[A comparative study of artificial intelligence nasal polyp classification based on whole-slide imaging and JESREC...

Objective: To explore the types and clinical characteristics of chronic rhinosinusitis with nasal polyps (CRSwNP) based on artificial intelligence and whole-slide imaging (WSI), and to explore the consistency of the diagnostic criteria of the Japanese epidemiological survey of refractory eosinophilic chronic rhinosinusitis (JESREC) in Chinese CRSwNP patients. Methods: The data of 136 patients with CRSwNP (101 males and 35 females, aging 14 to 70 years) who underwent endoscopic sinus surgery from 2018 to 2019 in the Department of Otorhinolaryngology Head and Neck Surgery, the Third Affiliated Hospital of Sun Yat-sen University were analysed retrospectively. The preoperative clinical characteristics of patients were collected, such as visual analogue scale (VAS) of nasal symptoms, peripheral blood inflammatory cell count, total immunoglobulin E (IgE), Lund-Kennedy score and Lund-Mackay score. The proportion of inflammatory cells such as eosinophils, lymphocytes, plasma cells and neutrophils were calculated on the WSI of each patient through artificial intelligence chronic rhinosinusitis evaluation platform 2.0 (AICEP 2.0), and the specific type of nasal polyps was then obtained as eosinophilic CRSwNP (eCRSwNP) or non-eosinophilic CRSwNP (non-eCRSwNP). In addition, the JESREC diagnostic criteria was used to classify the nasal polyps, and the classification results were compared with the current gold standard for nasal polyps diagnosis (pathological diagnosis based on WSI). The accuracy, sensitivity and specificity of the diagnostic criteria of JESREC were evaluated. The data were expressed in M (Q₁, Q₃) and statistically analyzed by SPSS 17.0. Results: There was no significant difference between eCRSwNP and non-eCRSwNP in age distribution, gender, time of onset, total VAS score, Lund-Kennedy score or Lund-Mackay score. However, there was a significant difference in the ratio of nasal polyp inflammatory cells (eosinophils 40.5% (22.8%, 54.7%) vs 2.5% (1.0%, 5.3%), neutrophils 0.3% (0.1%, 0.7%) vs 1.3% (0.5%, 3.6%), lymphocytes 49.9% (39.3%, 65.9%) vs 82.0% (72.8%, 87.5%), plasma cells 5.1% (3.6%, 10.5%) vs 13.0% (7.4%, 16.3%), χ² value was 9.91, 4.66, 8.28, 5.06, respectively, all P<0.05). In addition, eCRSwNP had a significantly higher level of proportion of allergic symptoms (nasal itching and sneezing), asthma, peripheral blood eosinophil and total IgE (all P<0.05). The overall accuracy, sensitivity and specificity of the JESREC diagnostic criteria was 74.3%, 81.3% and 64.3%, respectively. Conclusions: The eCRSwNP based on artificial intelligence and WSI has significant high level of allergic symptoms, asthma, peripheral blood eosinophils and total IgE, and the percentages of inflammatory cells in nasal polyps are different from that of non-eCRSwNP. The JESREC diagnostic criteria has good consistency in our research.