Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,228 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoImageProcessor, AutoModelForImageClassification
|
4 |
+
from PIL import Image
|
5 |
+
import json
|
6 |
+
import re
|
7 |
+
import pandas as pd
|
8 |
+
from datetime import datetime
|
9 |
+
import plotly.express as px
|
10 |
+
from io import StringIO
|
11 |
+
|
12 |
+
# Load text model
|
13 |
+
text_model_name = "microsoft/BiomedVLP-CXR-BERT-specialized"
|
14 |
+
text_tokenizer = AutoTokenizer.from_pretrained(text_model_name)
|
15 |
+
text_model = AutoModelForSequenceClassification.from_pretrained(text_model_name)
|
16 |
+
|
17 |
+
# Load image model
|
18 |
+
image_model_name = "aehrc/cxrmate-tf" # Replace with skin disease model if needed
|
19 |
+
image_processor = AutoImageProcessor.from_pretrained(image_model_name)
|
20 |
+
image_model = AutoModelForImageClassification.from_pretrained(image_model_name)
|
21 |
+
|
22 |
+
# Define labels
|
23 |
+
text_labels = ["Positive", "Negative", "Neutral", "Informative"] # For text analysis
|
24 |
+
image_labels = ["Normal", "Abnormal"] # For X-ray or skin images
|
25 |
+
|
26 |
+
# Store conversation state
|
27 |
+
conversation_state = {
|
28 |
+
"history": [],
|
29 |
+
"texts": [],
|
30 |
+
"image_uploaded": False,
|
31 |
+
"last_analysis": None,
|
32 |
+
"analysis_log": []
|
33 |
+
}
|
34 |
+
|
35 |
+
# Extract key terms
|
36 |
+
def extract_key_terms(text):
|
37 |
+
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)
|
38 |
+
return terms
|
39 |
+
|
40 |
+
# Generate context-aware follow-up questions
|
41 |
+
def generate_follow_up(terms, history):
|
42 |
+
if not terms:
|
43 |
+
return "Please provide medical text (e.g., symptoms, abstract) or upload an image."
|
44 |
+
if "astronaut" in [t.lower() for t in terms] or "microgravity" in [t.lower() for t in terms]:
|
45 |
+
return "Are you researching space medicine? Please describe physiological data or symptoms in microgravity."
|
46 |
+
if len(terms) < 3:
|
47 |
+
return "Can you provide more details (e.g., duration of symptoms or study context)?"
|
48 |
+
if not conversation_state["image_uploaded"]:
|
49 |
+
return "Would you like to upload an image (e.g., X-ray or skin photo) for analysis?"
|
50 |
+
return "Would you like to analyze another text or image, or export the analysis log?"
|
51 |
+
|
52 |
+
# Main analysis function
|
53 |
+
def analyze_medical_input(user_input, image=None, chat_history=None, export_format="None"):
|
54 |
+
global conversation_state
|
55 |
+
if not chat_history:
|
56 |
+
chat_history = []
|
57 |
+
|
58 |
+
# Process text input
|
59 |
+
text_response = ""
|
60 |
+
text_chart = ""
|
61 |
+
if user_input.strip():
|
62 |
+
terms = extract_key_terms(user_input)
|
63 |
+
conversation_state["texts"].extend(terms)
|
64 |
+
inputs = text_tokenizer(user_input, return_tensors="pt", padding=True, truncation=True, max_length=512)
|
65 |
+
with torch.no_grad():
|
66 |
+
outputs = text_model(**inputs)
|
67 |
+
logits = outputs.logits
|
68 |
+
predicted_class_idx = logits.argmax(-1).item()
|
69 |
+
confidence = torch.softmax(logits, dim=-1)[0][predicted_class_idx].item()
|
70 |
+
scores = torch.softmax(logits, dim=-1)[0].tolist()
|
71 |
+
conversation_state["last_analysis"] = {
|
72 |
+
"type": "text",
|
73 |
+
"label": text_labels[predicted_class_idx],
|
74 |
+
"confidence": confidence,
|
75 |
+
"scores": scores,
|
76 |
+
"input": user_input,
|
77 |
+
"timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
78 |
+
}
|
79 |
+
conversation_state["analysis_log"].append(conversation_state["last_analysis"])
|
80 |
+
text_response = f"Text Analysis: {text_labels[predicted_class_idx]} (Confidence: {confidence:.2%})"
|
81 |
+
|
82 |
+
# Text visualization (Chart.js)
|
83 |
+
chart_data = {
|
84 |
+
"type": "bar",
