OCT angiography preferred method for detecting glaucoma in optic disc, macula
Click Here to Manage Email Alerts
Key takeaways:
- OCTA outperformed OCT alone or with OCTA in glaucoma detection based on accuracy, sensitivity, specificity and AUC.
- Integrating OCTA with deep learning may contribute to more effective glaucoma diagnosis.
OCT angiography alone demonstrated superior sensitivity in detecting glaucomatous changes in the optic disc and macula compared with OCT and OCT angiography combined, according to a study published in Clinical Ophthalmology.
“Several investigations have employed deep learning, especially convolutional neural networks (CNN), for analyses of OCT and OCT angiography (OCTA) readings to identify ocular illnesses such as retinal vein occlusion or diabetic retinopathy,” Sayeh Pourjavan, of the department of ophthalmology at Catholic University of Louvain in Brussels, and colleagues wrote. “However, although deep learning has already been used on OCT images to detect glaucoma, this is much rarer in the case of OCTA images.
“This study marks an initial validation of deep learning techniques for distinguishing between healthy and glaucomatous states, paving the way for more sophisticated investigations into disease detection at earlier, subclinical stages,” they added.
To compare the diagnostic accuracy of a CNN architecture for OCT and OCTA as well as assess the value of CNN-derived combined OCT and OCTA images in diagnosing glaucoma, Pourjavan and colleagues used the AngioVue imaging system (Solix Optovue Inc.) to conduct OCTA and spectral domain OCT imaging of the nerve head and macula of 135 eyes from 45 patients with glaucoma (mean age, 66.4 ± 12 years; 47.7% women) and 103 eyes from 51 healthy individuals without glaucoma (mean age, 49.7 ± 17 years; 72.5% women).
For the dataset, they used preprocessed, resized and normalized en face images of the superficial and choroid layers for OCTA-based vessel density and OCT-based structural thickness of the macula and optic disc.
According to the researchers, the first part of the CNN architecture extracted features for each image type (OCT-disc, OCT-macula, OCTA-disc and OCTA-macula), and the second part combined the features to classify eyes as healthy or glaucomatous.
They measured performance by accuracy, sensitivity, specificity and area under the curve.
For OCT images, the disc and macula combination outperformed disc or macula alone with an accuracy of 87.82%, AUC of 95.49% and sensitivity of 89.66%.
For OCTA images, the disc and macula combination outperformed disc or macula alone with an accuracy of 92.44%, AUC of 96.94%, sensitivity of 94.71% and specificity of 89.34%.
Similarly, for OCT and OCTA combined, the disc and macula combination outperformed the disc or macula alone with an accuracy of 89.92%, AUC of 96.92%, sensitivity of 92.61% and specificity of 86.51%.
The researchers noted that based on these data, OCTA consistently outperformed OCT, as well as OCT and OCTA combined, across all metrics for combined disc and macula images.
“This finding suggests that vascular information provided by OCTA plays a crucial role in early glaucoma detection, particularly in cases where OCT’s structural data is less indicative of early-stage disease,” the researchers wrote.
Pourjavan and colleagues acknowledged that the difference in average ages between patients with glaucoma and those without glaucoma may be a potential limitation.
“Future studies may benefit from age-matching the control and glaucoma groups to minimize this bias,” they wrote. “In ‘real life’ most of the glaucoma patients are elderly. However, the significant structural and vascular differences detected by OCT and OCTA in our study support the conclusion that these methods are effective for glaucoma diagnosis, even with age as a variable.”
Perspective
Back to Top
Bridget Boamah Appiah, OD, FAAO
Although early diagnosis and intervention has been shown to significantly reduce the progression of glaucoma, the equivocal nature of its symptoms during its initial stages makes prompt diagnosis a challenge for optometrists.
This study addressed the integration of OCT and OCTA with deep learning for glaucoma diagnosis and highlighted that OCTA alone outperformed OCT combined with OCTA in terms of diagnostic sensitivity. The clinical significance is that vascular information provided by OCTA plays an important role in early glaucoma detection, particularly in cases where OCT’s structural data are less indicative of early-stage disease.
The study, however, had some shortcomings. Although OCTA relies on structural OCT data, deep-learning analysis unexplainably showed that OCTA alone outperformed OCT plus OCTA in sensitivity, particularly when assessing the macula and disc areas. The study also relied on high-quality standardized images, which may not reflect real clinical conditions. More so, the participants differed significantly in age, which may introduce some bias, as age is a risk factor for glaucoma.
The adoption of OCT and OCTA combined with deep-learning techniques has great potential for early and accurate glaucoma diagnosis, but further studies with age-matched controls and diverse datasets are needed for its validation and generalizability.
Collapse
Click Here to Manage Email Alerts