Best Practices for Using Neolens AI

Use Neolens safely, effectively, and meaningfully by following these best practices.

---

💡 1. Understand Your Model

• Know which version of the model is deployed (check model_id in metadata).
• Be aware of its training scope (e.g., adult vs pediatric data).
• Understand its known strengths and blind spots (see AI Limitations).

tip

Use the /model/info endpoint to retrieve current version details and changelogs.

---

🔬 2. Validate in Context

• Always validate AI outputs with domain experts.
• Compare with prior studies and clinical records.
• Evaluate performance across different devices and populations.

warning

Never use Neolens in clinical decision-making without human oversight.

---

⚙️ 3. Configure Responsibly

• Tune thresholds and filters to reduce false positives in your workflow.
• Document all configuration changes (e.g., min_confidence: 0.85).
• Use test datasets for performance evaluation after changes.

---

🔄 4. Monitor Continuously

• Set up logging and feedback collection loops.
• Track user edits vs. AI suggestions to detect drift or mismatch.
• Re-evaluate regularly as the model or data changes.

---

🛠️ 5. Integrate with Purpose

• Only automate where AI confidence is consistently high.
• Avoid “black box” integrations — always allow result inspection.
• Include fallback mechanisms and clinician override options.

---

🤝 6. Communicate Clearly

• Label AI-generated insights in UIs and reports.
• Inform patients and staff when AI is used in diagnosis or triage.
• Provide links to documentation and audit logs.

---

📊 7. Stay Compliant

• Align usage with your local data protection laws (e.g., GDPR, HIPAA).
• Apply RBAC for sensitive endpoints.
• Keep audit logs for all API calls involving patient data.