362 |
+
.encode(
|
363 |
+
alt.X("Value:Q", bin=alt.Bin(maxbins=30)),
|
364 |
+
alt.Y('count()'),
|
365 |
+
alt.Facet('Column:N', columns=2),
|
366 |
+
tooltip=["Value"]
|
367 |
+
)
|
368 |
+
.properties(
|
369 |
+
title='Distribution of Numeric Columns',
|
370 |
+
width=300,
|
371 |
+
height=200
|
372 |
+
)
|
373 |
+
.interactive()
|
374 |
+
)
|
375 |
+
|
376 |
+
return summary_text, corr_chart, distribution_chart
|
377 |
+
|
378 |
+
except Exception as e:
|
379 |
+
logger.error(f"Enhanced EDA Error: {e}")
|
380 |
+
return f"Enhanced EDA failed: {e}", None, None
|
381 |
+
|
382 |
+
### File Handling ###
|
383 |
+
def read_uploaded_file(uploaded_file: Optional[gr.File]) -> str:
|
384 |
+
"""
|
385 |
+
Reads the content of an uploaded file (txt, csv, xls, xlsx, pdf).
|
386 |
+
Returns the extracted text or CSV-like content.
|
387 |
+
"""
|
388 |
+
if uploaded_file is None:
|
389 |
+
return ""
|
390 |
+
|
391 |
+
file_name = uploaded_file.name
|
392 |
+
file_ext = os.path.splitext(file_name)[1].lower()
|
393 |
+
|
394 |
+
try:
|
395 |
+
# For text
|
396 |
+
if file_ext == ".txt":
|
397 |
+
return uploaded_file.read().decode("utf-8")
|
398 |
+
|
399 |
+
# For CSV
|
400 |
+
elif file_ext == ".csv":
|
401 |
+
return uploaded_file.read().decode("utf-8")
|
402 |
+
|
403 |
+
# For Excel
|
404 |
+
elif file_ext in [".xls", ".xlsx"]:
|
405 |
+
# We'll just return empty here and parse it later into a DataFrame
|
406 |
+
# because we can read the binary directly into pd.read_excel().
|
407 |
+
# Or store as bytes for later use in EDA.
|
408 |
+
return "EXCEL_FILE_PLACEHOLDER" # We'll handle it differently in EDA step
|
409 |
+
|
410 |
+
# For PDF
|
411 |
+
elif file_ext == ".pdf":
|
412 |
+
pdf_reader = PyPDF2.PdfReader(uploaded_file)
|
413 |
+
text_content = []
|
414 |
+
for page in pdf_reader.pages:
|
415 |
+
text_content.append(page.extract_text())
|
416 |
+
return "\n".join(text_content)
|
417 |
+
|
418 |
+
else:
|
419 |
+
return f"Unsupported file format: {file_ext}"
|
420 |
+
except Exception as e:
|
421 |
+
logger.error(f"File read error: {e}")
|
422 |
+
return f"Error reading file: {e}"
|
423 |
+
|
424 |
+
def parse_excel_file(uploaded_file) -> pd.DataFrame:
|
425 |
+
"""
|
426 |
+
Parse an Excel file into a pandas DataFrame.
|
427 |
+
We assume the user wants the first sheet or we can guess.
|
428 |
+
"""
|
429 |
+
try:
|
430 |
+
# For Excel, we can do:
|
431 |
+
df = pd.read_excel(uploaded_file, engine="openpyxl")
|
432 |
+
return df
|
433 |
+
except Exception as e:
|
434 |
+
logger.error(f"Excel parsing error: {e}")
|
435 |
+
raise
|
436 |
+
|
437 |
+
def parse_csv_content(csv_content: str) -> pd.DataFrame:
|
438 |
+
"""
|
439 |
+
Attempt to parse CSV content with both utf-8 and utf-8-sig to handle BOM issues.
