Self-supervised learning of 3D structure from 2D OCT slices for retinal disease diagnosis on UK biobank scans

Basit Öğe Kaydı
dc.authorid0009-0007-4212-5484
dc.authorid0000-0001-6309-4524
dc.authorid0000-0002-8649-6013
dc.contributor.authorNazlı, Muhammet Serdaren_US
dc.contributor.authorTurkan, Yaseminen_US
dc.contributor.authorTek, Faik Borayen_US
dc.date.accessioned2026-03-06T11:35:11Z
dc.date.available2026-03-06T11:35:11Z
dc.date.issued2025-09-21
dc.departmentIşık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans Programıen_US
dc.departmentIşık University, School of Graduate Studies, Master’s Program in Computer Engineeringen_US
dc.descriptionUK Biobank data handling and experiments was performed by Yasemin Turkan as part of her PhD thesis, using the UK Biobank Resource under Application Number 82266. Serdar Nazli developed the model and conducted experiments on the OCTA-500 dataset. Computational resources were provided by the Turkish National High-Performance Computing Center (UHEM) under Grant Number 1017802024. This study was also supported by the Scientific and Technological Research Council of Turkey (T\u00FC BI?TAK) under Grant Number 122E509.en_US
dc.description.abstractThis study presents a self-supervised learning framework for retinal disease classification using Optical Coherence Tomography (OCT) scans. To balance the contextual richness of 3D volumes with the computational efficiency of 2D architectures, we introduce a quasi-3D input generation strategy. Each input is constructed by stacking three OCT slices, sampled from channel-specific Gaussian distributions centered on the volume midplane, and arranged in a standard three-channel 2D format compatible with existing pre-trained models. These quasi-3D images are used to pre-train a Vision Transformer (ViT-Base) via a Masked Autoencoder (MAE) with a shared masking pattern, encouraging the model to reconstruct masked regions by encoding anatomical continuity across slices. Pre-training is conducted on 10,000 unlabeled OCT volumes from the UK Biobank. The encoder is then fine-tuned on the OCTA-500 dataset for three-class and four-class retinal disease classification tasks, including macular degeneration and diabetic retinopathy. The model achieves 92.57% accuracy on the three-class task, matching the performance of RETFound while using over 150 times less pre-training data and a smaller backbone.en_US
dc.description.sponsorshipTürkiye Bilimsel ve Teknolojik Araştırma Kurumuen_US
dc.description.versionPublisher's Versionen_US
dc.identifier.citationNazlı, M. S., Turkan, Y. & Tek, F. B. (2025). Self-supervised learning of 3D structure from 2D OCT slices for retinal disease diagnosis on UK biobank scans. Paper presented at the International Conference on Computer Science and Engineering, UBMK, 930-934. doi:https://doi.org/10.1109/UBMK67458.2025.11206892en_US
dc.identifier.doi10.1109/UBMK67458.2025.11206892
dc.identifier.endpage934
dc.identifier.isbn9798331599751
dc.identifier.issn2521-1641
dc.identifier.issue2025
dc.identifier.scopus2-s2.0-105030861655
dc.identifier.scopusqualityN/A
dc.identifier.startpage930
dc.identifier.urihttps://hdl.handle.net/11729/7106
dc.identifier.urihttps://doi.org/10.1109/UBMK67458.2025.11206892
dc.indekslendigikaynakScopusen_US
dc.institutionauthorTurkan, Yaseminen_US
dc.institutionauthorid0000-0001-6309-4524
dc.language.isoenen_US
dc.peerreviewedYesen_US
dc.publicationstatusPublisheden_US
dc.publisherInstitute of Electrical and Electronics Engineers Inc.en_US
dc.relation.ispartofInternational Conference on Computer Science and Engineering, UBMKen_US
dc.relation.publicationcategoryKonferans Öğesi - Uluslararası - Öğrencien_US
dc.rightsinfo:eu-repo/semantics/closedAccessen_US
dc.subjectMasked autoencoderen_US
dc.subjectMedical image analysisen_US
dc.subjectOptical coherence tomographyen_US
dc.subjectRetinal diseaseen_US
dc.subjectSelf-supervised learningen_US
dc.subjectVision transformeren_US
dc.subjectClassification (of information)en_US
dc.subjectComputer aided diagnosisen_US
dc.subjectDiseasesen_US
dc.subjectEye protectionen_US
dc.subjectLearning algorithmsen_US
dc.subjectLearning systemsen_US
dc.subjectMedical image processingen_US
dc.subjectOphthalmologyen_US
dc.subjectOptical tomographyen_US
dc.subjectSupervised learningen_US
dc.subjectAuto encodersen_US
dc.subjectBiobanksen_US
dc.subjectCoherence tomographyen_US
dc.subjectDisease classificationen_US
dc.subjectOptical-en_US
dc.subjectQuasi-3Den_US
dc.subjectRetinal diseaseen_US
dc.subjectComputational efficiencyen_US
dc.titleSelf-supervised learning of 3D structure from 2D OCT slices for retinal disease diagnosis on UK biobank scansen_US
dc.typeConference Objecten_US
dspace.entity.typePublicationen_US

Dosyalar

Orijinal paket
Listeleniyor 1 - 1 / 1
Küçük Resim Yok
Ä°sim:
Self_Supervised_Learning_of_3D_Structure_from_2D_OCT_Slices_for_Retinal_Disease_Diagnosis_on_UK_Biobank_Scans.pdf
Boyut:
2.71 MB
Biçim:
Adobe Portable Document Format
Ä°ndir
Lisans paketi
Listeleniyor 1 - 1 / 1
Küçük Resim Yok
Ä°sim:
license.txt
Boyut:
1.17 KB
Biçim:
Item-specific license agreed upon to submission
Açıklama:
Ä°ndir

Koleksiyon

Öğrenci Yayınları Bildiri Koleksiyonu
Lisansüstü Eğitim Enstitüsü Diğer Yayınlar Koleksiyonu
Scopus İndeksli Yayınlar Koleksiyonu