🛡️ RadShield - Prostate Cancer Radiotherapy Risk Assessment System

📋 Project Overview

RadShield is an integrated clinical decision support system designed for risk assessment and patient care management in prostate cancer radiotherapy. The system integrates multi-dimensional information including clinical factors, genomic markers, and dosimetric parameters to provide real-time radiation proctitis risk prediction and personalized treatment recommendations.

Core Values

• Precise Risk Assessment: Integrates multi-dimensional risk factors to provide acute and chronic toxicity risk stratification
• Clinical Decision Support: Automatically generates optimized treatment recommendations based on risk assessment results
• Patient Engagement: Tracks post-treatment symptom changes through electronic patient-reported outcomes (ePRO)
• Visual Analytics: Intuitive dashboards and charts to help clinicians quickly understand patient risk status

---

✨ Key Features

1. Clinical Decision Support Mode

Risk Assessment Engine

• Multi-dimensional Risk Factor Integration

  • Clinical factors: Diabetes, Inflammatory Bowel Disease (IBD), Anticoagulant use, Smoking history
  • Genomic markers: Mutation detection for ATM, NBN, LIG1, LIG4, PCNA, REV3L, POLH, XPC
  • Dosimetric parameters: Rectal V70, V50, and other DVH metrics
  • Treatment modalities: VMAT, IMRT, Proton Therapy

• Risk Stratification

  • Acute Toxicity Risk: Assesses risk during treatment and up to 3 months post-treatment
  • Late Toxicity Risk: Assesses long-term risk 6 months or more after treatment
  • Risk levels: Low risk (< 30%), Intermediate risk (30-60%), High risk (> 60%)

Visualization Dashboard

• Risk Gauge Charts

  • Real-time display of acute and chronic risk scores (0-100%)
  • Color coding: Green (low risk), Orange (intermediate risk), Red (high risk)
  • Alert threshold indicator (60% as high-risk threshold)

• Population Risk Positioning

  • Displays patient's risk distribution position among 1,200 prostate cancer radiotherapy patients
  • Normal distribution curve visualization
  • Model confidence display (87.4%)

• Multi-dimensional Risk Radar Chart

  • Five-dimensional assessment: Dosimetry, Genomics, Comorbidities, Medications, Age
  • Intuitive presentation of relative intensity across risk dimensions

Clinical Decision Support (CDS)

The system automatically generates the following recommendations based on risk assessment results:

• High-Risk Alerts

  • Recommend use of rectal spacer gel (SpaceOAR) to increase physical distance
  • Re-optimize treatment plan: Limit Rectum V70 < 15%

• Genomic Sensitivity Recommendations

  • When high radiosensitivity gene mutations are detected, consider reducing single-fraction dose or switching to proton therapy

• Bleeding Risk Management

  • When anticoagulants are used and acute risk > 40%, recommend cardiology consultation to assess temporary suspension of anticoagulants

Clinical Report Generation

• PDF Report Download
  • Includes patient information, risk stratification, key risk drivers, and optimization recommendations
  • Supports export to Electronic Medical Record (EMR) systems
  • Report content can be edited before generation

2. Patient-Reported Outcomes Mode (ePRO - Electronic Patient-Reported Outcomes)

Symptom Reporting Function

Patients can report the following symptoms through the system:

• Rectal Bleeding Status

  • None, Mild (on tissue), Severe (dripping)

• Pain Score

  • 0-10 scale (0 = no pain, 10 = most severe pain)

• Bowel Urgency

  • 0-10 scale (0 = none, 10 = very severe)

• Daily Bowel Movement Frequency

  • Numerical input (0-20 times)

Symptom Trend Tracking

• Historical Trend Charts

  • Pain score trend line chart
  • Bowel urgency trend line chart
  • Historical record table

• Risk Alerts

  • System automatically alerts when pain score ≥ 7 or bleeding is "severe"
  • Recommends early referral to colorectal surgery department

Personalized Patient Education

• Recommended Reading Content
  • Low-residue diet guidelines
  • Probiotic supplementation recommendations
  • Personalized patient education information based on patient symptoms

---

🏗️ System Architecture

risk_eval/
├── integrated_app.py      # Main application (Streamlit)
├── logic.py               # Core risk calculation logic
├── styles.py              # Custom CSS styles
├── test.py                # Python prototype version
├── test.html              # HTML prototype version
└── README.md              # This file

Core Module Descriptions

integrated_app.py

• Streamlit main application
• Integrates Clinical Decision Support and ePRO modes
• UI/UX design and interaction logic

logic.py

• calculate_risk(): Core risk calculation function
• create_population_chart(): Population risk positioning chart generation
• create_pdf(): PDF report generation engine

styles.py

• Custom CSS styles
• Modern UI design (gradient backgrounds, animation effects)

---

🚀 Installation and Usage

Environment Requirements

• Python 3.8+
• Streamlit
• Plotly
• NumPy
• Pandas
• FPDF (optional, for PDF report generation)

Installation Steps

1. Install Dependencies

pip install streamlit plotly numpy pandas fpdf

2. Launch Application

streamlit run risk_eval/integrated_app.py

Or use the full path:

streamlit run "/Users/vince/model test/risk_eval/integrated_app.py"

