Spaces:
Runtime error
Runtime error
File size: 10,067 Bytes
3cd3c35 |
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 |
import gradio as gr
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import json
import re
import pandas as pd
from datetime import datetime
import plotly.express as px
from io import StringIO
# Load text model
text_model_name = "microsoft/BiomedVLP-CXR-BERT-specialized"
text_tokenizer = AutoTokenizer.from_pretrained(text_model_name)
text_model = AutoModelForSequenceClassification.from_pretrained(text_model_name)
# Load image model
image_model_name = "aehrc/cxrmate-tf" # Replace with skin disease model if needed
image_processor = AutoImageProcessor.from_pretrained(image_model_name)
image_model = AutoModelForImageClassification.from_pretrained(image_model_name)
# Define labels
text_labels = ["Positive", "Negative", "Neutral", "Informative"] # For text analysis
image_labels = ["Normal", "Abnormal"] # For X-ray or skin images
# Store conversation state
conversation_state = {
"history": [],
"texts": [],
"image_uploaded": False,
"last_analysis": None,
"analysis_log": []
}
# Extract key terms
def extract_key_terms(text):
terms = re.findall(r'\b(fever|cough|fatigue|headache|sore throat|chest pain|shortness of breath|rash|lesion|study|treatment|trial|astronaut|microgravity)\b', text, re.IGNORECASE)
return terms
# Generate context-aware follow-up questions
def generate_follow_up(terms, history):
if not terms:
return "Please provide medical text (e.g., symptoms, abstract) or upload an image."
if "astronaut" in [t.lower() for t in terms] or "microgravity" in [t.lower() for t in terms]:
return "Are you researching space medicine? Please describe physiological data or symptoms in microgravity."
if len(terms) < 3:
return "Can you provide more details (e.g., duration of symptoms or study context)?"
if not conversation_state["image_uploaded"]:
return "Would you like to upload an image (e.g., X-ray or skin photo) for analysis?"
return "Would you like to analyze another text or image, or export the analysis log?"
# Main analysis function
def analyze_medical_input(user_input, image=None, chat_history=None, export_format="None"):
global conversation_state
if not chat_history:
chat_history = []
# Process text input
text_response = ""
text_chart = ""
if user_input.strip():
terms = extract_key_terms(user_input)
conversation_state["texts"].extend(terms)
inputs = text_tokenizer(user_input, return_tensors="pt", padding=True, truncation=True, max_length=512)
with torch.no_grad():
outputs = text_model(**inputs)
logits = outputs.logits
predicted_class_idx = logits.argmax(-1).item()
confidence = torch.softmax(logits, dim=-1)[0][predicted_class_idx].item()
scores = torch.softmax(logits, dim=-1)[0].tolist()
conversation_state["last_analysis"] = {
"type": "text",
"label": text_labels[predicted_class_idx],
"confidence": confidence,
"scores": scores,
"input": user_input,
"timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S")
}
conversation_state["analysis_log"].append(conversation_state["last_analysis"])
text_response = f"Text Analysis: {text_labels[predicted_class_idx]} (Confidence: {confidence:.2%})"
# Text visualization (Chart.js)
chart_data = {
"type": "bar",
"data": {
"labels": text_labels,
"datasets": [{
"label": "Confidence Scores",
"data": scores,
"backgroundColor": ["#4CAF50", "#F44336", "#2196F3", "#FF9800"],
"borderColor": ["#388E3C", "#D32F2F", "#1976D2", "#F57C00"],
"borderWidth": 1
}]
},
"options": {
"scales": {
"y": {"beginAtZero": True, "max": 1, "title": {"display": True, "text": "Confidence"}},
"x": {"title": {"display": True, "text": "Text Categories"}}
},
"plugins": {"title": {"display": True, "text": "Text Analysis Confidence"}}
}
}
text_chart = f"""
<canvas id='textChart' width='400' height='200'></canvas>
<script src='https://cdn.jsdelivr.net/npm/chart.js'></script>
<script>
new Chart(document.getElementById('textChart'), {json.dumps(chart_data)});
</script>
"""
# Process image input
image_response = ""
image_chart = ""
if image is not None:
conversation_state["image_uploaded"] = True
inputs = image_processor(images=image, return_tensors="pt")
with torch.no_grad():
outputs = image_model(**inputs)
logits = outputs.logits
predicted_class_idx = logits.argmax(-1).item()
