Identification of ocular refraction based on deep learning algorithm as a novel retinoscopy method
Tóm tắt
The evaluation of refraction is indispensable in ophthalmic clinics, generally requiring a refractor or retinoscopy under cycloplegia. Retinal fundus photographs (RFPs) supply a wealth of information related to the human eye and might provide a promising approach that is more convenient and objective. Here, we aimed to develop and validate a fusion model-based deep learning system (FMDLS) to identify ocular refraction via RFPs and compare with the cycloplegic refraction. In this population-based comparative study, we retrospectively collected 11,973 RFPs from May 1, 2020 to November 20, 2021. The performance of the regression models for sphere and cylinder was evaluated using mean absolute error (MAE). The accuracy, sensitivity, specificity, area under the receiver operating characteristic curve, and F1-score were used to evaluate the classification model of the cylinder axis. Overall, 7873 RFPs were retained for analysis. For sphere and cylinder, the MAE values between the FMDLS and cycloplegic refraction were 0.50 D and 0.31 D, representing an increase of 29.41% and 26.67%, respectively, when compared with the single models. The correlation coefficients (r) were 0.949 and 0.807, respectively. For axis analysis, the accuracy, specificity, sensitivity, and area under the curve value of the classification model were 0.89, 0.941, 0.882, and 0.814, respectively, and the F1-score was 0.88. The FMDLS successfully identified the ocular refraction in sphere, cylinder, and axis, and showed good agreement with the cycloplegic refraction. The RFPs can provide not only comprehensive fundus information but also the refractive state of the eye, highlighting their potential clinical value.
Tài liệu tham khảo
citation_journal_title=Lancet Glob Health; citation_title=The lancet global health commission on global eye health: vision beyond 2020; citation_author=MJ Burton, J Ramke, AP Marques; citation_volume=9; citation_issue=4; citation_publication_date=2021; citation_pages=e489-e551; citation_doi=10.1016/S2214-109X(20)30488-5; citation_id=CR1
citation_journal_title=Lancet Glob Health; citation_title=Global causes of blindness and distance vision impairment 1990–2020: a systematic review and meta-analysis; citation_author=SR Flaxman, R Bourne, S Resnikoff; citation_volume=5; citation_issue=12; citation_publication_date=2017; citation_pages=e1221-e1234; citation_doi=10.1016/S2214-109X(17)30393-5; citation_id=CR2
citation_journal_title=Annu Rev Vis Sci; citation_title=Origins of refractive errors: environmental and genetic factors; citation_author=EN Harb, CF Wildsoet; citation_volume=5; citation_publication_date=2019; citation_pages=47-72; citation_doi=10.1146/annurev-vision-091718-015027; citation_id=CR3
citation_journal_title=Ophthalmology; citation_title=Potential lost productivity resulting from the global burden of myopia: systematic review, meta-analysis, and modeling; citation_author=KS Naidoo, TR Fricke, KD Frick; citation_volume=126; citation_issue=3; citation_publication_date=2019; citation_pages=338-346; citation_doi=10.1016/j.ophtha.2018.10.029; citation_id=CR4
citation_journal_title=Acta Ophthalmol; citation_title=Cycloplegic refraction is the gold standard for epidemiological studies; citation_author=IG Morgan, R Iribarren, A Fotouhi; citation_volume=93; citation_issue=6; citation_publication_date=2015; citation_pages=581-585; citation_doi=10.1111/aos.12642; citation_id=CR5
citation_journal_title=Ophthalmology; citation_title=Accuracy of autorefraction in children: a report by the American Academy of Ophthalmology; citation_author=LB Wilson, M Melia, RT Kraker; citation_volume=127; citation_issue=9; citation_publication_date=2020; citation_pages=1259-1267; citation_doi=10.1016/j.ophtha.2020.03.004; citation_id=CR6
