TestData.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime, timedelta
import os

# Define save directory
save_directory = r'C:\Users\jsntg\OneDrive\Desktop\3YP\code\data'
os.makedirs(save_directory, exist_ok=True)  # make sure directory exists

# Set random seed for reproducibility
np.random.seed(42)

# Generate temporal data for 100 patients
N = 100
patients_data = []

for patient_id in range(1, N + 1):
    # Decide total monitoring hours (between 6 to 24 hours)
    total_hours = np.random.randint(6, 25)
    
    # Decide sepsis onset hour (after hour 3)
    sepsis_start_hour = np.random.randint(3, max(4, total_hours - 2))
    
    # Decide sepsis duration (at least 2 hours)
    sepsis_duration = np.random.randint(2, total_hours - sepsis_start_hour + 1)
    
    sepsis_end_hour = sepsis_start_hour + sepsis_duration
    
    # Generate data for each time point
    for hour in range(total_hours):
        in_sepsis = sepsis_start_hour <= hour < sepsis_end_hour
        
        # Base data
        base_data = {
            'patient_id': patient_id,
            'time_hour': hour,
            'timestamp': datetime.now() - timedelta(hours=total_hours-hour),
        }
        
        if not in_sepsis:
            # Normal state
            vital_signs = {
                'respiratory_rate': np.clip(np.random.normal(18, 2), 12, 20),
                'oxygen_saturations': np.clip(np.random.normal(97, 1), 94, 100),
                'supplemental_oxygen': np.random.choice([0, 1], p=[0.95, 0.05]),
                'temperature': np.clip(np.random.normal(36.8, 0.3), 36.0, 37.5),
                'systolic_bp': np.clip(np.random.normal(115, 10), 100, 130),
                'diastolic_bp': np.clip(np.random.normal(75, 8), 60, 85),
                'heart_rate': np.clip(np.random.normal(75, 8), 60, 90),
                'consciousness': np.random.choice([1, 2], p=[0.9, 0.1]),
                'pain': np.random.randint(0, 4),
                'discharge_lochia': np.random.choice([0, 1], p=[0.5, 0.5]),
                'proteinuria': np.random.choice([0, 1], p=[0.9, 0.1]),
                'sepsis': 0
            }
        else:
            # Sepsis state
            sepsis_progress = (hour - sepsis_start_hour) / sepsis_duration
            
            if sepsis_progress < 0.3:  # Early stage
                vital_signs = {
                    'respiratory_rate': np.clip(np.random.normal(22, 2), 20, 26),
                    'oxygen_saturations': np.clip(np.random.normal(94, 1.5), 90, 97),
                    'supplemental_oxygen': np.random.choice([0, 1], p=[0.7, 0.3]),
                    'temperature': np.clip(np.random.normal(38.2, 0.4), 37.5, 39.0),
                    'systolic_bp': np.clip(np.random.normal(110, 12), 95, 125),
                    'diastolic_bp': np.clip(np.random.normal(70, 10), 55, 80),
                    'heart_rate': np.clip(np.random.normal(95, 10), 85, 110),
                    'consciousness': np.random.choice([1, 2, 3], p=[0.6, 0.3, 0.1]),
                    'pain': np.random.randint(4, 8),
                    'discharge_lochia': np.random.choice([0, 1], p=[0.3, 0.7]),
                    'proteinuria': np.random.choice([0, 1], p=[0.7, 0.3]),
                    'sepsis': 1
                }
            elif sepsis_progress < 0.7:  # Mid stage
                vital_signs = {
                    'respiratory_rate': np.clip(np.random.normal(26, 3), 22, 32),
                    'oxygen_saturations': np.clip(np.random.normal(91, 2), 85, 95),
                    'supplemental_oxygen': np.random.choice([0, 1], p=[0.4, 0.6]),
                    'temperature': np.clip(np.random.normal(38.8, 0.5), 37.8, 39.8),
                    'systolic_bp': np.clip(np.random.normal(100, 15), 85, 115),
                    'diastolic_bp': np.clip(np.random.normal(65, 12), 50, 75),
                    'heart_rate': np.clip(np.random.normal(110, 12), 95, 130),
                    'consciousness': np.random.choice([2, 3, 4], p=[0.5, 0.3, 0.2]),
                    'pain': np.random.randint(6, 10),
                    'discharge_lochia': np.random.choice([0, 1], p=[0.2, 0.8]),
                    'proteinuria': np.random.choice([0, 1], p=[0.5, 0.5]),
                    'sepsis': 1
                }
            else:  # Late stage
                vital_signs = {
                    'respiratory_rate': np.clip(np.random.normal(30, 4), 24, 38),
                    'oxygen_saturations': np.clip(np.random.normal(88, 3), 80, 92),
                    'supplemental_oxygen': 1,
                    'temperature': np.clip(np.random.normal(39.5, 0.6), 38.5, 40.5),
                    'systolic_bp': np.clip(np.random.normal(90, 18), 75, 105),
                    'diastolic_bp': np.clip(np.random.normal(60, 15), 45, 70),
                    'heart_rate': np.clip(np.random.normal(125, 15), 110, 150),
                    'consciousness': np.random.choice([3, 4, 5, 6], p=[0.3, 0.3, 0.2, 0.2]),
                    'pain': np.random.randint(8, 11),
                    'discharge_lochia': 1,
                    'proteinuria': 1,
                    'sepsis': 1
                }
        
        record = {**base_data, **vital_signs}
        patients_data.append(record)

