Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Nov;281(11):5815-5821.
doi: 10.1007/s00405-024-08809-4. Epub 2024 Jul 13.

Artificial intelligence for automatic detection and segmentation of nasal polyposis: a pilot study

Vittorio Rampinelli #  1 , Alberto Paderno #  2 , Carlo Conti  3 , Gabriele Testa  3 , Claudia Lodovica Modesti  3 , Edoardo Agosti  4 , Isabelle Dohin  3 , Tommaso Saccardo  5 , Alessandro Vinciguerra  6 , Marco Ferrari  5 , Alberto Schreiber  3 , Davide Mattavelli  3 , Piero Nicolai  5 , Chris Holsinger  7 , Cesare Piazza  3
Affiliations

Artificial intelligence for automatic detection and segmentation of nasal polyposis: a pilot study

Vittorio Rampinelli et al. Eur Arch Otorhinolaryngol. 2024 Nov.

Abstract

Purpose: Accurate diagnosis and quantification of polyps and symptoms are pivotal for planning the therapeutic strategy of Chronic rhinosinusitis with nasal polyposis (CRSwNP). This pilot study aimed to develop an artificial intelligence (AI)-based image analysis system capable of segmenting nasal polyps from nasal endoscopy videos.

Methods: Recorded nasal videoendoscopies from 52 patients diagnosed with CRSwNP between 2019 and 2022 were retrospectively analyzed. Images extracted were manually segmented on the web application Roboflow. A dataset of 342 images was generated and divided into training (80%), validation (10%), and testing (10%) sets. The Ultralytics YOLOv8.0.28 model was employed for automated segmentation.

Results: The YOLOv8s-seg model consisted of 195 layers and required 42.4 GFLOPs for operation. When tested against the validation set, the algorithm achieved a precision of 0.91, recall of 0.839, and mean average precision at 50% IoU (mAP50) of 0.949. For the segmentation task, similar metrics were observed, including a mAP ranging from 0.675 to 0.679 for IoUs between 50% and 95%.

Conclusions: The study shows that a carefully trained AI algorithm can effectively identify and delineate nasal polyps in patients with CRSwNP. Despite certain limitations like the focus on CRSwNP-specific samples, the algorithm presents a promising complementary tool to existing diagnostic methods.

Keywords: Artificial intelligence; Data analysis; Diagnosis; Nasal polyps; Sinusitis.

PubMed Disclaimer

References

    1. Fokkens WJ, Viskens AS, Backer V et al (2023) EPOS/EUFOREA update on indication and evaluation of Biologics in Chronic Rhinosinusitis with nasal polyps. Rhinology 61(3):194–202. https://doi.org/10.4193/Rhin22.489 - DOI - PubMed
    1. Topol EJ (2019) High-performance medicine: the convergence of human and artificial intelligence. Nat Med 25(1):44–56. https://doi.org/10.1038/s41591-018-0300-7 - DOI - PubMed
    1. Litjens G, Kooi T, Bejnordi BE et al (2017) A survey on deep learning in medical image analysis. Med Image Anal 42:60–88. https://doi.org/10.1016/j.media.2017.07.005 - DOI - PubMed
    1. Xu J, Wang J, Bian X et al (2022) Deep learning for nasopharyngeal Carcinoma Identification using both white light and narrow-Band Imaging Endoscopy. Laryngoscope 132(5):999–1007. https://doi.org/10.1002/lary.29894 - DOI - PubMed
    1. Wu Q, Wang X, Liang G et al (2023) Advances in image-based Artificial Intelligence in Otorhinolaryngology-Head and Neck surgery: a systematic review. Otolaryngol Head Neck Surg 169(5):1132–1142. https://doi.org/10.1002/ohn.391 - DOI - PubMed

LinkOut - more resources