Jae-Seung Yun1, Jaesik Kim2, Sang-Hyuk Jung3, Seon-Ah Cha4, Seung-Hyun Ko4, Yu-Bae Ahn4, Hong-Hee Won5, Kyung-Ah Sohn6, Dokyoon Kim7. 1. Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA; Division of Endocrinology and Metabolism, Department of Internal Medicine, St. Vincent's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea. 2. Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA; Department of Computer Engineering, Ajou University, Suwon, Republic of Korea; Institute for Biomedical Informatics, University of Pennsylvania, Philadelphia, PA, USA. 3. Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA; Institute for Biomedical Informatics, University of Pennsylvania, Philadelphia, PA, USA; Samsung Advanced Institute for Health Sciences and Technology (SAIHST), Sungkyunkwan University, Samsung Medical Center, Seoul, Republic of Korea. 4. Division of Endocrinology and Metabolism, Department of Internal Medicine, St. Vincent's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea. 5. Samsung Advanced Institute for Health Sciences and Technology (SAIHST), Sungkyunkwan University, Samsung Medical Center, Seoul, Republic of Korea. 6. Department of Computer Engineering, Ajou University, Suwon, Republic of Korea; Department of Artificial Intelligence, Ajou University, Suwon, Republic of Korea. Electronic address: kasohn@ajou.ac.kr. 7. Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA; Institute for Biomedical Informatics, University of Pennsylvania, Philadelphia, PA, USA. Electronic address: dokyoon.kim@pennmedicine.upenn.edu.
Abstract
BACKGROUND AND AIMS: We aimed to develop and evaluate a non-invasive deep learning algorithm for screening type 2 diabetes in UK Biobank participants using retinal images. METHODS AND RESULTS: The deep learning model for prediction of type 2 diabetes was trained on retinal images from 50,077 UK Biobank participants and tested on 12,185 participants. We evaluated its performance in terms of predicting traditional risk factors (TRFs) and genetic risk for diabetes. Next, we compared the performance of three models in predicting type 2 diabetes using 1) an image-only deep learning algorithm, 2) TRFs, 3) the combination of the algorithm and TRFs. Assessing net reclassification improvement (NRI) allowed quantification of the improvement afforded by adding the algorithm to the TRF model. When predicting TRFs with the deep learning algorithm, the areas under the curve (AUCs) obtained with the validation set for age, sex, and HbA1c status were 0.931 (0.928-0.934), 0.933 (0.929-0.936), and 0.734 (0.715-0.752), respectively. When predicting type 2 diabetes, the AUC of the composite logistic model using non-invasive TRFs was 0.810 (0.790-0.830), and that for the deep learning model using only fundus images was 0.731 (0.707-0.756). Upon addition of TRFs to the deep learning algorithm, discriminative performance was improved to 0.844 (0.826-0.861). The addition of the algorithm to the TRFs model improved risk stratification with an overall NRI of 50.8%. CONCLUSION: Our results demonstrate that this deep learning algorithm can be a useful tool for stratifying individuals at high risk of type 2 diabetes in the general population.
BACKGROUND AND AIMS: We aimed to develop and evaluate a non-invasive deep learning algorithm for screening type 2 diabetes in UK Biobank participants using retinal images. METHODS AND RESULTS: The deep learning model for prediction of type 2 diabetes was trained on retinal images from 50,077 UK Biobank participants and tested on 12,185 participants. We evaluated its performance in terms of predicting traditional risk factors (TRFs) and genetic risk for diabetes. Next, we compared the performance of three models in predicting type 2 diabetes using 1) an image-only deep learning algorithm, 2) TRFs, 3) the combination of the algorithm and TRFs. Assessing net reclassification improvement (NRI) allowed quantification of the improvement afforded by adding the algorithm to the TRF model. When predicting TRFs with the deep learning algorithm, the areas under the curve (AUCs) obtained with the validation set for age, sex, and HbA1c status were 0.931 (0.928-0.934), 0.933 (0.929-0.936), and 0.734 (0.715-0.752), respectively. When predicting type 2 diabetes, the AUC of the composite logistic model using non-invasive TRFs was 0.810 (0.790-0.830), and that for the deep learning model using only fundus images was 0.731 (0.707-0.756). Upon addition of TRFs to the deep learning algorithm, discriminative performance was improved to 0.844 (0.826-0.861). The addition of the algorithm to the TRFs model improved risk stratification with an overall NRI of 50.8%. CONCLUSION: Our results demonstrate that this deep learning algorithm can be a useful tool for stratifying individuals at high risk of type 2 diabetes in the general population.
