Publication: Design and Development of a Mobile Application for Accessible Pterygium Screening Using Pre-trained Deep Learning Models
8
0
Issued Date
2025-01-01
Resource Type
ISSN
03029743
eISSN
16113349
Scopus ID
2-s2.0-86000445875
Journal Title
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume
15432 LNAI
Start Page
348
End Page
359
Rights Holder(s)
SCOPUS
Bibliographic Citation
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol.15432 LNAI (2025) , 348-359
Suggested Citation
Withunchettanan T., Charoenruengkit W. Design and Development of a Mobile Application for Accessible Pterygium Screening Using Pre-trained Deep Learning Models. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Vol.15432 LNAI (2025) , 348-359. 359. doi:10.1007/978-981-96-0695-5_28 Retrieved from: https://hdl.handle.net/20.500.14740/20768
Author(s)
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
This research presents the development of a mobile application designed for the pre-screening of pterygium with the aim of enhancing early detection and improving model accuracy through continuous data collection. Leveraging the power of pre-trained deep learning model approach, the application integrates a smartphone camera with API service for image classification to provide an accessible and user-friendly tool for pterygium detection. The study evaluates several deep learning architectures, including EfficientNetB0, ResNet50, and VGG16, through 5-fold cross-validation and on a separate test set, assessing their precision, recall, and F1 scores. Evaluation classification results and evidence analysis with Grad-CAM function demonstrate that this approach offers a promising rapid solution for enhancing pterygium detection and allows for the data collection of anonymized patient images to continuously refine the model.