Multimodal Imaging in Diabetic Macular Edema

Abstract

Throughout ophthalmic history it has been shown that progress has gone hand in hand with technological breakthroughs. In the past, fluorescein angiography and fundus photographs were the most commonly used imaging modalities in the management of diabetic macular edema (DME). Today, despite the moderate correlation between macular thickness and functional outcomes, spectral domain optical coherence tomography (SD-OCT) has become the DME workhorse in clinical practice. Several SD-OCT biomarkers have been looked at including presence of epiretinal membrane, vitreomacular adhesion, disorganization of the inner retinal layers, central macular thickness, integrity of the ellipsoid layer, and subretinal fluid, among others. Emerging imaging modalities include fundus autofluorescence, macular pigment optical density, fluorescence lifetime imaging ophthalmoscopy, OCT angiography, and adaptive optics. Technological advances in imaging of the posterior segment of the eye have enabled ophthalmologists to develop hypotheses about pathological mechanisms of disease, monitor disease progression, and assess response to treatment. Spectral domain OCT is the most commonly performed imaging modality in the management of DME. However, reliable biomarkers have yet to be identified. Machine learning may provide treatment algorithms based on multimodal imaging.

Challenges in Diabetic Macular Edema Management: An Expert Consensus Report

“…Although CMT reduction, measured with OCT, is a very useful marker and is commonly used for monitoring treatment response in eyes with DME, there is not enough scientific evidence to support the relationship between CMT and chronicity or CMT and inflammation in DME. [6][7][8][9][10][11][12][13][14][15]40,41 Panel recommendation:…”

“…4,5 The prevalence of DME in Europe was estimated in 3.7% and its pooled mean annual incidence in type-2-diabetes patients was 0.4%. 3 The changes in the paradigm of DME treatment [6][7][8][9][10][11] as well as the development of technological advances for diagnosis [12][13][14][15] makes, from our point of view, necessary to reconsider the approach to the daily practice management of these patients.…”

Ensemble of deep convolutional neural networks using multi-spectral vision transformer for diabetic macular edema classification from multicolor image

“…Multicolor images (MCIs) have effectively addressed these challenges by utilizing three color lasers from a confocal scanning laser ophthalmoscope (cSLO) to capture distinct retina layers. The use of a multi-wavelength laser enables the acquisition of fundus images in the macular region, consisting of blue reflection (BR) at 488 nm, green reflection (GR) at 515 nm, and infrared reflection (IR) at 820 nm [3] . Furthermore, these three images are automatically processed to generate an additional pseudo-color (PC) image, making MCI (Figure 1) a collective term encompassing these four images.…”

Multimodal, multitask, multiattention (M3) deep learning detection of reticular pseudodrusen: Toward automated and accessible classification of age-related macular degeneration

“…Most deep learning models in the medical informatics domain are single-modality approaches, as demonstrated in recent literature reviews [21][22][23][24][25]. For instance, a detailed review summarized 15 deep learning models for retinal disease diagnosis; all were single-modality, using either CFP alone or optical coherence tomography (OCT) images alone [22].…”