Authors: Donald M Lloyd-Jones; Yuling Hong; Darwin Labarthe; Dariush Mozaffarian; Lawrence J Appel; Linda Van Horn; Kurt Greenlund; Stephen Daniels; Graham Nichol; Gordon F Tomaselli; Donna K Arnett; Gregg C Fonarow; P Michael Ho; Michael S Lauer; Frederick A Masoudi; Rose Marie Robertson; Véronique Roger; Lee H Schwamm; Paul Sorlie; Clyde W Yancy; Wayne D Rosamond Journal: Circulation Date: 2010-01-20 Impact factor: 29.690
Authors: Daniel Shu Wei Ting; Carol Yim-Lui Cheung; Gilbert Lim; Gavin Siew Wei Tan; Nguyen D Quang; Alfred Gan; Haslina Hamzah; Renata Garcia-Franco; Ian Yew San Yeo; Shu Yen Lee; Edmund Yick Mun Wong; Charumathi Sabanayagam; Mani Baskaran; Farah Ibrahim; Ngiap Chuan Tan; Eric A Finkelstein; Ecosse L Lamoureux; Ian Y Wong; Neil M Bressler; Sobha Sivaprasad; Rohit Varma; Jost B Jonas; Ming Guang He; Ching-Yu Cheng; Gemmy Chui Ming Cheung; Tin Aung; Wynne Hsu; Mong Li Lee; Tien Yin Wong Journal: JAMA Date: 2017-12-12 Impact factor: 56.272
Authors: Carol Yim-Lui Cheung; Yingfeng Zheng; Wynne Hsu; Mong Li Lee; Qiangfeng Peter Lau; Paul Mitchell; Jie Jin Wang; Ronald Klein; Tien Yin Wong Journal: Ophthalmology Date: 2010-12-13 Impact factor: 12.079
Authors: Ryan Poplin; Avinash V Varadarajan; Katy Blumer; Yun Liu; Michael V McConnell; Greg S Corrado; Lily Peng; Dale R Webster Journal: Nat Biomed Eng Date: 2018-02-19 Impact factor: 25.671
Authors: M Kamran Ikram; Xueling Sim; Sim Xueling; Richard A Jensen; Mary Frances Cotch; Alex W Hewitt; M Arfan Ikram; Jie Jin Wang; Ronald Klein; Barbara E K Klein; Monique M B Breteler; Ning Cheung; Gerald Liew; Paul Mitchell; Andre G Uitterlinden; Fernando Rivadeneira; Albert Hofman; Paulus T V M de Jong; Cornelia M van Duijn; Linda Kao; Ching-Yu Cheng; Albert Vernon Smith; Nicole L Glazer; Thomas Lumley; Barbara McKnight; Bruce M Psaty; Fridbert Jonasson; Gudny Eiriksdottir; Thor Aspelund; Tamara B Harris; Lenore J Launer; Kent D Taylor; Xiaohui Li; Sudha K Iyengar; Quansheng Xi; Theru A Sivakumaran; David A Mackey; Stuart Macgregor; Nicholas G Martin; Terri L Young; Josh C Bis; Kerri L Wiggins; Susan R Heckbert; Christopher J Hammond; Toby Andrew; Samantha Fahy; John Attia; Elizabeth G Holliday; Rodney J Scott; F M Amirul Islam; Jerome I Rotter; Annie K McAuley; Eric Boerwinkle; E Shyong Tai; Vilmundur Gudnason; David S Siscovick; Johannes R Vingerling; Tien Y Wong Journal: PLoS Genet Date: 2010-10-28 Impact factor: 5.917
Authors: Chan-Hee Jung; Jang Won Son; Shinae Kang; Won Jun Kim; Hun-Sung Kim; Hae Soon Kim; Mihae Seo; Hye-Jung Shin; Seong-Su Lee; Su Jin Jeong; Yongin Cho; Seung Jin Han; Hyang Mi Jang; Mira Rho; Shinbi Lee; Mihyun Koo; Been Yoo; Jung-Wha Moon; Hye Young Lee; Jae-Seung Yun; Sun Young Kim; Sung Rae Kim; In-Kyung Jeong; Ji-Oh Mok; Kun Ho Yoon Journal: Diabetes Metab J Date: 2021-01-13 Impact factor: 5.376
Authors: Amit V Khera; Mark Chaffin; Krishna G Aragam; Mary E Haas; Carolina Roselli; Seung Hoan Choi; Pradeep Natarajan; Eric S Lander; Steven A Lubitz; Patrick T Ellinor; Sekar Kathiresan Journal: Nat Genet Date: 2018-08-13 Impact factor: 38.330