Predicting the Risk of Diabetic Neurodegenerative Disorders by Artificial Intelligence Tools Based on Retinal Imaging

Ospedale San Raffaele250 enrolled

Overview

Diabetic Retinopathy (DR) is the most frequent complication of diabetes, and its presence and severity are related to the appearance of both micro and macrovascular events. Risk profiles have been suggested as a major direction for research in diabetes, based on non- invasive retinal imaging evaluations. There has been promising evidence that artificial intelligence (AI) based on fundus photographs can detect clinical metrics and systemic conditions invisible to expert human observers. Notably, deep-learning (DL) convolutional neural networks (CNNs) developed for retinal photographs have been shown superior performance in the detection of DR compared with human assessment. The relationship between retinal vascular abnormalities and neurovascular complications of diabetes has been reported. The retina is a window to the body that allows a non-invasive observation of microvascular and neural tissues. However, in clinical practice there are no reported phenotypic indicators or reliable examinations to identify type 2 diabetic (T2D) patients with neurodegenerative/cognitive impairment. The presence of cognitive Impairment is a very frequent complication in diabetic patients, reported up to 60% of the diabetics when compared to only 11 % in the non-diabetics (OR of 8.78). Furthermore, AI based on retinal imaging has never been applied before to predict the onset and worsening of neurodegenerative/cognitive impairment of T2D in a real-world setting. The aim of this project is to develop trustworthy AI tools for predicting the risk of developing and progressing of neurodegenerative/cognitive diabetic impairment based on retinal images, in T2D population. For the development and validation of these tools, T2D patients will be enrolled from 4 well-established Italian centers. The proposal of this study is addressed to health care systems, in order to improve their consciousness about diabetic neurodegenerative/cognitive complications and reduce the related economic burden. Since the huge majority of these disorders remain undiagnosed, DINEURET will provide new cost-effective screening strategies to identify these patients. 4 centers will be involved: * 75 patients will be included in the IRCCS Ospedale San Raffaele, Milan; * 75 patients will be included in the IRCCS MultiMedica, Milan; * 50 patients will be included in the Ospedale Della Murgia "Fabio Perinei", Altamura; * 50 patients will be included in the Azienda Ospedaliero-Universitaria (AOUI) of Cagliari, Cagliari.

Study Type

OBSERVATIONAL

Enrollment

Eligibility

Sex: ALLMin age: 45 Years

Medical Language ↔ Plain English

Inclusion Criteria: 1. Male or female \> 45 years-old; 2. Diagnosis of type 2 DM; 3. No previous treatment for diabetic retinopathy; 4. Clear ocular media; 5. Ability to communicate well with the Investigator and to understand and comply with the requirements of the study; 6. Ability to provide written informed consent in accordance with institutional, local and national regulatory guidelines and to attend all study visits Exclusion Criteria: 1. Patients affected by other retinal disease than diabetic retinopathy; 2. Presence of diabetic macular edema; 3. Presence of proliferative diabetic retinopathy; 4. Any media opacities, including corneal opacity, cataract formation and hemorrhage in the vitreous body, which may interfere with viewing by the laser surgeon of the target structures in the study eye(s). Subject requiring cataract surgery in the next 12 months must be excluded; 5. Aphakic eye(s) with vitreous in the anterior chamber; 6. Neovascular glaucoma; 7. Glaucoma caused by congenital angle anomalies; 8. Open angle of less than 90º or extensive peripheral anterior and low synechia, present circumferentially around the corner; 9. Glaucoma secondary to active uveitis; 10. Any other ocular condition that would progress in the study period and confound visual acuity assessment a part from diabetic retinopathy; 11. Presence of idiopathic or autoimmune-associated uveitis; 12. Any ocular or systemic medication known to be toxic to the lens, retina or optic nerve; 13. Any intra-ocular surgery on a qualifying eye within three months prior to entry in the study; 14. Any prior thermal laser in the macula or intravitreal injections or panphotocoagulation; 15. History of vitrectomy, filtering surgery, corneal transplant or retinal detachment surgery; 16. Previous therapeutic radiation in the ocular region in either eye; 17. Participation in an investigational drug, biologic, or device study within 6 months prior to baseline \[Note: observational clinical studies solely involving over-the-counter vitamins, supplements, or diets are not exclusionary\]; 18. Has a serious medical illness that will prevent the subject from performing study activities (including cardiac, hepatic, renal, respiratory, endocrinologic, neurologic, or hematologic disease) or, in the judgement of the Investigator, is likely to require surgical intervention or hospitalization at any point during the study; 19. In the opinion of the Investigator, is unlikely to comply with the study protocol.

Outcomes

Primary Outcomes

Accuracy of AI based model

To evaluate the accuracy (sensibility and specificity) of the AI based model in the prediction of a worsening of neurodegenerative/cognitive impairment (defined as \> 2 point decrease at Montreal Cognitive Assessment scale) based on the retinal imaging acquired at the baseline.

Time frame: 30 August 2026

Secondary Outcomes

Reliability and reproducibility of the AI based model To characterize clinical phenotypes within T2D based on the risk of developing and worsening of cognitive decline.

To assess the reliability and reproducibility of the AI based model on different sub-samples (patients with different degrees of neurocognitive impairment).

Time frame: 30 August 2026

Clinical phenotypes of T2D patients

To select clinical phenotypes of T2D patients based on retinal imaging that are characterized by higher risk of developing and worsening of cognitive decline.

Time frame: 30 August 2026

Predicting the Risk of Diabetic Neurodegenerative Disorders by Artificial Intelligence Tools Based on Retinal Imaging

Overview

Conditions

Interventions

Eligibility

Locations (1)

Outcomes

Primary Outcomes

Secondary Outcomes

Central Contacts