AI and Imaging in Managing Respiratory Disease Outbreaks: A Scoping Review

Overview

This review evaluates the role of radiology and artificial intelligence (AI) in managing respiratory disease outbreaks over the past two decades, focusing on diagnosis, therapy guidance, prognosis, and data sharing. It highlights the evolving use of imaging in early pandemic phases and the challenges and opportunities for AI to enhance preparedness for future respiratory threats.

Background

Respiratory disease outbreaks such as SARS, H1N1, MERS, and COVID-19 have been declared Public Health Emergencies of International Concern by the WHO. Medical imaging has been pivotal in diagnosing and managing these diseases, especially during early pandemic stages. AI has rapidly advanced in chest imaging, offering potential to detect outbreaks, forecast cases, and optimize clinical decisions. However, limitations in data diversity and quality have hindered AI applicability, underscoring the need for improved international data collaboration and algorithm development.

Data Highlights

The review identified timelines for pathogen identification, imaging findings, and AI applications across four major respiratory pandemics. For example, SARS spread to 29 countries, infecting 8,096 people and causing 774 deaths. Early imaging findings were incorporated into case definitions and therapy monitoring. AI studies were limited but included early attempts at automated detection using neural networks on multi-center datasets.

Key Findings

• Chest radiography was integral to early diagnosis and triage during SARS, sometimes detecting abnormalities before respiratory symptoms.
• Radiographic imaging guided therapy decisions and discharge criteria, although standardized therapies were lacking during early outbreaks.
• International data sharing efforts faced legal and procedural barriers, limiting rapid AI development and validation.
• Early AI applications for SARS included neural network models for automated detection on chest radiographs, but details and validation were insufficient.
• Time from pathogen identification to genome sequencing and PCR test development has shortened significantly, yet initial case identification remains challenging.
• Routine surveillance and data collection are critical for early outbreak detection and AI training.

Clinical Implications

Clinicians should recognize the value of chest imaging in early respiratory outbreak management for diagnosis and monitoring. Enhanced international collaboration and standardized data sharing are essential to develop robust AI tools that can support rapid response. Integrating AI with imaging may improve early detection and prognosis but requires high-quality, diverse datasets and clear algorithm validation.

Conclusion

Imaging has played a crucial role in managing respiratory pandemics, and AI holds promise to augment this role. Addressing data sharing challenges and improving AI algorithm development are key to enhancing preparedness for future respiratory disease outbreaks.

References

1. WHO/CDC/2003-2021 -- Pandemic Preparedness and Imaging in Respiratory Outbreaks
2. Levac et al 2010 -- Scoping Review Methodology
3. Early SARS Imaging Studies 2003-2005 -- Role of Radiology and AI