Retinal OCT Shows Promise as Diagnostic Aid for Mild Cognitive Impairment

Aided by artificial intelligence, analysis of retinal images and related data identified condition-specific features that distinguished patients with mild cognitive impairment from those with normal cognition, a prospective study showed.

The analytic model, used to evaluate optical coherence tomography (OCT) and OCT angiography (OCTA), detected changes in the retina and microcirculation that distinguished mild cognitive impairment from normal cognition with 79% sensitivity and 83% specificity. Comparison of matched images using intraclass correlation coefficients (ICC) analysis (reflecting measurement reliability) produced values of moderate to good in most cases and excellent for some of the measures.

The results build on previous work showing good repeatability of OCT and OCTA analysis for features that distinguish Alzheimer's disease from normal cognition, reported Sharon Fekrat, MD, of the Duke Eye Center in Durham, North Carolina, and co-authors in Ophthalmology Science.

"This is particularly exciting work because we have previously been unable to differentiate mild cognitive impairment from normal cognition in previous models," Fekrat said in a statement. "This work brings us one step closer to detecting cognitive impairment earlier before it progresses to Alzheimer's dementia."

In a previous study, Fekrat and colleagues used OCT and OCTA to identify differences in the superficial capillary plexus of the retina between patients with Alzheimer's disease and normal cognition. No significant differences were observed between patients with mild cognitive impairment and the control group with normal cognition.

A number of studies have identified changes in retinal anatomy and vasculature in patients with Alzheimer's disease. Increasingly, researchers have focused on identifying biomarkers to distinguish mild cognitive impairment from normal cognition. If realized, that goal could pave the way to a noninvasive diagnostic test for preclinical dementia. A study reported several years ago took an intermediate step in the discovery process, showing differences in retinal parameters between patients with Alzheimer's and those with mild cognitive impairment.

Though OCTA has shown promise, measurement reliability can be affected by patient factors, scan signal strength, imaging artifact, and level of cooperation during the imaging process, Fekrat and co-authors noted in their introduction. Unacceptable variation in measurements could affect clinical adoption of the technology. As a result, studies validating repeatability of OCTA measures are essential.

In the current study, investigators examined retinal perfusion in patients with Alzheimer's disease, mild cognitive impairment, and Parkinson's disease, as well as a control group of individuals with normal cognition. They performed repeat OCTA imaging for both eyes of each study participant and measured perfusion density (PFD), vessel density (VD), and foveal avascular zone (FAZ), using 3×3-mm and 6×6-mm OCTA images.

The primary outcome was repeatability of PFD, VD, and FAZ measurements obtained by OCTA, reflected by ICC analysis. The analysis included 3×3-mm scans for 22 patients with Alzheimer's, 40 with mild cognitive impairment, 21 with Parkinson's, and 26 with normal cognition, as well as 6×6-mm scans for 29, 44, 29, and 30 participants in the same four patient groups.

For the 3×3-mm images, ICC values for repeatability ranged from 0.64 (moderate) for PFD and VD in patients with mild cognitive impairment to 0.87 (good) for PFD in the Alzheimer's group and did not differ significantly by diagnostic group. Repeatability of FAZ area was excellent (ICC 0.97-0.99) for all diagnostic groups.

For the 6×6-mm images, most ICC values for PFD and VD ranged between moderate (0.56 for outer ring PFD and inner ring VD of patients with normal cognition) to good (0.85 for PFD in patients with mild cognitive impairment). No significant differences in repeatability were observed among or between the different diagnostic groups.

"[U]ntil now, the repeatability of these measurements had not been assessed in neurodegenerative diseases and compared among diagnostic groups," the authors wrote. "In this study, we conclude that the overall repeatability of macular OCTA was moderate to good and, notably, repeatability was not worse in individuals with neurodegenerative disease compared to NC [normal cognition] across all imaging parameters."

Comment