|
440 |
+
"""
|
441 |
+
from io import StringIO
|
442 |
+
errors = []
|
443 |
+
for encoding_try in ["utf-8", "utf-8-sig"]:
|
444 |
+
try:
|
445 |
+
df = pd.read_csv(StringIO(csv_content), encoding=encoding_try)
|
446 |
+
return df
|
447 |
+
except Exception as e:
|
448 |
+
errors.append(f"Encoding {encoding_try} failed: {e}")
|
449 |
+
error_msg = "Could not parse CSV content.\n" + "\n".join(errors)
|
450 |
+
logger.error(error_msg)
|
451 |
+
raise ValueError(error_msg)
|
452 |
+
|
453 |
+
### Gradio Interface ###
|
454 |
+
with gr.Blocks() as demo:
|
455 |
+
gr.Markdown("# ✨ Advanced Clinical Research Assistant with Enhanced EDA ✨")
|
456 |
+
gr.Markdown("""
|
457 |
+
Welcome to the **Enhanced** AI-Powered Clinical Assistant!
|
458 |
+
- **Summarize** large blocks of clinical text.
|
459 |
+
- **Predict** outcomes with a fine-tuned model.
|
460 |
+
- **Translate** text between English & French.
|
461 |
+
- **Perform Named Entity Recognition** with spaCy.
|
462 |
+
- **Fetch** from PubMed, Crossref, Europe PMC.
|
463 |
+
- **Generate** professional PDF reports.
|
464 |
+
- **Perform Enhanced EDA** on CSV/Excel data with correlation heatmaps & distribution plots.
|
465 |
+
""")
|
466 |
+
|
467 |
+
# Inputs
|
468 |
+
with gr.Row():
|
469 |
+
text_input = gr.Textbox(label="Input Text", lines=5, placeholder="Enter clinical text or query...")
|
470 |
+
file_input = gr.File(
|
471 |
+
label="Upload File (txt/csv/xls/xlsx/pdf)",
|
472 |
+
file_types=[".txt", ".csv", ".xls", ".xlsx", ".pdf"]
|
473 |
+
)
|
474 |
+
|
475 |
+
action = gr.Radio(
|
476 |
+
[
|
477 |
+
"Summarize",
|
478 |
+
"Predict Outcome",
|
479 |
+
"Generate Report",
|
480 |
+
"Translate",
|
481 |
+
"Perform Named Entity Recognition",
|
482 |
+
"Perform Enhanced EDA",
|
483 |
+
"Fetch Clinical Studies",
|
484 |
+
"Fetch PubMed Articles (Legacy)",
|
485 |
+
"Fetch PubMed by Query",
|
486 |
+
"Fetch Crossref by Query",
|
487 |
+
],
|
488 |
+
label="Select an Action",
|
489 |
+
)
|
490 |
+
translation_option = gr.Dropdown(
|
491 |
+
choices=list(LANGUAGE_MAP.keys()),
|
492 |
+
label="Translation Option",
|
493 |
+
value="English to French"
|
494 |
+
)
|
495 |
+
query_params_input = gr.Textbox(
|
496 |
+
label="Query Parameters (JSON Format)",
|
497 |
+
placeholder='{"term": "cancer", "retmax": "5"}'
|
498 |
+
)
|
499 |
+
nct_id_input = gr.Textbox(label="NCT ID for Article Search")
|
500 |
+
report_filename_input = gr.Textbox(
|
501 |
+
label="Report Filename",
|
502 |
+
placeholder="clinical_report.pdf",
|
503 |
+
value="clinical_report.pdf"
|
504 |
+
)
|
505 |
+
export_format = gr.Dropdown(["None", "CSV", "JSON"], label="Export Format")
|
506 |
+
|
507 |
+
# Outputs
|
508 |
+
output_text = gr.Textbox(label="Output", lines=10)
|
509 |
+
|
510 |
+
with gr.Row():
|
511 |
+
output_chart = gr.Plot(label="Visualization 1")
|
512 |
+
output_chart2 = gr.Plot(label="Visualization 2")
|
513 |
+
|
514 |
+
output_file = gr.File(label="Generated File")
|
515 |
+
|
516 |
+
submit_button = gr.Button("Submit")
|
517 |
+
|
518 |
+
# Async function for handling actions
|
519 |
+
async def handle_action(
|
520 |
+