citation_journal_title=Transl Vis Sci Technol; citation_title=Does the accuracy and repeatability of refractive error estimates depend on the measurement principle of autorefractors?; citation_author=D Padhy, SR Bharadwaj, S Nayak; citation_volume=10; citation_issue=1; citation_publication_date=2021; citation_pages=2; citation_doi=10.1167/tvst.10.1.2; citation_id=CR7
citation_journal_title=Prog Retin Eye Res; citation_title=The epidemics of myopia: aetiology and prevention; citation_author=IG Morgan, AN French, RS Ashby; citation_volume=62; citation_publication_date=2018; citation_pages=134-149; citation_doi=10.1016/j.preteyeres.2017.09.004; citation_id=CR8
citation_journal_title=Invest Ophthalmol Vis Sci; citation_title=Morphological characteristics of the optic nerve head and choroidal thickness in high myopia; citation_author=G Hu, Q Chen, X Xu; citation_volume=61; citation_issue=4; citation_publication_date=2020; citation_pages=46; citation_doi=10.1167/iovs.61.4.46; citation_id=CR9
citation_journal_title=Ophthalmology; citation_title=Myopia-related optic disc and retinal changes in adolescent children from singapore; citation_author=C Samarawickrama, P Mitchell, L Tong; citation_volume=118; citation_issue=10; citation_publication_date=2011; citation_pages=2050-2057; citation_doi=10.1016/j.ophtha.2011.02.040; citation_id=CR10
citation_journal_title=Invest Ophthalmol Vis Sci; citation_title=Quantification of retinal nerve fiber and retinal artery trajectories using second-order polynomial equation and its association with axial length; citation_author=T Yamashita, T Sakamoto, H Terasaki; citation_volume=55; citation_issue=8; citation_publication_date=2014; citation_pages=5176-5182; citation_doi=10.1167/iovs.14-14105; citation_id=CR11
citation_journal_title=Sci Rep; citation_title=Machine learning prediction of pathologic myopia using tomographic elevation of the posterior sclera; citation_author=YC Kim, DJ Chang, SJ Park; citation_volume=11; citation_issue=1; citation_publication_date=2021; citation_pages=6950; citation_doi=10.1038/s41598-021-85699-0; citation_id=CR12
citation_journal_title=Nat Med; citation_title=A guide to deep learning in healthcare; citation_author=A Esteva, A Robicquet, B Ramsundar; citation_volume=25; citation_issue=1; citation_publication_date=2019; citation_pages=24-29; citation_doi=10.1038/s41591-018-0316-z; citation_id=CR13
citation_journal_title=Nature; citation_title=Dermatologist-level classification of skin cancer with deep neural networks; citation_author=A Esteva, B Kuprel, R Novoa; citation_volume=542; citation_issue=7639; citation_publication_date=2017; citation_pages=115-118; citation_doi=10.1038/nature21056; citation_id=CR14
citation_journal_title=BJR Open; citation_title=Deep learning in breast imaging; citation_author=A Bhowmik, S Eskreis-Winkler; citation_volume=4; citation_issue=1; citation_publication_date=2022; citation_pages=20210060; citation_id=CR15
citation_journal_title=Am J Ophthalmol; citation_title=Applying machine learning techniques in nomogram prediction and analysis for SMILE treatment; citation_author=T Cui, Y Wang, S Ji; citation_volume=210; citation_publication_date=2020; citation_pages=71-77; citation_doi=10.1016/j.ajo.2019.10.015; citation_id=CR16
citation_journal_title=Eur Radiol; citation_title=Deep learning radiomics of ultrasonography can predict response to neoadjuvant chemotherapy in breast cancer at an early stage of treatment: a prospective study; citation_author=J Gu, T Tong, C He; citation_volume=32; citation_publication_date=2022; citation_pages=2099-2109; citation_doi=10.1007/s00330-021-08293-y; citation_id=CR17
citation_journal_title=Gastric Cancer; citation_title=Endoscopic three-categorical diagnosis of Helicobacter pylori infection using linked color imaging and deep learning: a single-center prospective study (with video); citation_author=H Nakashima, H Kawahira, H Kawachi; citation_volume=23; citation_publication_date=2020; citation_pages=1033-1040; citation_doi=10.1007/s10120-020-01077-1; citation_id=CR18
