virtualhealth.py

# -*- coding: utf-8 -*-
"""VirtualHealth.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1yVSYiPz-WUlO7U0uAKe9LmfMXHd5eyAA
"""
!pip install streamlit

import xgboost as xgb
import pickle
import numpy as np
import pandas as pd

# Load the trained model
model = xgb.XGBClassifier()
model.load_model("symptom_disease_model.json")

# Load the label encoder
label_encoder = pickle.load(open("label_encoder.pkl", "rb"))

# Load symptom names (from preprocessed training data)
X_train = pd.read_csv("X_train.csv")  # Get feature names
symptom_list = X_train.columns.tolist()

# Function to Predict Disease
def predict_disease(user_symptoms):
    # Convert user symptoms into one-hot encoded format
    input_vector = np.zeros(len(symptom_list))

    for symptom in user_symptoms:
        if symptom in symptom_list:
            input_vector[symptom_list.index(symptom)] = 1

    input_vector = input_vector.reshape(1, -1)  # Reshape for model

    # Predict disease (returns a numerical class)
    predicted_class = model.predict(input_vector)[0]

    # Convert number to disease name
    predicted_disease = label_encoder.inverse_transform([predicted_class])[0]

    return predicted_disease

# Example Usage
user_symptoms = ["itching", "skin_rash", "nodal_skin_eruptions"]
predicted_disease = predict_disease(user_symptoms)
print(f"Predicted Disease: {predicted_disease}")

!pip install zipfile36
import sys
if sys.version_info >= (3, 6):
    import zipfile
else:
    import zipfile36 as zipfile
import os

zip_file_path = '/content/disease symptom.zip'  # Update with your path
extracted_dir = '/content'  # Where to extract the files

with zipfile.ZipFile(zip_file_path, 'r') as zip_ref:
    zip_ref.extractall(extracted_dir)

# Load the precaution dataset
precaution_df = pd.read_csv("Disease precaution.csv")

# Convert to dictionary for fast lookup
precaution_dict = {}
for _, row in precaution_df.iterrows():
    disease = row["Disease"].strip().lower()
    precautions = [row[f"Precaution_{i}"] for i in range(1, 5) if pd.notna(row[f"Precaution_{i}"])]
    precaution_dict[disease] = precautions

# Function to Get Precautions
def get_precautions(disease_name):
    disease_name = disease_name.strip().lower()
    return precaution_dict.get(disease_name, ["No precautions found"])

# Example Usage
precautions = get_precautions(predicted_disease)
print(f"Precautions for {predicted_disease}: {precautions}")

!pip install nltk

import re
import nltk
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize

# Download stopwords if not already downloaded
nltk.download("stopwords")
nltk.download("punkt")

# Load English stopwords
stop_words = set(stopwords.words("english"))
nltk.download('punkt_tab')

import xgboost as xgb
import pickle
import numpy as np
import pandas as pd
import torch
from transformers import AutoTokenizer, AutoModelForQuestionAnswering
import re  # Import regex module for better input processing

# ============================
# 🔹 1. Load Pretrained Medical Q&A Model
# ============================
qa_model_name = "deepset/roberta-base-squad2"  # Better model for medical Q&A
tokenizer = AutoTokenizer.from_pretrained(qa_model_name)
qa_model = AutoModelForQuestionAnswering.from_pretrained(qa_model_name)

# ============================
# 🔹 2. Load Symptom Checker Model & Label Encoder (Fixed)
# ============================
# Load trained XGBoost model from JSON
model = xgb.XGBClassifier()
model.load_model("symptom_disease_model.json")
common_symptoms = ["fever", "cough", "headache", "pain", "vomiting", "fatigue", "nausea", "rash", "chills", "dizziness", "sore throat", "diarrhea"]

# Load Corrected Label Encoder
label_encoder = pickle.load(open("label_encoder.pkl", "rb"))

# Load symptom names from training data
X_train = pd.read_csv("X_train.csv")  # Get feature names
symptom_list = X_train.columns.tolist()

# ============================
# 🔹 3. Load Precaution Data
# ============================
precaution_df = pd.read_csv("Disease precaution.csv")
precaution_dict = {
    row["Disease"].strip().lower(): [row[f"Precaution_{i}"] for i in range(1, 5) if pd.notna(row[f"Precaution_{i}"])]
    for _, row in precaution_df.iterrows()
}

# ============================
# 🔹 4. Load Medical Context
# ============================
def load_medical_context():
    with open("medical_context.txt", "r", encoding="utf-8") as file:
        return file.read()

medical_context = load_medical_context()

