Artificial Intelligence (AI), Brain Tumors, Deep Learning, Machine Learning, Optic Atrophy, Optic Nerve Diseases, Optic Neuritis, Optic Neuropathy, Optic Neuropathy, Ischemic, Papilledema, Retinal Photograph
Conditions
Keywords
fundus imaging, automated diagnosis, multi-class classification, clinical decision support
Brief summary
The research team, recognized as a world leader in Artificial Intelligence for neuro-ophthalmology, has shown that it is possible to diagnose certain neuro-ophthalmologic or neurologic disorders from a single retinal fundus image (Milea et al, New England Journal of Medicine, 2020). However, clinical practice requires identifying a broader spectrum of diseases (inflammatory, ischemic, hereditary, neurodegenerative) within the same analysis. The main objective is to develop, through a new algorithm capable of classifying multiple disorders from a smaller set of conventional retinal images. This project meets a significant public health need: the global shortage of neuro-ophthalmologists. It aims to provide healthcare professionals with a rapid triage tool to detect serious and treatable conditions, enabling timely intervention. The study will include patients with clearly defined neuro-ophthalmologic or neurologic conditions, confirmed diagnoses, and retinal imaging. Clinical, paraclinical, and imaging data collected during standard care will be used, with strict anonymization according to legal and institutional requirements. Specific Objectives : 1. Evaluate the performance of a diagnostic classification algorithm trained on retinal images. 2. Assess the ability to detect multiple pathologies from a single retinal image. 3. Support the development of advanced computer vision tools for medical diagnostics.
Interventions
OTHERDeep learning algorithm applied on retrospectively collected color fundus photographs
Deep learning algorithm applied on retrospectively collected color fundus photographs
Sponsors
Fondation Ophtalmologique Adolphe de Rothschild
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Healthy volunteers
No
Inclusion criteria
* Patients with well-defined neuro-ophthalmologic or neurologic conditions, including different forms of optic neuropathies and various neurodegenerative diseases. * Patients with a robust reference diagnosis confirmed by clinical experts. * Patients with available retinal fundus images collected during routine care.
Exclusion criteria
* Patients without a confirmed diagnosis or unclear clinical classification. * Patients without retinal fundus images or with images that are completely unreadable. * Patients whose data cannot be anonymized according to legal and institutional protocols.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Diagnostic performance of the Artificial Intelligence algorithm in detecting multiple neuro-ophthalmologic and neurologic conditions from retinal imaging. | baseline | Evaluation of the algorithm's sensitivity, specificity, and area under the receiver operating caracteristics curve for classifying multiple neuro-ophthalmologic and neurologic pathologies using retinal fundus photography and Optical Coherence Tomography images, compared with expert-established reference diagnoses. |
Countries
France
Outcome results
None listed