Hassan ON, Menten MJ, Bogunovic H, Schmidt-Erfurth U, Lotery A, Rueckert D (2021)
Publication Type: Conference contribution
Publication year: 2021
Publisher: IEEE Computer Society
Book Volume: 2021-April
Pages Range: 238-242
Conference Proceedings Title: Proceedings - International Symposium on Biomedical Imaging
Event location: Nice, FRA
ISBN: 9781665412469
DOI: 10.1109/ISBI48211.2021.9434107
Convolutional neural networks (CNNs) have achieved remarkable success in predicting clinical information and individuals' characteristics from medical images. Previous ophthalmological studies have suggested that age and sex have retinal manifestations that can be observed in retinal optical coherence tomography (OCT) scans. Following on these studies, we evaluated the use of three-dimensional CNNs for predicting the subject's age and sex directly from 3D retinal OCT scans. We also assessed the effect of the receptive field size on the model performance. In addition, we adopted a robust and simple bias-adjustment scheme for further performance enhancement of eye age prediction. We used a large dataset consisting of 66, 767 subjects with OCT scans from the UK Biobank data and evaluated our model on 10% of the dataset (i.e. 6, 676 subjects). An accurate prediction was obtained for age (mean absolute error (MAE): 3.3 years, coefficient of determination R2: 0.89) while an acceptable performance was achieved for sex (area under the curve (AUC): 0.86).
APA:
Hassan, O.N., Menten, M.J., Bogunovic, H., Schmidt-Erfurth, U., Lotery, A., & Rueckert, D. (2021). Deep learning prediction of age and sex from optical coherence tomography. In Proceedings - International Symposium on Biomedical Imaging (pp. 238-242). Nice, FRA: IEEE Computer Society.
MLA:
Hassan, Osama N., et al. "Deep learning prediction of age and sex from optical coherence tomography." Proceedings of the 18th IEEE International Symposium on Biomedical Imaging, ISBI 2021, Nice, FRA IEEE Computer Society, 2021. 238-242.
BibTeX: Download