app.py

import os
import streamlit as st
from PIL import Image
import torch
from torchvision import transforms, models
import numpy as np
from groq import Groq

# Set up environment variables
os.environ["GROQ_API_KEY"] = "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"

# Initialize Groq client
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))

# Load Pretrained Models
@st.cache_resource
# Load Pretrained Model for Organ Recognition
@st.cache_resource
def load_organ_model():
    model = models.resnet18(pretrained=True)  # Load pretrained ResNet18
    num_features = model.fc.in_features       # Get the number of input features to the final layer
    model.fc = torch.nn.Linear(num_features, 4)  # Modify the final layer for 4 classes
    model.eval()  # Set the model to evaluation mode
    return model



# Image Preprocessing
def preprocess_image(image):
    transform = transforms.Compose([
        transforms.Resize((224, 224)),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ])
    return transform(image).unsqueeze(0)

# Groq API for AI Insights
def get_ai_insights(text_prompt):
    try:
        response = client.chat.completions.create(
            messages=[{"role": "user", "content": text_prompt}],
            model="llama-3.3-70b-versatile"
        )
        return response.choices[0].message.content
    except Exception as e:
        return f"Error: {e}"
        
# Organ Recognition Prediction
def predict_organ(image):
    with torch.no_grad():
        input_tensor = preprocess_image(image)
        output = organ_model(input_tensor)

        # Check the output dimensions
        st.write(f"Model output shape: {output.shape}")

        # Ensure the output matches the number of classes
        classes = ["Lungs", "Heart", "Spine", "Other"]
        if output.size(1) != len(classes):
            raise ValueError(
                f"Model output size ({output.size(1)}) does not match the number of classes ({len(classes)})."
            )

        # Get the prediction
        prediction_index = output.argmax().item()
        prediction = classes[prediction_index]
        return prediction


# Predict Normal/Abnormal
def predict_normal_abnormal(image):
    with torch.no_grad():
        output = chexnet_model(preprocess_image(image))
        classes = ["Normal", "Abnormal"]
        prediction = classes[output.argmax().item()]
    return prediction

# Streamlit App
st.title("Medical X-ray Analysis App")
st.sidebar.title("Navigation")
task = st.sidebar.radio("Select a task", ["Upload X-ray", "AI Insights"])

if task == "Upload X-ray":
    uploaded_file = st.file_uploader("Upload an X-ray image", type=["jpg", "png", "jpeg"])

    if uploaded_file:
        image = Image.open(uploaded_file)
        st.image(image, caption="Uploaded X-ray", use_column_width=True)

        # Predict Organ
        st.subheader("Step 1: Identify the Organ")
        organ = predict_organ(image)
        st.write(f"Predicted Organ: **{organ}**")

        # Predict Normal/Abnormal
        st.subheader("Step 2: Analyze the X-ray")
        classification = predict_normal_abnormal(image)
        st.write(f"X-ray Status: **{classification}**")

        if classification == "Abnormal":
            st.subheader("Step 3: AI-Based Insights")
            ai_prompt = f"Explain why this X-ray of the {organ} is abnormal."
            insights = get_ai_insights(ai_prompt)
            st.write(insights)

elif task == "AI Insights":
    st.subheader("Ask AI")
    user_input = st.text_area("Enter your query for AI insights")
    if user_input:
        response = get_ai_insights(user_input)
        st.write(response)