Multimodal retinal imaging: synthesis of some methods
for the automatic screening of ocular pathologies.

Abstract
Robust and reliable diagnosis of ocular pathologies and their origin, assessment of
their severity, choice of appropriate treatment, as well as monitoring of their
progression require the use of multimodal retinal imaging.
This imaging uses a variety of imaging techniques, the main ones being OCT (Optical
coherence tomography), OCT – A (Optical coherence Tomography Angiography),
OCT - 3D, Fluorescein Angiography (FA) and Indocyanine Green Angiography
(ICGA), as well as ultra-fast and ultra-widefield fundus imaging.
Increasingly, ophthalmological platforms are using a combination of these
techniques; we are talking about Combined Multimodal Imaging platforms. Using
these platforms, we aim to combine these different imaging sources to apply different
imaging techniques in order to locate, extract, segment and analyze the different
morphological structures of the ocular fundus, for the detection and monitoring of eye
diseases.
We present in this plenary these different imaging techniques, illustrated by
Framework examples, as well as their advantages and limitations. Subsequently,
through some research work, methods based on deep learning will be described
while showing how these methods exploit these different sources of retinal imaging to
make a precise diagnosis, ensuring monitoring of the evolution of the pathology, its
signs, as well as the prediction of its stages of severity. We will mainly deal with
multimodal methods based on deep learning for early screening, diagnosis and
detection of the severity of the three main pathologies namely Glaucoma, Diabetic
Retinopathy (DR) and Age-related Macular Degeneration (AMD).

Mohamed AKIL
Emeritus Professor
LIGM Laboratory (UMR CNRS 8049)
Gustave Eiffel University, ESIEE Paris