Artificial Intelligence-Based Pathological Subtype Diagnosis of Nasal Polyps: A Multidimensional and Micro-Visualization Study - PubMed
5 days ago
- #Pathology
- #Nasal Polyps
- #Artificial Intelligence
- AI-based system (NPSS) developed for diagnosing nasal polyp (NP) subtypes using microscopic images (MI) and whole slide images (WSI).
- NPSS-MI and NPSS-WSI achieved accuracies of 90% and 91%, significantly reducing diagnostic time from 193 to 8 seconds and 10,450 to 250 seconds, respectively.
- Junior pathologists improved diagnostic accuracy from 50% to 89% using NPSS.
- 3D reconstruction (3DNP) enabled spatial quantification of inflammatory cells, revealing distinct patterns.
- NPSS-WSI prognostic model outperformed MI-based model with AUC of 86.64% vs. 79.81% (p = 0.039).
- Study utilized 2457 slides from 20 hospitals, with internal and external validation showing robust performance (F1-score: 0.809-0.792, IoU: 0.827-0.815).