citation_journal_title=Comput Methods Programs Biomed; citation_title=Neuroimaging and deep learning for brain stroke detection—a review of recent advancements and future prospects; citation_author=R Karthik, R Menaka, A Johnson; citation_volume=197; citation_publication_date=2020; citation_pages=105-728; citation_doi=10.1016/j.cmpb.2020.105728; citation_id=CR19
citation_journal_title=J Neurol Neurosurg Psychiatry; citation_title=Retinal imaging in Alzheimer’s disease; citation_author=CY Cheung, V Mok, PJ Foster; citation_volume=92; citation_issue=9; citation_publication_date=2021; citation_pages=983-994; citation_doi=10.1136/jnnp-2020-325347; citation_id=CR20
citation_journal_title=Nat Biomed Eng; citation_title=Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning; citation_author=R Poplin, AV Varadarajan, K Blumer; citation_volume=2; citation_issue=3; citation_publication_date=2018; citation_pages=158-164; citation_doi=10.1038/s41551-018-0195-0; citation_id=CR21
citation_journal_title=Biomed Eng Online; citation_title=Deep learning for predicting refractive error from multiple photorefraction images; citation_author=D Xu, S Ding, T Zheng; citation_volume=21; citation_issue=1; citation_publication_date=2022; citation_pages=55; citation_doi=10.1186/s12938-022-01025-3; citation_id=CR22
citation_journal_title=JMIR Med Inform; citation_title=Deep learning-based prediction of refractive error using photorefraction images captured by a smartphone: model development and validation study; citation_author=J Chun, Y Kim, KY Shin; citation_volume=8; citation_issue=5; citation_publication_date=2020; citation_doi=10.2196/16225; citation_id=CR23
citation_journal_title=Invest Ophthalmol Vis Sci; citation_title=Deep learning for predicting refractive error from retinal fundus images; citation_author=AV Varadarajan, R Poplin, K Blumer; citation_volume=59; citation_issue=7; citation_publication_date=2018; citation_pages=2861-2868; citation_doi=10.1167/iovs.18-23887; citation_id=CR24
citation_journal_title=Ann Transl Med; citation_title=Automatic identification of myopia based on ocular appearance images using deep learning; citation_author=Y Yang, R Li, D Lin; citation_volume=8; citation_issue=11; citation_publication_date=2020; citation_pages=705; citation_doi=10.21037/atm.2019.12.39; citation_id=CR25
citation_journal_title=Front Med; citation_title=Prediction of refractive error based on ultrawide field images with deep learning models in myopia patients; citation_author=D Yang, M Li, W Li; citation_volume=9; citation_publication_date=2022; citation_doi=10.3389/fmed.2022.834281; citation_id=CR26
citation_journal_title=J Thorac Oncol; citation_title=Receiver operating characteristic curve in diagnostic test assessment; citation_author=JN Mandrekar; citation_volume=5; citation_issue=9; citation_publication_date=2010; citation_pages=1315-1316; citation_doi=10.1097/JTO.0b013e3181ec173d; citation_id=CR27
citation_journal_title=Ophthalmology; citation_title=Noncycloplegic compared with cycloplegic refraction in a Chicago school-aged population; citation_author=X Guo, AF Shakarchi, SS Block; citation_volume=129; citation_issue=7; citation_publication_date=2022; citation_pages=813-820; citation_doi=10.1016/j.ophtha.2022.02.027; citation_id=CR28
citation_journal_title=Nature; citation_title=Deep learning; citation_author=Y LeCun, Y Bengio, G Hinton; citation_volume=521; citation_issue=7553; citation_publication_date=2015; citation_pages=436-444; citation_doi=10.1038/nature14539; citation_id=CR29
citation_journal_title=Int J Numer Method Biomed Eng; citation_title=A method for the automatic detection of myopia in Optos fundus images based on deep learning; citation_author=Z Shi, T Wang, Z Huang; citation_volume=37; citation_issue=6; citation_publication_date=2021; citation_doi=10.1002/cnm.3460; citation_id=CR30