# ============================
# 🔹 5. Doctor Database (For Appointments)
# ============================
doctor_database = {
    "malaria": [{"name": "Dr. Rajesh Kumar", "specialty": "Infectious Diseases", "location": "Apollo Hospital", "contact": "9876543210"}],
    "diabetes": [{"name": "Dr. Anil Mehta", "specialty": "Endocrinologist", "location": "AIIMS Delhi", "contact": "9876543233"}],
    "heart attack": [{"name": "Dr. Vikram Singh", "specialty": "Cardiologist", "location": "Medanta Hospital", "contact": "9876543255"}],
    "hepatitis e": [{"name": "Dr. Sunil Agarwal", "specialty": "Hepatologist", "location": "Fortis Hospital", "contact": "9876543266"}],
    "pneumonia": [{"name": "Dr. Priya Sharma", "specialty": "Pulmonologist", "location": "Max Healthcare", "contact": "9876543277"}],
    "heartattack": [{"name": "Dr. Vikram Singh", "specialty": "Cardiologist", "location": "Medanta Hospital", "contact": "9876543255"}],
}

# ============================
# 🔹 6. Predict Disease from Symptoms (Fully Fixed)
# ============================
def predict_disease(user_symptoms):
    """Predicts the disease based on user symptoms using the trained XGBoost model."""
    input_vector = np.zeros(len(symptom_list))

    for symptom in user_symptoms:
        if symptom in symptom_list:
            input_vector[symptom_list.index(symptom)] = 1  # One-hot encoding

    input_vector = input_vector.reshape(1, -1)  # Reshape for model input

    # Predict disease (returns a numerical class)
    predicted_class = model.predict(input_vector)[0]

    # Convert number to disease name
    predicted_disease = label_encoder.inverse_transform([predicted_class])[0]

    return predicted_disease

# ============================
# 🔹 7. Get Precautions for a Disease
# ============================
def get_precautions(disease):
    """Returns the precautions for a given disease."""
    return precaution_dict.get(disease.lower(), ["No precautions available"])

# ============================
# 🔹 8. Answer Medical Questions (Q&A Model)
# ============================
def get_medical_answer(question):
    """Uses the pre-trained Q&A model to answer general medical questions."""
    inputs = tokenizer(question, medical_context, return_tensors="pt", truncation=True, max_length=512)
    with torch.no_grad():
        outputs = qa_model(**inputs)

    answer_start = torch.argmax(outputs.start_logits)
    answer_end = torch.argmax(outputs.end_logits) + 1

    answer = tokenizer.convert_tokens_to_string(
        tokenizer.convert_ids_to_tokens(inputs["input_ids"][0][answer_start:answer_end])
    )

    return answer if answer.strip() and answer != "[CLS]" else "No reliable answer found."

# ============================
# 🔹 9. Book a Doctor's Appointment
# ============================
def book_appointment(disease):
    """Finds a doctor for the given disease and returns appointment details."""
    disease = disease.lower().strip()
    doctors = doctor_database.get(disease, [])
    if not doctors:
        return f"Sorry, no available doctors found for {disease}."

    doctor = doctors[0]
    return f"Appointment booked with **{doctor['name']}** ({doctor['specialty']}) at **{doctor['location']}**.\nContact: {doctor['contact']}"

# ============================
# 🔹 10. Handle User Queries
# ============================
def extract_treatment_from_context(disease):
    """Extracts treatment details for a given disease from `medical_context.txt`."""
    with open("medical_context.txt", "r", encoding="utf-8") as file:
        lines = file.readlines()

    treatment_section = []
    found_disease = False
    found_treatment = False

    for line in lines:
        line = line.strip()

        #  Check if we found the disease name
        if f"## {disease.lower()}" in line.lower():
            found_disease = True

        #  If we found the disease, now look for "Treatment"
        if found_disease and "**Treatment**" in line:
            found_treatment = True
            continue  # Skip the "**Treatment**:" line itself

        #  If found, keep extracting treatment details
        if found_treatment:
            # Stop at blank line or the next section (## New Disease Name)
            if line == "" or line.startswith("## "):
                break
            treatment_section.append(line)

    return "\n".join(treatment_section) if treatment_section else None


def extract_disease_name(user_query):
    """Extracts the disease name by removing unnecessary words, but keeps medical terms."""
    user_query_cleaned = re.sub(r"[^\w\s]", "", user_query.lower())  # Remove punctuation
    words = word_tokenize(user_query_cleaned)