|
85 |
+
"data": {
|
86 |
+
"labels": text_labels,
|
87 |
+
"datasets": [{
|
88 |
+
"label": "Confidence Scores",
|
89 |
+
"data": scores,
|
90 |
+
"backgroundColor": ["#4CAF50", "#F44336", "#2196F3", "#FF9800"],
|
91 |
+
"borderColor": ["#388E3C", "#D32F2F", "#1976D2", "#F57C00"],
|
92 |
+
"borderWidth": 1
|
93 |
+
}]
|
94 |
+
},
|
95 |
+
"options": {
|
96 |
+
"scales": {
|
97 |
+
"y": {"beginAtZero": True, "max": 1, "title": {"display": True, "text": "Confidence"}},
|
98 |
+
"x": {"title": {"display": True, "text": "Text Categories"}}
|
99 |
+
},
|
100 |
+
"plugins": {"title": {"display": True, "text": "Text Analysis Confidence"}}
|
101 |
+
}
|
102 |
+
}
|
103 |
+
text_chart = f"""
|
104 |
+
<canvas id='textChart' width='400' height='200'></canvas>
|
105 |
+
<script src='https://cdn.jsdelivr.net/npm/chart.js'></script>
|
106 |
+
<script>
|
107 |
+
new Chart(document.getElementById('textChart'), {json.dumps(chart_data)});
|
108 |
+
</script>
|
109 |
+
"""
|
110 |
+
|
111 |
+
# Process image input
|
112 |
+
image_response = ""
|
113 |
+
image_chart = ""
|
114 |
+
if image is not None:
|
115 |
+
conversation_state["image_uploaded"] = True
|
116 |
+
inputs = image_processor(images=image, return_tensors="pt")
|
117 |
+
with torch.no_grad():
|
118 |
+
outputs = image_model(**inputs)
|
119 |
+
logits = outputs.logits
|
120 |
+
predicted_class_idx = logits.argmax(-1).item()
|
121 |
+
confidence = torch.softmax(logits, dim=-1)[0][predicted_class_idx].item()
|
122 |
+
scores = torch.softmax(logits, dim=-1)[0].tolist()
|
123 |
+
conversation_state["last_analysis"] = {
|
124 |
+
"type": "image",
|
125 |
+
"label": image_labels[predicted_class_idx],
|
126 |
+
"confidence": confidence,
|
127 |
+
"scores": scores,
|
128 |
+
"input": "image",
|
129 |
+
"timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
130 |
+
}
|
131 |
+
conversation_state["analysis_log"].append(conversation_state["last_analysis"])
|
132 |
+
image_response = f"Image Analysis: {image_labels[predicted_class_idx]} (Confidence: {confidence:.2%})"
|
133 |
+
|
134 |
+
# Image visualization (Chart.js)
|
135 |
+
chart_data = {
|
136 |
+
"type": "bar",
|
137 |
+
"data": {
|
138 |
+
"labels": image_labels,
|
139 |
+
"datasets": [{
|
140 |
+
"label": "Confidence Scores",
|
141 |
+
"data": scores,
|
142 |
+
"backgroundColor": ["#4CAF50", "#F44336"],
|
143 |
+
"borderColor": ["#388E3C", "#D32F2F"],
|
144 |
+
"borderWidth": 1
|
145 |
+
}]
|
146 |
+
},
|
147 |
+
"options": {
|
148 |
+
"scales": {
|
149 |
+
"y": {"beginAtZero": True, "max": 1, "title": {"display": True, "text": "Confidence"}},
|
150 |
+
"x": {"title": {"display": True, "text": "Image Categories"}}
|
151 |
+
},
|
152 |
+
"plugins": {"title": {"display": True, "text": "Image Analysis Confidence"}}
|
153 |
+
}
|
154 |
+
}
|
155 |
+
image_chart = f"""
|
156 |
+
<canvas id='imageChart' width='400' height='200'></canvas>
|
157 |
+
<script src='https://cdn.jsdelivr.net/npm/chart.js'></script>
|
158 |
+
<script>
|
159 |
+
new Chart(document.getElementById('imageChart'), {json.dumps(chart_data)});
|
160 |
+
</script>
|
161 |
+
"""
|
162 |
+
|
163 |
+
# Generate trend visualization (Plotly)
|
164 |
+
trend_html = ""
|
165 |
+
if len(conversation_state["analysis_log"]) > 1:
|
166 |
+
df = pd.DataFrame(conversation_state["analysis_log"])
|
167 |
+
fig = px.line(
|
168 |
+
df, x="timestamp", y="confidence", color="type",
|
169 |
+
title="Analysis Confidence Over Time",
|
170 |
+
labels={"confidence": "Confidence Score", "timestamp": "Time"}
|
171 |
+
)
|
172 |
+
trend_html = fig.to_html(full_html=False)
|
173 |
+
|
174 |
+
# Combine responses
|
175 |
+
response = "\n".join([r for r in [text_response, image_response] if r])
|
176 |
+
if not response:
|
177 |
+
response = "No analysis yet. Please provide text or upload an image."
