Deep Learning-based Prediction of Axial Length Using Ultra-widefield Fundus Photography
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Purpose: To develop a deep learning model that can predict the axial lengths of eyes using ultra-widefield (UWF) fundus photography. Methods: We retrospectively enrolled patients who visited the ophthalmology clinic at the Seoul National University Hospitalbetween September 2018 and December 2021. Patients with axial length measurements and UWF images taken within 3months of axial length measurement were included in the study. The dataset was divided into a development set and a testset at an 8:2 ratio while maintaining an equal distribution of axial lengths (stratified splitting with binning). We used transferlearning-based on EfficientNet B3 to develop the model. We evaluated the model’s performance using mean absolute error(MAE), R-squared (R2), and 95% confidence intervals (CIs). We used vanilla gradient saliency maps to illustrate the regions predominantlyused by convolutional neural network. Results: In total, 8,657 UWF retinal fundus images from 3,829 patients (mean age, 63.98 ±15.25 years) were included in thestudy. The deep learning model predicted the axial lengths of the test dataset with MAE and R2 values of 0.744 mm (95% CI,0.709–0.779 mm) and 0.815 (95% CI, 0.785–0.840), respectively. The model’s accuracy was 73.7%, 95.9%, and 99.2% in prediction,with error margins of ±1.0, ±2.0, and ±3.0 mm, respectively. Conclusions: We developed a deep learning-based model for predicting the axial length from UWF images with good performance.
Abstract
Purpose: To develop a deep learning model that can predict the axial lengths of eyes using ultra-widefield (UWF) fundus photography. Methods: We retrospectively enrolled patients who visited the ophthalmology clinic at the Seoul National University Hospitalbetween September 2018 and December 2021. Patients with axial length measurements and UWF images taken within 3months of axial length measurement were included in the study. The dataset was divided into a development set and a testset at an 8:2 ratio while maintaining an equal distribution of axial lengths (stratified splitting with binning). We used transferlearning-based on EfficientNet B3 to develop the model. We evaluated the model’s performance using mean absolute error(MAE), R-squared (R2), and 95% confidence intervals (CIs). We used vanilla gradient saliency maps to illustrate the regions predominantlyused by convolutional neural network. Results: In total, 8,657 UWF retinal fundus images from 3,829 patients (mean age, 63.98 ±15.25 years) were included in thestudy. The deep learning model predicted the axial lengths of the test dataset with MAE and R2 values of 0.744 mm (95% CI,0.709–0.779 mm) and 0.815 (95% CI, 0.785–0.840), respectively. The model’s accuracy was 73.7%, 95.9%, and 99.2% in prediction,with error margins of ±1.0, ±2.0, and ±3.0 mm, respectively. Conclusions: We developed a deep learning-based model for predicting the axial length from UWF images with good performance.
참고문헌 (16)
- Olsen T, "Calculation of intraocular lens power : a review", Acta Ophthalmol Scand, 85, pp.472~485 (2007)
- Haarman AE, "The complications of myopia : a review and meta-analysis", Invest Ophthalmol Vis Sci, 61, pp.49~ (2020)
- Hashimoto S, "Association between axial length and myopic maculopathy : the Hisayama Study", Ophthalmol Retina, 3, pp.867~873 (2019)
- Oku Y, "Long axial length as risk factor for normal tension glaucoma", Graefes Arch Clin Exp Ophthalmol, 247, pp.781~787 (2009)
- Li M, "Retinal microvascular network and microcirculation assessments in high myopia", Am J Ophthalmol, 174, pp.56~67 (2017)
- Liu M, "Myopia-related stepwise and quadrant retinal microvascular alteration and its correlation with axial length", Eye(Lond), 35, pp.2196~2205 (2021)
- Yang Y, "Retinal microvasculature alteration in high myopia", Invest Ophthalmol Vis Sci, 57, pp.6020~6030 (2016)
- Moon JY, "Wide-field swept-source optical coherence tomography angiography in the assessment of retinal microvasculature and choroidal thickness in patients with myopia", Br J Ophthalmol, 107, pp.102~108 (2023)
- Dong L, "Deep Learning-based estimation of axial length and subfoveal choroidal thickness from color fundus photographs", Front Cell Dev Biol, 9, pp.653692~ (2021)
- Nussenblatt RB, "Standardization of vitreal inflammatory activity in intermediate and posterior uveitis", Ophthalmology, 92, pp.467~471 (1985)
- Tan M, "EfficientNet : rethinking model scaling for convolutional neural networks", PMLR, 97, pp.6105~6114 (2019)
- "Deep inside convolutional networks: isualizing image classification models and saliency maps" (2014)[URL]
- Jeong Y, "Ocular axial length prediction based on visual interpretation of retinal fundus images via deep neural network", IEEE J Sel Top Quantum Electron, 27, pp.1~7 (2020)
- Gordon RA, "Refractive development of the human eye", Arch Ophthalmol, 103, pp.785~789 (1985)
- Jagadeesh D, "Tessellated fundus appearance and its association with myopic refractive error", Clin Exp Optom, 102, pp.378~384 (2019)
- Lee K M, "Relationship between three-dimensional magnetic resonance imaging eyeball shape and optic nerve head morphology", Ophthalmology, 128, pp.532~544 (2021)