action: str,
|
521 |
+
text: str,
|
522 |
+
file_up: gr.File,
|
523 |
+
translation_opt: str,
|
524 |
+
query_params: str,
|
525 |
+
nct_id: str,
|
526 |
+
report_filename: str,
|
527 |
+
export_format: str
|
528 |
+
) -> Tuple[Optional[str], Optional[Any], Optional[Any], Optional[str]]:
|
529 |
+
|
530 |
+
# Read the uploaded file
|
531 |
+
file_content = read_uploaded_file(file_up)
|
532 |
+
combined_text = (text + "\n" + file_content).strip() if file_content else text
|
533 |
+
|
534 |
+
# Branch by action
|
535 |
+
if action == "Summarize":
|
536 |
+
return summarize_text(combined_text), None, None, None
|
537 |
+
|
538 |
+
elif action == "Predict Outcome":
|
539 |
+
predictions = predict_outcome(combined_text)
|
540 |
+
if isinstance(predictions, dict):
|
541 |
+
chart = visualize_predictions(predictions)
|
542 |
+
return json.dumps(predictions, indent=2), chart, None, None
|
543 |
+
return predictions, None, None, None
|
544 |
+
|
545 |
+
elif action == "Generate Report":
|
546 |
+
file_path = generate_report(combined_text, filename=report_filename)
|
547 |
+
msg = f"Report generated: {file_path}" if file_path else "Report generation failed."
|
548 |
+
return msg, None, None, file_path
|
549 |
+
|
550 |
+
elif action == "Translate":
|
551 |
+
return translate_text(combined_text, translation_opt), None, None, None
|
552 |
+
|
553 |
+
elif action == "Perform Named Entity Recognition":
|
554 |
+
ner_result = perform_named_entity_recognition(combined_text)
|
555 |
+
return ner_result, None, None, None
|
556 |
+
|
557 |
+
elif action == "Perform Enhanced EDA":
|
558 |
+
# We expect the user to either upload a CSV or Excel, or paste CSV content.
|
559 |
+
if file_up is None and not combined_text:
|
560 |
+
return "No data provided for EDA.", None, None, None
|
561 |
+
|
562 |
+
# If Excel was uploaded
|
563 |
+
if file_up and file_up.name.lower().endswith((".xls", ".xlsx")):
|
564 |
+
try:
|
565 |
+
df_excel = parse_excel_file(file_up)
|
566 |
+
eda_summary, corr_chart, dist_chart = perform_enhanced_eda(df_excel)
|
567 |
+
return eda_summary, corr_chart, dist_chart, None
|
568 |
+
except Exception as e:
|
569 |
+
return f"Excel EDA failed: {e}", None, None, None
|
570 |
+
|
571 |
+
# If CSV was uploaded
|
572 |
+
if file_up and file_up.name.lower().endswith(".csv"):
|
573 |
+
try:
|
574 |
+
df_csv = parse_csv_content(file_content)
|
575 |
+
eda_summary, corr_chart, dist_chart = perform_enhanced_eda(df_csv)
|
576 |
+
return eda_summary, corr_chart, dist_chart, None
|
577 |
+
except Exception as e:
|
578 |
+
return f"CSV EDA failed: {e}", None, None, None
|
579 |
+
|
580 |
+
# If user just pasted CSV content (no file)
|
581 |
+
if not file_up and "," in combined_text:
|
582 |
+
try:
|
583 |
+
df_csv = parse_csv_content(combined_text)
|
584 |
+
eda_summary, corr_chart, dist_chart = perform_enhanced_eda(df_csv)
|
585 |
+
return eda_summary, corr_chart, dist_chart, None
|
586 |
+
except Exception as e:
|
587 |
+
return f"CSV EDA failed: {e}", None, None, None
|
588 |
+
|
589 |
+
# Otherwise, not supported
|
590 |
+
return "No valid CSV/Excel data found for EDA.", None, None, None