# Create DataFrame
df_data = pd.DataFrame(patients_data)

# Print summary statistics
print(f"Total number of records: {len(df_data)}")
print(f"Number of patients: {df_data['patient_id'].nunique()}")
print(f"Sepsis positive records: {df_data['sepsis'].sum()} ({df_data['sepsis'].mean()*100:.1f}%)")

print("\nPatient time series statistics:")
patient_statistics = df_data.groupby('patient_id').agg({
    'time_hour': 'count',
    'sepsis': 'sum'
}).rename(columns={'time_hour': 'total_hours', 'sepsis': 'sepsis_hours'})
print(patient_statistics)

# =============================================================================
# Step 2: Compute a multi‑parameter sepsis score and re‑label based on threshold
# =============================================================================
def calculate_sepsis_score(row):
    """
    Calculate a simplified sepsis score based on five core parameters.
    Each parameter contributes 0, 1, or 2 points.
    """
    score = 0
    
    # Respiratory rate (breaths/min)
    rr = row['respiratory_rate']
    if rr >= 24:
        score += 2
    elif rr >= 20:
        score += 1
    
    # Oxygen saturation (%)
    spo2 = row['oxygen_saturations']
    if spo2 <= 90:
        score += 2
    elif spo2 <= 94:
        score += 1
    
    # Temperature (°C)
    temp = row['temperature']
    if temp >= 38.5 or temp <= 36.0:
        score += 2
    elif temp >= 38.0:
        score += 1
    
    # Heart rate (bpm)
    hr = row['heart_rate']
    if hr >= 120:
        score += 2
    elif hr >= 100:
        score += 1
    
    # Consciousness level (1=alert, 2=lethargic, 3=obtunded, 4=stupor, 5=coma, 6=deep coma)
    consc = row['consciousness']
    if consc >= 4:
        score += 2
    elif consc >= 2:
        score += 1
    
    return score

# Apply the score function
df_data['sepsis_score'] = df_data.apply(calculate_sepsis_score, axis=1)

# Choose a threshold (adjust to obtain a reasonable sepsis prevalence)
threshold = 4   # You can change this value (e.g., 3, 4, or 5)
df_data['sepsis_new'] = (df_data['sepsis_score'] >= threshold).astype(int)

# Compare old vs new labels
changed = (df_data['sepsis'] != df_data['sepsis_new']).sum()
print("=== After score‑based relabeling ===")
print(f"Threshold = {threshold}")
print(f"Samples with changed label: {changed} ({changed/len(df_data):.2%})")
print(f"New sepsis‑positive rate: {df_data['sepsis_new'].mean()*100:.1f}%")

# Replace the original sepsis column with the new one
df_data['sepsis'] = df_data['sepsis_new']
df_data.drop(columns=['sepsis_score', 'sepsis_new'], inplace=True)

# =============================================================================
# Step 3: Add 5‑10% random label noise (flip some sepsis values)
# =============================================================================
# Use the same random seed for reproducibility (already set to 42 at the top)
flip_ratio = np.random.uniform(0.05, 0.1)          # random between 5% and 10%
n_flip = int(len(df_data) * flip_ratio)
flip_indices = np.random.choice(df_data.index, size=n_flip, replace=False)
df_data.loc[flip_indices, 'sepsis'] = 1 - df_data.loc[flip_indices, 'sepsis']

print("\n=== After adding 5‑10% random noise ===")
print(f"Flipped {n_flip} records ({flip_ratio:.2%})")
print(f"Final sepsis‑positive records: {df_data['sepsis'].sum()} ({df_data['sepsis'].mean()*100:.1f}%)\n")

# Save data to CSV
df_data.to_csv(os.path.join(save_directory, 'test_data_damn.csv'), index=False)
print("\nData saved as: 'test_data_damn.csv'")
print("\nFirst 5 rows of data:")
print(df_data.head())

# Visualize temporal data series
def plot_patient_timeline(patient_id, df):
    plot_data = df[df['patient_id'] == patient_id].sort_values('time_hour')
    
    fig, axes = plt.subplots(3, 2, figsize=(15, 12))
    fig.suptitle(f'Patient Id: {patient_id}', fontsize=16)
    
    # Some of the key metrics to plot
    metrics = [
        ('respiratory_rate', 'Respiratory Rate'),
        ('oxygen_saturations', 'Oxygen Saturations'),
        ('temperature', 'Temperature'),
        ('heart_rate', 'Heart Rate'),
        ('systolic_bp', 'Systolic Blood Pressure'),
        ('pain', 'Pain')
    ]
    
    for idx, (metric, title) in enumerate(metrics):
        ax = axes[idx//2, idx%2]
        ax.plot(plot_data['time_hour'], plot_data[metric], marker='o', linewidth=2)
        ax.set_title(title)
        ax.set_xlabel('Time (hours)')
        ax.set_ylabel(title)
        ax.grid(True, alpha=0.3)
        
        # Mark sepsis duration
        sepsis_period = plot_data[plot_data['sepsis'] == 1]
        if not sepsis_period.empty:
            start = sepsis_period['time_hour'].min()
            end = sepsis_period['time_hour'].max()
            ax.axvspan(start, end, alpha=0.3, color='red', label='Sepsis')
            ax.legend(loc='upper left')
    
    fig.tight_layout(rect=[0, 0.04, 1, 0.96], h_pad=3.0)
    plt.show()

# Plot timelines for each patient
for patient_id in df_data['patient_id'].unique():
    plot_patient_timeline(patient_id, df_data)