    #  Remove stopwords but keep diseases/symptoms
    filtered_words = [word for word in words if word not in stop_words or word in common_symptoms]

    return " ".join(filtered_words).strip()

def find_best_match(query, database):
    """Finds the best matching disease from the database based on query words."""
    query_words = query.split()  # Split query into words

    # Check for exact match first
    if query in database:
        return query  # Exact match found

    # Check if any word in query exists in database keys
    for disease in database:
        for word in query_words:
            if word in disease:  # Partial match found
                return disease

    return None  # No match found


def handle_user_query(user_query):
    """Handles user queries related to symptoms, diseases, and doctor appointments."""

    user_query = user_query.lower().strip()

    #  Skip Cleaning for "I have..." and "experiencing..." Cases
    if "i have" in user_query or "experiencing" in user_query:
        symptoms = user_query.replace("I have", "").replace("experiencing", "").strip()
        disease = predict_disease(symptoms.split(", "))  # Convert to list
        precautions = get_precautions(disease)
        return f"**Predicted Disease:** {disease}\n**Precautions:** {', '.join(precautions)}\n{book_appointment(disease)}"

    #  Extract Disease Name for Queries
    user_query_cleaned = extract_disease_name(user_query)

    #  Handle "Who should I see for..." Queries (Improved with Partial Matching)
    if "who should i see " in user_query:
        disease_query = user_query.replace("who should i see", "").strip()
        disease = find_best_match(disease_query, doctor_database)  # Get best match

        if disease:
            doctor = doctor_database[disease][0]
            return f"You should see a **{doctor['specialty']}** for {disease}.\nExample: {doctor['name']} at {doctor['location']}."
        else:
            return "I'm not sure. Please consult a general physician for more guidance."

    #  Book Appointment (Improved with Partial Matching)
    elif "book appointment" in user_query_cleaned:
        disease_query = user_query_cleaned.replace("book appointment", "").strip()
        disease = find_best_match(disease_query, doctor_database)
        return book_appointment(disease) if disease else "Sorry, no matching doctor found."

    #  Symptoms Query
    elif "symptoms" in user_query_cleaned or "signs" in user_query_cleaned:
        disease = user_query_cleaned.replace("symptoms", "").replace("signs", "").strip()
        return get_medical_answer(f"What are the symptoms of {disease}?")

    #  Precautions Query
    elif "precautions" in user_query_cleaned or "prevent" in user_query_cleaned:
        disease = user_query_cleaned.replace("precautions", "").replace("prevent", "").strip()
        return ", ".join(get_precautions(disease))

    #  Treatment Query
    if "treatment" in user_query_cleaned or "treat" in user_query_cleaned:
        disease = user_query_cleaned.replace("treatment", "").replace("treat", "").strip()

        # 🔹 First, try to extract treatment from `medical_context.txt`
        treatment_answer = extract_treatment_from_context(disease)
        if treatment_answer:
            return treatment_answer  #  Use direct extraction first

        # 🔹 If no treatment info found, use the Q&A Model
        model_answer = get_medical_answer(f"What is the treatment for {disease}?")
        if model_answer in ["<s>", "", "No reliable answer found."]:
            return f"I'm not sure, but common treatments for {disease} include medication, therapy, or consulting a specialist."
        return model_answer

    #  General Medical Questions (Fallback)
    else:
        response = get_medical_answer(user_query)
        if response in ["<s>", "", "No reliable answer found."]:
            return "I'm not sure, but you may consult a specialist for better guidance."
        return response

# ============================
# 🔹 11. Test Cases (Run Examples)
# ============================
print(handle_user_query("I have fever, chills, and muscle aches"))  # Should predict disease & precautions
print(handle_user_query("What are the symptoms of pneumonia?"))  # Should return pneumonia symptoms
print(handle_user_query("Book an appointment for diabetes"))  # Should book a diabetes specialist
print(handle_user_query("Who should I see for  heart attack"))  # Should return "Cardiologist"
print(handle_user_query("what is the treatment for tuberculosis"))  # Should return correct treatment

print(handle_user_query("What is the treatment for tuberculosis?"))  # Should return correct treatment
print(handle_user_query("What is the treatment for malaria?"))  # Should also work
print(handle_user_query("What is the treatment for cancer?"))  # Should return something useful

