Can a Rule-Based Expert System Diagnose Nasal Obstruction from Nasoendoscopy Videos?

Area of Science:

• Otolaryngology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Nasal obstruction presents diverse etiologies necessitating specialized endoscopic evaluation.
• Access to Ear, Nose, and Throat (ENT) specialists can be limited in certain healthcare settings.
• A rule-based expert system (RB-ES) offers a potential solution to assist in diagnosing nasal conditions.

Purpose of the Study:

• To evaluate the diagnostic performance of an RB-ES for common nasal obstruction conditions.
• To compare the RB-ES diagnostic accuracy against experienced ENT specialists.
• To assess the system's utility in settings with restricted access to specialist care.

Main Methods:

• A prospective cohort study involving 71 participants (65 patients, 6 controls).
• Nasal endoscopy was performed before and after decongestion.
• Diagnoses were established by four ENT specialists and compared with RB-ES outputs.

Main Results:

• The RB-ES exhibited comparable sensitivity to ENT specialists across all evaluated conditions (p > 0.05).
• Highest sensitivity was observed for diagnosing chronic rhinosinusitis with nasal polyps (CRSwNP).
• Overall specificity of the RB-ES remained high, matching specialist performance for CRSwNP.

Conclusions:

• The RB-ES demonstrates potential as a diagnostic aid for nasal obstruction, particularly CRSwNP.
• Further validation through expanded datasets and integration of artificial intelligence is recommended.
• RB-ES may improve diagnostic capabilities in resource-limited environments.