Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,84 +1,100 @@
|
|
1 |
import os
|
2 |
import streamlit as st
|
3 |
-
import tensorflow as tf
|
4 |
-
from tensorflow.keras.applications import ResNet50
|
5 |
-
from tensorflow.keras.applications.resnet50 import preprocess_input, decode_predictions
|
6 |
-
from tensorflow.keras.models import Model
|
7 |
-
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
|
8 |
-
from tensorflow.keras.preprocessing.image import img_to_array, load_img
|
9 |
-
import numpy as np
|
10 |
-
import requests
|
11 |
from PIL import Image
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
-
#
|
14 |
-
os.environ
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
predictions = Dense(1, activation='sigmoid')(x)
|
23 |
-
classification_model = Model(inputs=base_model.input, outputs=predictions)
|
24 |
-
|
25 |
-
# Load pre-trained ResNet50 for organ recognition
|
26 |
-
organ_model = ResNet50(weights='imagenet')
|
27 |
-
|
28 |
-
def classify_image(image_path):
|
29 |
-
"""Classify the image as normal or abnormal."""
|
30 |
-
image = load_img(image_path, target_size=(224, 224))
|
31 |
-
image_array = img_to_array(image)
|
32 |
-
image_array = preprocess_input(image_array)
|
33 |
-
image_array = np.expand_dims(image_array, axis=0)
|
34 |
-
prediction = classification_model.predict(image_array)
|
35 |
-
return 'Abnormal' if prediction[0][0] > 0.5 else 'Normal'
|
36 |
-
|
37 |
-
def recognize_organ(image_path):
|
38 |
-
"""Recognize the organ in the image."""
|
39 |
-
image = load_img(image_path, target_size=(224, 224))
|
40 |
-
image_array = img_to_array(image)
|
41 |
-
image_array = preprocess_input(image_array)
|
42 |
-
image_array = np.expand_dims(image_array, axis=0)
|
43 |
-
prediction = organ_model.predict(image_array)
|
44 |
-
decoded = decode_predictions(prediction, top=3)[0]
|
45 |
-
return decoded[0][1] # Top predicted class
|
46 |
-
|
47 |
-
def get_ai_insights(organ):
|
48 |
-
"""Fetch AI-based insights about the organ using Groq API."""
|
49 |
-
url = "https://api.groq.com/v1/insights"
|
50 |
-
headers = {"Authorization": f"Bearer {GROQ_API_KEY}", "Content-Type": "application/json"}
|
51 |
-
data = {"query": f"Provide detailed insights about {organ} X-ray, its diseases, and treatments."}
|
52 |
-
response = requests.post(url, headers=headers, json=data)
|
53 |
-
if response.status_code == 200:
|
54 |
-
return response.json().get("insights", "No insights available.")
|
55 |
-
else:
|
56 |
-
return "Failed to fetch insights. Please try again later."
|
57 |
-
|
58 |
-
def main():
|
59 |
-
st.title("Medical Image Classification App")
|
60 |
-
st.sidebar.title("Navigation")
|
61 |
|
62 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
if uploaded_file:
|
65 |
image = Image.open(uploaded_file)
|
66 |
-
st.image(image, caption="Uploaded
|
67 |
-
|
68 |
-
with open("temp_image.jpg", "wb") as f:
|
69 |
-
f.write(uploaded_file.getbuffer())
|
70 |
|
71 |
-
|
72 |
-
|
73 |
-
|
|
|
74 |
|
75 |
-
|
76 |
-
|
77 |
-
|
|
|
78 |
|
79 |
-
|
80 |
-
|
81 |
-
|
|
|
|
|
82 |
|
83 |
-
|
84 |
-
|
|
|
|
|
|
|
|
|
|
1 |
import os
|
2 |
import streamlit as st
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
from PIL import Image
|
4 |
+
import torch
|
5 |
+
from torchvision import transforms, models
|
6 |
+
import numpy as np
|
7 |
+
from groq import Groq
|
8 |
+
|
9 |
+
# Set up environment variables
|
10 |
+
os.environ["GROQ_API_KEY"] = "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"