print(handle_user_query("What is the treatment for tuberculosis?"))  # Should return correct treatment
print(handle_user_query("What is the treatment for malaria?"))  # Should also work
print(handle_user_query("What is the treatment for cancer?"))  # Should return something useful
print(handle_user_query("How to treat diabetes?"))  # Should return proper treatment
print(handle_user_query("Tell me the cure for pneumonia?"))  # Should return treatment
print(handle_user_query("Treatment for typhoid?"))  # Should extract treatment

print(handle_user_query("What are the symptoms of pneumonia?"))  # Should return correct symptoms
print(handle_user_query("Signs of heart attack?"))  # Should return expected symptoms
print(handle_user_query("How do I know if I have typhoid?"))  # Should return typhoid symptoms
print(handle_user_query("What symptoms should I check for tuberculosis?"))  # Should work
print(handle_user_query("Symptoms of dengue?"))  # Should return symptoms of dengue

print(handle_user_query("Who should I see for a heart attack?"))  # Should return "Cardiologist"
print(handle_user_query("Which doctor should I visit for diabetes?"))  # Should return "Endocrinologist"
print(handle_user_query("Who should I consult for a skin rash?"))  # Should return "Dermatologist"
print(handle_user_query("What kind of doctor treats pneumonia?"))  # Should return "Pulmonologist"
print(handle_user_query("Who specializes in treating migraines?"))  # Should return "Neurologist"

print(handle_user_query("Book an appointment for malaria"))  # Should book doctor for malaria
print(handle_user_query("I need a doctor for high blood pressure"))  # Should book doctor for hypertension
print(handle_user_query("Schedule a consultation for fever"))  # Should book general physician
print(handle_user_query("Find a doctor for diabetes treatment"))  # Should book endocrinologist
print(handle_user_query("Book an appointment for pneumonia treatment"))  # Should book pulmonologist

print(handle_user_query("I have fever, cough, and chills"))  # Should predict disease correctly
print(handle_user_query("Experiencing blurry vision and excessive thirst"))  # Should return "Diabetes"
print(handle_user_query("I am experiencing severe chest pain and difficulty breathing"))  # Should return "Heart Attack"
print(handle_user_query("Feeling tired, cold, and gaining weight"))  # Should return "Hypothyroidism"
print(handle_user_query("I have rash, joint pain, and headache"))  # Should return "Dengue"

print(handle_user_query("What does a doctor do?"))  # Should return general doctor description
print(handle_user_query("What are antibiotics?"))  # Should explain antibiotics
print(handle_user_query("How does the immune system work?"))  # Should explain immunity
print(handle_user_query("What is the function of the liver?"))  # Should explain liver function
print(handle_user_query("Explain how blood pressure works?"))  # Should provide useful explanation




# Commented out IPython magic to ensure Python compatibility.
# %%writefile app.py
# import streamlit as st
# import requests
# 
# st.set_page_config(page_title="AI Health Assistant", page_icon="🤖")
# 
# st.title("🩺 AI Health Assistant")
# st.write("Ask any medical-related questions:")
# 
# # User Input
# user_input = st.text_input("Your Question:")
# 
# # Button to Send Query
# if st.button("Ask"):
#     response = requests.post("https://b7da-35-232-247-117.ngrok-free.app/query/", json={"user_input": user_input})
#     bot_response = response.json().get("response", "Error fetching response")
# 
#     st.markdown(f"**🤖 Bot:** {bot_response}")
#

"""✅ Steps to Deploy on Hugging Face Spaces
📌 Step 1: Create a Hugging Face Space
1️⃣ Go to Hugging Face Spaces
2️⃣ Click "New Space"
3️⃣ Name the Space (e.g., AI-Health-Assistant)
4️⃣ Select "Streamlit" as the SDK
5️⃣ Click "Create Space" ✅

📌 Step 2: Clone the Repository Locally
After creating the Space, clone it to your local machine or Google Colab:

```bash
git clone https://huggingface.co/spaces/YOUR_USERNAME/AI-Health-Assistant
cd AI-Health-Assistant
```
Replace YOUR_USERNAME with your Hugging Face username!

📌 Step 3: Add app.py (Your Streamlit Chatbot)
Inside the cloned folder, create app.py and paste the following:

📌 Step 4: Create requirements.txt
Create a new file requirements.txt inside the same folder and add:
```bash
streamlit
requests
```

📌 Step 5: Push Your Code to Hugging Face
Run these commands to push the code:

```bash
git add .
git commit -m "Initial commit"
git push
```
🚀 Your Space will automatically start building!
"""