|
591 |
+
|
592 |
+
elif action == "Fetch Clinical Studies":
|
593 |
+
if nct_id:
|
594 |
+
result = await fetch_articles_by_nct_id(nct_id)
|
595 |
+
elif query_params:
|
596 |
+
result = await fetch_articles_by_query(query_params)
|
597 |
+
else:
|
598 |
+
return "Provide either an NCT ID or valid query parameters.", None, None, None
|
599 |
+
|
600 |
+
articles = result.get("resultList", {}).get("result", [])
|
601 |
+
if not articles:
|
602 |
+
return "No articles found.", None, None, None
|
603 |
+
|
604 |
+
formatted_results = "\n\n".join(
|
605 |
+
f"Title: {a.get('title')}\nJournal: {a.get('journalTitle')} ({a.get('pubYear')})"
|
606 |
+
for a in articles
|
607 |
+
)
|
608 |
+
return formatted_results, None, None, None
|
609 |
+
|
610 |
+
elif action in ["Fetch PubMed Articles (Legacy)", "Fetch PubMed by Query"]:
|
611 |
+
pubmed_result = await fetch_pubmed_by_query(query_params)
|
612 |
+
xml_data = pubmed_result.get("result")
|
613 |
+
if xml_data:
|
614 |
+
articles = parse_pubmed_xml(xml_data)
|
615 |
+
if not articles:
|
616 |
+
return "No articles found.", None, None, None
|
617 |
+
formatted = "\n\n".join(
|
618 |
+
f"{a['Title']} - {a['Journal']} ({a['PublicationDate']})"
|
619 |
+
for a in articles if a['Title']
|
620 |
+
)
|
621 |
+
return formatted if formatted else "No articles found.", None, None, None
|
622 |
+
return "No articles found or error fetching data.", None, None, None
|
623 |
+
|
624 |
+
elif action == "Fetch Crossref by Query":
|
625 |
+
crossref_result = await fetch_crossref_by_query(query_params)
|
626 |
+
items = crossref_result.get("message", {}).get("items", [])
|
627 |
+
if not items:
|
628 |
+
return "No results found.", None, None, None
|
629 |
+
formatted = "\n\n".join(
|
630 |
+
f"Title: {item.get('title', ['No title'])[0]}, DOI: {item.get('DOI')}"
|
631 |
+
for item in items
|
632 |
+
)
|
633 |
+
return formatted, None, None, None
|
634 |
+
|
635 |
+
return "Invalid action.", None, None, None
|
636 |
+
|
637 |
+
submit_button.click(
|
638 |
+
handle_action,
|
639 |
+
inputs=[
|
640 |
+
action,
|
641 |
+
text_input,
|
642 |
+
file_input,
|
643 |
+
translation_option,
|
644 |
+
query_params_input,
|
645 |
+
nct_id_input,
|
646 |
+
report_filename_input,
|
647 |
+
export_format,
|
648 |
+
],
|
649 |
+
outputs=[output_text, output_chart, output_chart2, output_file],
|
650 |
+
)
|
651 |
+
|
652 |
+
# Launch the Gradio app
|
653 |
+
demo.launch(server_name="0.0.0.0", server_port=7860, share=True)
|
requirements.txt
ADDED
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
streamlit
|
2 |
+
gradio
|
3 |
+
openai>=0.27.8
|
4 |
+
torch>=2.0.0
|
5 |
+
transformers>=4.33.0
|
6 |
+
huggingface-hub>=0.16.0
|
7 |
+
python-dotenv>=1.0.0
|
8 |
+
reportlab>=3.6.0
|
9 |
+
matplotlib>=3.7.1
|
10 |
+
pandas>=2.0.3
|
11 |
+
altair>=4.2.2
|
12 |
+
loguru>=0.7.0
|
13 |
+
spacy>=3.6.0
|
14 |
+
PyPDF2>=3.0.0
|
15 |
+
pdfplumber>=0.9.0
|
16 |
+
Pillow>=10.0.0
|
17 |
+
sentencepiece
|
18 |
+
sacremoses>=0.0.53
|
19 |
+
httpx
|
20 |
+
numpy
|
21 |
+
reportlab
|
22 |
+
requests
|
23 |
+
openpyxl
|
24 |
+
|
25 |
+
|
26 |
+
|