A review of AI in chronic rhinosinusitis (CRS) finds models now span diagnosis, endotyping, therapy selection, and surgery — but in recurrence prediction after endoscopic sinus surgery, they consistently converge on asthma, radiographic sinus opacification, and eosinophilic or neutrophilic inflammation as key signals. The review also notes that, across domains, complex algorithms do not consistently outperform simpler models, and limited external validation remains the main barrier to real-world use.

Why It Matters To Your Practice

• AI tools in CRS are moving beyond image interpretation into pre-visit prediction, endotype inference, treatment-response modeling, and surgical prognostication.

• For clinicians, the near-term value is likely risk stratification and decision support rather than fully autonomous diagnosis or treatment selection.

• Asthma appears to be a recurring marker in models forecasting postoperative recurrence, making it a practical clinical feature to watch when assessing risk.

Clinical Implications

• Pre-visit and imaging-based models may help identify likely CRS and support earlier triage before specialist evaluation.

• Deep-learning systems can interpret CT, nasal endoscopy, and histopathology, with some models inferring eosinophilic endotype noninvasively.

• In practice, AI may help refine which patients are more likely to respond to medical therapy, biologics, or endoscopic sinus surgery.

• Because simpler models can perform comparably to more complex ones, clinicians should favor tools with clear inputs, transparent performance, and external validation.

Insights

• The review describes four current capability areas: diagnosis, endotyping, therapeutics, and surgery.

• Across endotyping studies, AI repeatedly identified IL-5 and mixed type 2/neutrophilic inflammation as important biologic patterns.

• Emerging frontiers include multimodal foundation models, federated learning, and computer-vision-guided intraoperative navigation.

• The authors call for international consensus on how AI in CRS is developed and evaluated, reflecting ongoing concerns about bias, reproducibility, and generalizability.

The Bottom Line

• AI in CRS is broadening fast, but most tools are not yet ready to drive routine care independently.

• Right now, the most actionable takeaway is that asthma is a consistent recurrence flag in surgical prediction models, while external validation remains the key hurdle before wider adoption.