citation_journal_title=Med Image Anal; citation_title=A review of machine learning methods for retinal blood vessel segmentation and artery/vein classification; citation_author=MRK Mookiah, S Hogg, TJ MacGillivray; citation_volume=68; citation_publication_date=2021; citation_pages=101-905; citation_doi=10.1016/j.media.2020.101905; citation_id=CR31
citation_journal_title=Pattern Recognit; citation_title=Automated segmentation of the optic disc from fundus images using an asymmetric deep learning network; citation_author=L Wang, J Gu, Y Chen; citation_volume=112; citation_publication_date=2021; citation_pages=107-810; citation_doi=10.1016/j.patcog.2020.107810; citation_id=CR32
citation_journal_title=Br J Ophthalmol; citation_title=Correlation between the pattern of myopic fundal changes and the axis of astigmatism of the eye; citation_author=R Ahmad, MA Al-Aqaba, U Fares; citation_volume=94; citation_issue=3; citation_publication_date=2010; citation_pages=307-310; citation_doi=10.1136/bjo.2009.161794; citation_id=CR33
citation_journal_title=Diagn Pathol; citation_title=The influence of corneal astigmatism on retinal nerve fiber layer thickness and optic nerve head parameter measurements by spectral-domain optical coherence tomography; citation_author=L Lin, Z Jun, H Hui; citation_volume=7; citation_publication_date=2012; citation_pages=55; citation_doi=10.1186/1746-1596-7-55; citation_id=CR34
citation_journal_title=Cornea; citation_title=Age-related changes in astigmatism and potential causes; citation_author=H Namba, A Sugano, T Murakami; citation_volume=39; citation_issue=Suppl 1; citation_publication_date=2020; citation_pages=S34-S38; citation_doi=10.1097/ICO.0000000000002507; citation_id=CR35
citation_journal_title=Invest Ophthalmol Vis Sci; citation_title=IMI—defining and classifying myopia: a proposed set of standards for clinical and epidemiologic studies; citation_author=DI Flitcroft, M He, JB Jonas; citation_volume=60; citation_issue=3; citation_publication_date=2019; citation_pages=M20-M30; citation_doi=10.1167/iovs.18-25957; citation_id=CR36
citation_journal_title=Br J Ophthalmol; citation_title=Cycloplegic autorefraction versus subjective refraction: the Tehran eye study; citation_author=H Hashemi, M Khabazkhoob, A Asharlous; citation_volume=100; citation_issue=8; citation_publication_date=2016; citation_pages=1122-1127; citation_doi=10.1136/bjophthalmol-2015-307871; citation_id=CR37
citation_journal_title=Lancet Digit Health; citation_title=Application of comprehensive artificial intelligence retinal expert (CARE) system: a national real-world evidence study; citation_author=D Lin, J Xiong, C Liu; citation_volume=3; citation_issue=8; citation_publication_date=2021; citation_pages=e486-e495; citation_doi=10.1016/S2589-7500(21)00086-8; citation_id=CR38
citation_journal_title=BMJ Open; citation_title=Developing a reporting guideline for artificial intelligence-centred diagnostic test accuracy studies: the STARD-AI protocol; citation_author=V Sounderajah, H Ashrafian, RM Golub; citation_volume=11; citation_issue=6; citation_publication_date=2021; citation_doi=10.1136/bmjopen-2020-047709; citation_id=CR39
citation_journal_title=Nat Rev Dis Prim; citation_title=Myopia; citation_author=PN Baird, SM Saw, C Lanca; citation_volume=6; citation_issue=1; citation_publication_date=2020; citation_pages=99; citation_doi=10.1038/s41572-020-00231-4; citation_id=CR40
citation_journal_title=IEEE Trans Pattern Anal Mach Intell; citation_title=Focal loss for dense object detection; citation_author=TY Lin, P Goyal, R Girshick; citation_volume=42; citation_issue=2; citation_publication_date=2020; citation_pages=318-327; citation_doi=10.1109/TPAMI.2018.2858826; citation_id=CR41
citation_journal_title=Lancet; citation_title=Statistical methods for assessing agreement between two methods of clinical measurement; citation_author=JM Bland, DG Altman; citation_volume=1; citation_issue=8476; citation_publication_date=1986; citation_pages=307-310; citation_doi=10.1016/S0140-6736(86)90837-8; citation_id=CR42