Improving the Accuracy of Otitis Media with Effusion Diagnosis in Pediatric Patients Using Deep Learning
Bioengineering · 20.11.2023 · 10 (11) · 1337
DOI: 10.3390/bioengineering10111337 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bioengineering, online erschienen am 20.11.2023. Verfasst von Jae-Hyuk Shim und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Biomedical Engineering, Gil Medical Center, Gachon University College of Medicine, Incheon 21565, Republic of Korea. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: GRRC, Korea Medical Device Development Fund grant funded by the Korea government (MSIT, MOTIE, and MFDS), Korea Medical Device Development Fund.
Abstract (Originalsprache)
Otitis media with effusion (OME), primarily seen in children aged 2 years and younger, is characterized by the presence of fluid in the middle ear, often resulting in hearing loss and aural fullness. While deep learning networks have been explored to aid OME diagnosis, prior work did not often specify if pediatric images were used for training, causing uncertainties about their clinical relevance, especially due to important distinctions between the tympanic membranes of small children and adults. We trained cross-validated ResNet50, DenseNet201, InceptionV3, and InceptionResNetV2 models on 1150 pediatric tympanic membrane images from otoendoscopes to classify OME. When assessed using a separate dataset of 100 pediatric tympanic membrane images, the models achieved mean accuracies of 92.9% (ResNet50), 97.2% (DenseNet201), 96.0% (InceptionV3), and 94.8% (InceptionResNetV2), compared to the seven otolaryngologists that achieved accuracies between 84.0% and 69.0%. The results showed that even the worst-performing model trained on fold 3 of InceptionResNetV2 with an accuracy of 88.0% exceeded the accuracy of the highest-performing otolaryngologist at 84.0%. Our findings suggest that these specifically trained deep learning models can potentially enhance the clinical diagnosis of OME using pediatric otoendoscopic tympanic membrane images.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Biomedical Engineering, Gil Medical Center, Gachon University College of Medicine, Incheon 21565, Republic of Korea.
- Department of Otorhinolaryngology-Head and Neck Surgery, Gil Medical Center, Gachon University College of Medicine, Incheon 21565, Republic of Korea.
- Department of Otorhinolaryngology-Head and Neck Surgery, Seoul National University Bundang Hospital, Seongnam 13620, Republic of Korea.
ORCID-Profile: Woongsang Sunwoo, Kwang Gi Kim
Förderung
- GRRC
- Korea Medical Device Development Fund grant funded by the Korea government (MSIT, MOTIE, and MFDS)
- Korea Medical Device Development Fund
- GRRC program of Gyeonggi province
Zitieren
APA
Shim, J. H., Sunwoo, W., Choi, B. Y., Kim, K. G., & Kim, Y. J. (2023). Improving the Accuracy of Otitis Media with Effusion Diagnosis in Pediatric Patients Using Deep Learning. Bioengineering, 10(11), 1337. https://doi.org/10.3390/bioengineering10111337
Vancouver
Shim JH, Sunwoo W, Choi BY, Kim KG, Kim YJ. Improving the Accuracy of Otitis Media with Effusion Diagnosis in Pediatric Patients Using Deep Learning. Bioengineering. 2023;10(11):1337. doi:10.3390/bioengineering10111337
BibTeX
@article{Shim2023,
author = {Shim, Jae-Hyuk and Sunwoo, Woongsang and Choi, Byung Yoon and Kim, Kwang Gi and Kim, Young Jae},
title = {Improving the Accuracy of Otitis Media with Effusion Diagnosis in Pediatric Patients Using Deep Learning},
journal = {Bioengineering},
year = {2023},
volume = {10},
number = {11},
pages = {1337},
doi = {10.3390/bioengineering10111337}
}
EUVIMED