|
178 |
+
response += f"\n\nFollow-Up: {generate_follow_up(conversation_state['texts'], conversation_state['history'])}"
|
179 |
+
response += f"\n\n{text_chart}\n{image_chart}\n{trend_html}"
|
180 |
+
|
181 |
+
# Handle export
|
182 |
+
if export_format != "None":
|
183 |
+
df = pd.DataFrame(conversation_state["analysis_log"])
|
184 |
+
if export_format == "JSON":
|
185 |
+
export_data = df.to_json(orient="records")
|
186 |
+
return response, gr.File(value=StringIO(export_data), file_name="analysis_log.json")
|
187 |
+
elif export_format == "CSV":
|
188 |
+
export_data = df.to_csv(index=False)
|
189 |
+
return response, gr.File(value=StringIO(export_data), file_name="analysis_log.csv")
|
190 |
+
|
191 |
+
# Add disclaimer
|
192 |
+
disclaimer = "β οΈ This tool is for research purposes only and does not provide medical diagnoses. Consult a healthcare professional for medical advice."
|
193 |
+
response += f"\n\n{disclaimer}"
|
194 |
+
|
195 |
+
conversation_state["history"].append((user_input, response))
|
196 |
+
return response
|
197 |
+
|
198 |
+
# Custom CSS for professional UI
|
199 |
+
css = """
|
200 |
+
body { background-color: #f0f2f5; font-family: 'Segoe UI', Arial, sans-serif; }
|
201 |
+
.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); }
|
202 |
+
h1 { color: #1a3c5e; text-align: center; font-size: 2em; }
|
203 |
+
input, textarea { border-radius: 8px; border: 1px solid #ccc; padding: 10px; }
|
204 |
+
button { background: linear-gradient(90deg, #3498db, #2980b9); color: white; border-radius: 8px; padding: 12px; font-weight: bold; }
|
205 |
+
button:hover { background: linear-gradient(90deg, #2980b9, #1a6ea6); }
|
206 |
+
#export_dropdown { width: 150px; margin-top: 10px; }
|
207 |
+
"""
|
208 |
+
|
209 |
+
# Create Gradio interface
|
210 |
+
with gr.Blocks(css=css) as iface:
|
211 |
+
gr.Markdown("# Ultra-Advanced Medical Research Chatbot")
|
212 |
+
gr.Markdown("Analyze medical texts or images for research purposes. Supports symptom analysis, literature review, or space medicine research. Not for medical diagnosis.")
|
213 |
+
with gr.Row():
|
214 |
+
with gr.Column(scale=2):
|
215 |
+
text_input = gr.Textbox(lines=5, placeholder="Enter symptoms, medical abstract, or space medicine data...")
|
216 |
+
image_input = gr.Image(type="pil", label="Upload X-ray or Skin Image")
|
217 |
+
export_dropdown = gr.Dropdown(choices=["None", "JSON", "CSV"], label="Export Log", value="None")
|
218 |
+
submit_button = gr.Button("Analyze")
|
219 |
+
with gr.Column(scale=3):
|
220 |
+
output = gr.HTML()
|
221 |
+
submit_button.click(
|
222 |
+
fn=analyze_medical_input,
|
223 |
+
inputs=[text_input, image_input, gr.State(), export_dropdown],
|
224 |
+
outputs=[output, gr.File()]
|
225 |
+
)
|
226 |
+
|
227 |
+
# Launch the interface
|
228 |
+
iface.launch()
|