Using deep learning to detect diabetic retinopathy on handheld non-mydriatic retinal images acquired by field workers in community settings

TY - JOUR

T1 - Using deep learning to detect diabetic retinopathy on handheld non-mydriatic retinal images acquired by field workers in community settings

AU - for the SMART India Study Group

AU - Nunez do Rio, Joan M.

AU - Nderitu, Paul

AU - Raman, Rajiv

AU - Rajalakshmi, Ramachandran

AU - Kim, Ramasamy

AU - Rani, Padmaja K.

AU - Sivaprasad, Sobha

AU - Bergeles, Christos

AU - Raman, Rajiv

AU - Bhende, Pramod

AU - Surya, Janani

AU - Gopal, Lingam

AU - Ramakrishnan, Radha

AU - Roy, Rupak

AU - Das, Supita

AU - Manayath, George

AU - Vignesh, T. P.

AU - Anantharaman, Giridhar

AU - Gopalakrishnan, Mahesh

AU - Natarajan, Sundaram

AU - Krishnan, Radhika

AU - Mani, Sheena Liz

AU - Agarwal, Manisha

AU - Behera, Umesh

AU - Bhattacharjee, Harsha

AU - Barman, Manabjyoti

AU - Sen, Alok

AU - Saxena, Moneesh

AU - Sil, Asim K.

AU - Chakabarty, Subhratanu

AU - Cherian, Thomas

AU - Jitesh, Reesha

AU - Naigaonkar, Rushikesh

AU - Desai, Abishek

AU - Kulkarni, Sucheta

N1 - Funding Information: The authors would like to thank all the SMART India collaborators, including fieldworkers, each centre staff, and reading centre staff, for the study database and the study participants. Publisher Copyright: © 2023, The Author(s).

PY - 2023/12/1

Y1 - 2023/12/1

N2 - Diabetic retinopathy (DR) at risk of vision loss (referable DR) needs to be identified by retinal screening and referred to an ophthalmologist. Existing automated algorithms have mostly been developed from images acquired with high cost mydriatic retinal cameras and cannot be applied in the settings used in most low- and middle-income countries. In this prospective multicentre study, we developed a deep learning system (DLS) that detects referable DR from retinal images acquired using handheld non-mydriatic fundus camera by non-technical field workers in 20 sites across India. Macula-centred and optic-disc-centred images from 16,247 eyes (9778 participants) were used to train and cross-validate the DLS and risk factor based logistic regression models. The DLS achieved an AUROC of 0.99 (1000 times bootstrapped 95% CI 0.98–0.99) using two-field retinal images, with 93.86 (91.34–96.08) sensitivity and 96.00 (94.68–98.09) specificity at the Youden’s index operational point. With single field inputs, the DLS reached AUROC of 0.98 (0.98–0.98) for the macula field and 0.96 (0.95–0.98) for the optic-disc field. Intergrader performance was 90.01 (88.95–91.01) sensitivity and 96.09 (95.72–96.42) specificity. The image based DLS outperformed all risk factor-based models. This DLS demonstrated a clinically acceptable performance for the identification of referable DR despite challenging image capture conditions.

AB - Diabetic retinopathy (DR) at risk of vision loss (referable DR) needs to be identified by retinal screening and referred to an ophthalmologist. Existing automated algorithms have mostly been developed from images acquired with high cost mydriatic retinal cameras and cannot be applied in the settings used in most low- and middle-income countries. In this prospective multicentre study, we developed a deep learning system (DLS) that detects referable DR from retinal images acquired using handheld non-mydriatic fundus camera by non-technical field workers in 20 sites across India. Macula-centred and optic-disc-centred images from 16,247 eyes (9778 participants) were used to train and cross-validate the DLS and risk factor based logistic regression models. The DLS achieved an AUROC of 0.99 (1000 times bootstrapped 95% CI 0.98–0.99) using two-field retinal images, with 93.86 (91.34–96.08) sensitivity and 96.00 (94.68–98.09) specificity at the Youden’s index operational point. With single field inputs, the DLS reached AUROC of 0.98 (0.98–0.98) for the macula field and 0.96 (0.95–0.98) for the optic-disc field. Intergrader performance was 90.01 (88.95–91.01) sensitivity and 96.09 (95.72–96.42) specificity. The image based DLS outperformed all risk factor-based models. This DLS demonstrated a clinically acceptable performance for the identification of referable DR despite challenging image capture conditions.

UR - http://www.scopus.com/inward/record.url?scp=85146910939&partnerID=8YFLogxK

U2 - 10.1038/s41598-023-28347-z

DO - 10.1038/s41598-023-28347-z

M3 - Article

C2 - 36697482

AN - SCOPUS:85146910939

SN - 2045-2322

VL - 13

JO - Scientific Reports

JF - Scientific Reports

IS - 1

M1 - 1392

ER -