At the time of writing, the world is going through the heart of the COVID 19 pandemic. Nearly all educational meetings across the world are being cancelled, entire state school systems are shutting, and there is no toilet paper to be found in any store. Most of us are practicing social distancing to help protect those most vulnerable. Older individuals and those with compromised health are at most risk for developing serious complications from the virus.
Consequently, these are also the individuals we are likely to be monitoring for various types of glaucoma. This month we’ll take a look at recent articles on the topic, including one relevant to patients who are self isolating at home.
Using self tonometry to monitor increased IOP
In the February 2020 issue of Optometry and Vision Science, an article from The University of New South Wales in Australia investigated the use of self tonometry with the Icare® HOME in 498 subjects. Researchers wanted to help provide some guidance to practitioners on how to interpret the often confusing results. 175 normal patients, 286 glaucoma suspects, and 30 patients with glaucoma were measured with both the Icare® Home tonometer and applanation tonometry by trained optometrists.
The average Icare® HOME measurement was 1.7 mmHg below that of the average applanation measurement. Taking this slight bias into account along with the confidence intervals of measurements and normal diurnal fluctuations, the authors developed a reference table that practitioners can use to reference their patients' measurements (figure 1). Different patient types warrant different diurnal fluctuation values, which accounts for the 3 columns. For example, those with low tension glaucoma don’t fluctuate as much as those with ocular hypertension.1
| Target IOP | Threshold HOME IOP (+2.5 mmHg) | Threshold HOME IOP (+3.0 mmHg) | Threshold HOME IOP (+3.5 mmHg) |
|---|---|---|---|
| 10 | 15.0 | 15.5 | 16.1 |
| 11 | 16.1 | 16.6 | 17.1 |
| 12 | 17.1 | 17.7 | 18.2 |
| 13 | 18.2 | 18.7 | 19.3 |
| 14 | 19.3 | 19.8 | 20.3 |
| 15 | 20.3 | 20.8 | 21.4 |
| 16 | 21.4 | 21.9 | 22.4 |
| 17 | 22.4 | 22.9 | 23.5 |
| 18 | 23.5 | 24.0 | 24.5 |
| 19 | 24.5 | 25.0 | 25.6 |
| 20 | 25.6 | 26.1 | 26.6 |
The reference table helps docs taking advantage of ICare® Home to interpret the results. To simplify the process even more, if the value from the home tonometer exceeds a patient's target pressure + 6, chances are very good their true pressure is above target and warrants a change in treatment.
In times like these where routine care is suspended, home IOP monitoring devices may begin to play a larger role in more and more practices.
Looking at corneal speckle as predictor of glaucoma risk
A lot of news surrounding how OCT can further improve our ability to identify and measure progression of glaucoma often involves new features like angiography. But a group of researchers from Wroclaw University in Poland were interested in how existent OCT technology might give some insight.
18 subjects with open angle glaucoma, 24 suspects, and 22 age matched normal subjects were recruited for the study. Each subject had standard glaucoma screening tests done to confirm diagnosis (IOP, ONH OCT, HVF, etc.). Additionally, anterior segment OCT was performed and the central 2 mm of the cornea was analyzed. The pixel density of these corneal OCT images was found to be significantly different in the glaucoma and glaucoma suspect subjects compared to normals. The finding suggests that the microstructure of the corneal stroma, and more importantly, the overall collagen profile of an individual play a significant role in glaucoma.2
Depending on how well-refined this technology becomes, it would not be surprising to see software updates taking advantage of corneal speckle in many commercially available OCT instruments. In the same way ganglion cell scans improved our ability to detect early glaucoma, corneal speckle may become a screening tool that is quick, non-invasive, and can be performed with an instrument many practitioners already have.
Deep learning and glaucoma
Deep learning and artificial intelligence have been touted as the possible solution to many of our problems, including glaucoma management. However, deep learning often requires very large sets of data in order to train the algorithm. An article from early this year published in Acta Ophthalmologica outlined how a research group from Belgium tackled this issue.
Researchers collected 8433 color fundus images labelled as either glaucomatous vs non-glaucomatous to be used. Employing both transfer learning and active learning techniques, researchers were able to train a program to detect glaucoma with 98% sensitivity and 91% specificity using only 2072 images, significantly less than required in other studies. However, the researchers admit that most of the training was done using images of moderate to severe glaucoma, reducing the practicality for identifying early glaucoma with this practical program.
Another interesting aspect of the study came in the form of heat maps. Heat maps are images generated by the program to show researchers exactly what part of the image it considers important for distinguishing between glaucoma and non-glaucoma. The findings suggest that both the superior and inferior portions of the optic nerve as well as the superior and portions of the RNFL were the most important areas for the program. Intuitively, this makes sense to us as practitioners, considering that we often prioritize vertical C/D ratio and see most of RNFL thinning in the superior and inferior quadrants. Fortunately, efficiency of machine learning will continue to improve, data labelling costs will go down, and commercially available software should be an option for practitioners in the near future.3
Conclusion
As with many other issues we’ve faced in the past, technological advancement will likely help to save us—not only in the realm of glaucoma, but also in the current predicament. Rapid genome sequencing and electronic transfer of information to various pharmaceutical companies has fast tracked a potential vaccine in a matter of months that could have taken years to develop just 10 years ago. Regardless of how it turns out, how we look at both pandemics and glaucoma will continue to evolve in lock step with technological advancement.
Resources
- Huang J, Phu J, Kalloniatis M, Zangerl B. Determining Significant Elevation of Intraocular Pressure Using Self-tonometry. Optom Vis Sci 2020;97:86–93.
- Iskander DR, Kostyszak MA, Jesus DA, Majewska M, Danielewska ME, Krzyżanowska-Berkowska P. Assessing Corneal Speckle in Optical Coherence Tomography: A New Look at Glaucomatous Eyes. Optom Vis Sci 2020;97:62–7.
- Hemelings R, Elen B, Barbosa-Breda J, Lemmens S, Meire M, Pourjavan S, Vandewalle E, Van de Veire S, Blaschko MB, De Boever P, Stalmans I. Accurate prediction of glaucoma from colour fundus images with a convolutional neural network that relies on active and transfer learning. Acta Ophthalmol 2020;98:e94–100.