confidence = torch.softmax(logits, dim=-1)[0][predicted_class_idx].item()
scores = torch.softmax(logits, dim=-1)[0].tolist()
conversation_state["last_analysis"] = {
"type": "image",
"label": image_labels[predicted_class_idx],
"confidence": confidence,
"scores": scores,
"input": "image",
"timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S")
}
conversation_state["analysis_log"].append(conversation_state["last_analysis"])
image_response = f"Image Analysis: {image_labels[predicted_class_idx]} (Confidence: {confidence:.2%})"
# Image visualization (Chart.js)
chart_data = {
"type": "bar",
"data": {
"labels": image_labels,
"datasets": [{
"label": "Confidence Scores",
"data": scores,
"backgroundColor": ["#4CAF50", "#F44336"],
"borderColor": ["#388E3C", "#D32F2F"],
"borderWidth": 1
}]
},
"options": {
"scales": {
"y": {"beginAtZero": True, "max": 1, "title": {"display": True, "text": "Confidence"}},
"x": {"title": {"display": True, "text": "Image Categories"}}
},
"plugins": {"title": {"display": True, "text": "Image Analysis Confidence"}}
}
}
image_chart = f"""
<canvas id='imageChart' width='400' height='200'></canvas>
<script src='https://cdn.jsdelivr.net/npm/chart.js'></script>
<script>
new Chart(document.getElementById('imageChart'), {json.dumps(chart_data)});
</script>
"""
# Generate trend visualization (Plotly)
trend_html = ""
if len(conversation_state["analysis_log"]) > 1:
df = pd.DataFrame(conversation_state["analysis_log"])
fig = px.line(
df, x="timestamp", y="confidence", color="type",
title="Analysis Confidence Over Time",
labels={"confidence": "Confidence Score", "timestamp": "Time"}
)
trend_html = fig.to_html(full_html=False)
# Combine responses
response = "\n".join([r for r in [text_response, image_response] if r])
if not response:
response = "No analysis yet. Please provide text or upload an image."
response += f"\n\nFollow-Up: {generate_follow_up(conversation_state['texts'], conversation_state['history'])}"
response += f"\n\n{text_chart}\n{image_chart}\n{trend_html}"
# Handle export
if export_format != "None":
df = pd.DataFrame(conversation_state["analysis_log"])
if export_format == "JSON":
export_data = df.to_json(orient="records")
return response, gr.File(value=StringIO(export_data), file_name="analysis_log.json")
elif export_format == "CSV":
export_data = df.to_csv(index=False)
return response, gr.File(value=StringIO(export_data), file_name="analysis_log.csv")
# Add disclaimer
disclaimer = "β οΈ This tool is for research purposes only and does not provide medical diagnoses. Consult a healthcare professional for medical advice."
response += f"\n\n{disclaimer}"
conversation_state["history"].append((user_input, response))
return response
# Custom CSS for professional UI
css = """
body { background-color: #f0f2f5; font-family: 'Segoe UI', Arial, sans-serif; }
.gradio-container { max-width: 900px; margin: auto; padding: 30px; background: white; border-radius: 10px; box-shadow: 0 4px 12px rgba(0,0,0,0.1); }
h1 { color: #1a3c5e; text-align: center; font-size: 2em; }
input, textarea { border-radius: 8px; border: 1px solid #ccc; padding: 10px; }
button { background: linear-gradient(90deg, #3498db, #2980b9); color: white; border-radius: 8px; padding: 12px; font-weight: bold; }
button:hover { background: linear-gradient(90deg, #2980b9, #1a6ea6); }
#export_dropdown { width: 150px; margin-top: 10px; }
"""
# Create Gradio interface
with gr.Blocks(css=css) as iface:
gr.Markdown("# Ultra-Advanced Medical Research Chatbot")
gr.Markdown("Analyze medical texts or images for research purposes. Supports symptom analysis, literature review, or space medicine research. Not for medical diagnosis.")
with gr.Row():
with gr.Column(scale=2):
text_input = gr.Textbox(lines=5, placeholder="Enter symptoms, medical abstract, or space medicine data...")
image_input = gr.Image(type="pil", label="Upload X-ray or Skin Image")
export_dropdown = gr.Dropdown(choices=["None", "JSON", "CSV"], label="Export Log", value="None")
submit_button = gr.Button("Analyze")
with gr.Column(scale=3):
output = gr.HTML()
submit_button.click(
fn=analyze_medical_input,
inputs=[text_input, image_input, gr.State(), export_dropdown],
outputs=[output, gr.File()]
)
# Launch the interface
iface.launch() |