Findings from a recent study published in Ophthalmology Science evaluated the severity scales of diabetic macular ischemia (DMI) by analyzing the quantity and distribution of capillary nonperfusion using optical coherence tomography angiography (OCT-A) imaging.
Give me some background on DMI.
Capillary nonperfusion is an important sign of microvascular injury in eyes with diabetic retinopathy (DR) that represents the extent of ischemia—signifying the impairment of the capillary perfusion status of retinal vessels.
DMI often leads to visual impairment—including visual acuity (VA) reduction and poor retinal sensitivity associated with vessel density abnormalities—and in spite of its clinical significance, the characteristics of DMI progression remain unclear.
- In fact: Diagnostic criteria and severity grading have yet to be fully established.
Bring it back to this study.
Uniform manifold approximation and projection (UMAP) is a dimensionality reduction technique used to transform complex, high-dimensional data sets into more manageable representations to improve data visualization and organization.
In a previous study, UMAP was used to develop a novel continuous severity scaling system for macular telangiectasia (MacTel) type 2 by combining it with a deep learning (DL) model.
- Consequently: The study authors sought to use UMAP to create an objective severity scale for DMI with high-dimensional OCT-A images.
Now talk about the study.
In this single-center, prospective case series, investigators included 301 eyes from 301 patients with diabetic retinopathy (DR).
- Main outcome measure: Development of a severity scale for DMI
Researchers acquired 3 x 3-mm swept-source (SS-) OCT-A images and created en face images that featured a circle with a 2.55-mm diameter around the foveal center.
This circle was then divided into 15 x 15-pixel squares, and nonperfusion squares (NPSs) were defined as squares without retinal vessels.
Eyes were then divided into five groups based on the similarity of nonperfusion area distribution:
- Initial
- Mild
- Superficial
- Moderate
- Severe
Findings?
Nonperfusion square counts in the deep layer increased in the eyes of the initial, mild, moderate, and severe groups in a stepwise fashion.
Conversely, there were no significant changes in superficial NPS counts between the eyes of the initial and mild groups.
In eyes with intermediate DMI, the eyes of the superficial group exhibited higher NPS counts in the central sector of the superficial OCT-A layer compared with those of the moderate group.
Anything else?
The foveal avascular zone (FAZ) extended into the temporal subfield of the deep layer in the eyes of the moderate group.
Eyes of the severe group had significantly poorer VA that was more frequently accompanied with proliferative diabetic retinopathy (PDR).
Expert opinion?
“Eyes with retinal thinning were more frequent in the severe group, so neurodegeneration might contribute to the visual impairment at least in part,” noted the study authors.
They added that the different NPS characteristics between the superficial and moderate groups suggest that multiple mechanisms regulate the progression of capillary nonperfusion until the intermediate stages of DMI, including:
- Inherent FAZ size and vessel density in the parafovea
- Neuroglial responses to hyperglycemia
- Vascular permeability
Limitations?
These included:
- Selection bias due to the single-center study design and inclusion/exclusion criteria
- Only one SS-OCT-A device was used to collect data and apply the imaging processing
- Segmentation errors might have led to incorrect quantification in eyes with diabetic macular edema (DME)
- Other algorithms for dimensionality reduction and clustering could lead to other feature extractions or unsupervised classifications
Take home.
These findings suggest that applying UMAP and clustering facilitated the development of a novel severity scale for DMI based on the distribution of capillary nonperfusion in OCT-A images.
Next steps?
A future longitudinal study is necessary to validate these proposed severity scales and progression pathways.
Further, subsequent investigations could analyze the association between nonperfusion area distribution and retinal functional examinations other than VA.