3. Open Browser

The system will automatically open in the default browser, typically at http://localhost:8501

Usage Workflow

Clinical Decision Support Mode

1. Select Mode

   • Choose "Clinical Decision Support" in the sidebar

2. Enter Patient Information

   • Clinical Profile: Enter patient name, ID, risk factors (diabetes, IBD, anticoagulants, smoking history)
   • Radiogenomics Panel: Check relevant gene mutations
   • Dosimetry (DVH): Select treatment modality, adjust Rectum V70 and V50 parameters

3. View Risk Assessment Results

   • Risk Dashboard: View acute and chronic risk scores
   • Population Study: View patient's risk position in the population
   • Analysis Detail: View multi-dimensional risk radar chart and key risk factors
   • Decision Report: Preview and download PDF report

Patient-Reported Outcomes Mode

1. Select Mode

   • Choose "Patient-Reported Outcomes (ePRO)" in the sidebar

2. Enter Patient ID

   • Enter patient ID in the sidebar

3. Complete Symptom Report

   • Fill in today's symptoms in the "Symptom Report (ePRO)" tab
   • Click the "Submit Report" button

4. View Historical Trends

   • View symptom change charts and historical records in the "Historical Trends" tab

---

📊 Risk Calculation Logic

Risk Score Calculation Formula

The system uses a weighted scoring approach, integrating multiple risk factors:

Base Scores

• Acute risk base score: 10
• Chronic risk base score: 5

Clinical Factor Weights

• IBD: Acute +25, Chronic +30
• Diabetes: Chronic +10
• Anticoagulants: Acute +15, Chronic +25

Genomic Factor Weights

• ATM Mutation: Acute +20, Chronic +25
• NBN Mutation: Acute +15, Chronic +20
• LIG1 Mutation: Acute +12, Chronic +18
• LIG4 Mutation: Acute +30, Chronic +40 (highest risk)
• PCNA Mutation: Acute +10, Chronic +15
• REV3L Mutation: Acute +18, Chronic +25
• POLH Mutation: Acute +16, Chronic +20
• XPC Mutation: Acute +14, Chronic +15

Dosimetric Factor Weights

• Rectum V70 > 20%: Chronic risk + (V70 - 20) × 2
• Rectum V50 > 50%: Acute risk + (V50 - 50) × 1.5

Treatment Modality Adjustment

• VMAT / IMRT: Multiplier 1.0 (standard)
• Proton Therapy: Multiplier 0.5 (risk reduction)

Age Factor

• Age > 75 years: Chronic risk +5
• Age > 70 years: Acute risk +3

Risk Level Determination

• Low Risk: Score < 30
• Intermediate Risk: 30 ≤ Score < 60
• High Risk: Score ≥ 60

---

🎨 Interface Features

Visual Design

• Modern UI: Gradient backgrounds, card-based design, smooth animation effects
• Blue Theme: Professional medical system color scheme
• Responsive Design: Supports different screen sizes

Interactive Experience

• Real-time Calculation: Risk assessment results update immediately after parameter input
• Visualization Charts: Plotly interactive charts with zoom and hover tooltips
• Multi-tab Design: Clear information architecture for easy navigation

---

⚠️ Important Notes

Research Use Statement

• This system is a Prototype version, primarily for research and demonstration purposes
• Genomic marker weights are currently set relatively low, pending prospective study validation
• Complete validation and review processes are required before clinical use

Data Privacy

• This system is a locally deployed version; data is not uploaded to external servers
• Recommended for use in environments compliant with HIPAA/GDPR regulations
• Data encryption and access control should be strengthened for actual deployment

Model Limitations

• Risk calculation model is based on literature and clinical experience, not a machine learning model
• Population risk positioning charts use simulated data (normal distribution), not actual hospital data
• Model confidence (87.4%) is a demonstration value

Technical Limitations

• PDF generation requires installation of the fpdf package
• If fpdf is not installed, the system will provide plain text report download
• ePRO historical records are currently stored in Session State; restarting the application will clear them

---

🔧 Advanced Configuration

Custom Risk Weights

To adjust risk factor weights, edit the calculate_risk() function in logic.py:

# Example: Adjust IBD risk weight
if patient_data.get('ibd', False):
    acute_score += 30  # Originally 25
    chronic_score += 35  # Originally 30

Custom Styles

To modify UI styles, edit the CSS settings in styles.py.

EMR System Integration

Currently, PDF reports can be manually downloaded and imported into EMR. Future API integration can enable automatic upload functionality.

---

📚 References

Related Concepts

• Concept 1: Application of rectal spacer gel (SpaceOAR)
• Concept 3: Optimization of DVH constraints
• Concept 4: Pre-treatment genomic marker risk prediction

Clinical Guidelines

• Rectal toxicity risk assessment in prostate cancer radiotherapy
• Application of genomics in radiosensitivity prediction
• Value of electronic patient-reported outcomes (ePRO) in cancer care

---

📄 License Information

This project is for education & research purposes. Please comply with relevant regulations and ethical guidelines.

---

🤝 Contributing

Suggestions and feedback are welcome. To contribute code, please contact the development team first.

---

📧 Contact Information

For any questions or suggestions, please contact through the project management system or directly with the development team.

---

Last Updated: 2024

System Status: Prototype Version