|
11 |
|
12 |
+
# Initialize Groq client
|
13 |
+
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
|
14 |
+
|
15 |
+
# Load Pretrained Models
|
16 |
+
@st.cache_resource
|
17 |
+
def load_model():
|
18 |
+
# Pretrained EfficientNet for organ recognition
|
19 |
+
organ_model = models.efficientnet_b0(pretrained=True)
|
20 |
+
organ_model.eval()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
# Pretrained DenseNet (CheXNet) for normal/abnormal classification
|
23 |
+
chexnet_model = models.densenet121(pretrained=True)
|
24 |
+
chexnet_model.classifier = torch.nn.Linear(1024, 2) # Normal, Abnormal
|
25 |
+
chexnet_model.eval()
|
26 |
+
|
27 |
+
return organ_model, chexnet_model
|
28 |
+
|
29 |
+
organ_model, chexnet_model = load_model()
|
30 |
+
|
31 |
+
# Image Preprocessing
|
32 |
+
def preprocess_image(image):
|
33 |
+
transform = transforms.Compose([
|
34 |
+
transforms.Resize((224, 224)),
|
35 |
+
transforms.ToTensor(),
|
36 |
+
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
|
37 |
+
])
|
38 |
+
return transform(image).unsqueeze(0)
|
39 |
+
|
40 |
+
# Groq API for AI Insights
|
41 |
+
def get_ai_insights(text_prompt):
|
42 |
+
try:
|
43 |
+
response = client.chat.completions.create(
|
44 |
+
messages=[{"role": "user", "content": text_prompt}],
|
45 |
+
model="llama-3.3-70b-versatile"
|
46 |
+
)
|
47 |
+
return response.choices[0].message.content
|
48 |
+
except Exception as e:
|
49 |
+
return f"Error: {e}"
|
50 |
+
|
51 |
+
# Predict Organ
|
52 |
+
def predict_organ(image):
|
53 |
+
with torch.no_grad():
|
54 |
+
output = organ_model(preprocess_image(image))
|
55 |
+
classes = ["Lungs", "Heart", "Spine", "Other"] # Example classes
|
56 |
+
prediction = classes[output.argmax().item()]
|
57 |
+
return prediction
|
58 |
+
|
59 |
+
# Predict Normal/Abnormal
|
60 |
+
def predict_normal_abnormal(image):
|
61 |
+
with torch.no_grad():
|
62 |
+
output = chexnet_model(preprocess_image(image))
|
63 |
+
classes = ["Normal", "Abnormal"]
|
64 |
+
prediction = classes[output.argmax().item()]
|
65 |
+
return prediction
|
66 |
+
|
67 |
+
# Streamlit App
|
68 |
+
st.title("Medical X-ray Analysis App")
|
69 |
+
st.sidebar.title("Navigation")
|
70 |
+
task = st.sidebar.radio("Select a task", ["Upload X-ray", "AI Insights"])
|
71 |
+
|
72 |
+
if task == "Upload X-ray":
|
73 |
+
uploaded_file = st.file_uploader("Upload an X-ray image", type=["jpg", "png", "jpeg"])
|
74 |
|
75 |
if uploaded_file:
|
76 |
image = Image.open(uploaded_file)
|
77 |
+
st.image(image, caption="Uploaded X-ray", use_column_width=True)
|
|
|
|
|
|
|
78 |
|
79 |
+
# Predict Organ
|
80 |
+
st.subheader("Step 1: Identify the Organ")
|
81 |
+
organ = predict_organ(image)
|
82 |
+
st.write(f"Predicted Organ: **{organ}**")
|
83 |
|
84 |
+
# Predict Normal/Abnormal
|
85 |
+
st.subheader("Step 2: Analyze the X-ray")
|
86 |
+
classification = predict_normal_abnormal(image)
|
87 |
+
st.write(f"X-ray Status: **{classification}**")
|
88 |
|
89 |
+
if classification == "Abnormal":
|
90 |
+
st.subheader("Step 3: AI-Based Insights")
|
91 |
+
ai_prompt = f"Explain why this X-ray of the {organ} is abnormal."
|
92 |
+
insights = get_ai_insights(ai_prompt)
|
93 |
+
st.write(insights)
|
94 |
|
95 |
+
elif task == "AI Insights":
|
96 |
+
st.subheader("Ask AI")
|
97 |
+
user_input = st.text_area("Enter your query for AI insights")
|
98 |
+
if user_input:
|
99 |
+
response = get_ai_insights(user_input)
|
100 |
+
st.write(response)
|