# Teresa Chen Chen Lancaster Glaucoma Imaging 8 20 21 VIDEO FINAL **Channel:** Phương Trương **Source:** https://www.youtube.com/watch?v=jiyxEJD-MJw **Transcript page:** https://www.withtranscript.ai/video/jiyxEJD-MJw ## Chapters - 0:02 — Introduction and Talk Structure Overview - 1:11 — OCT Imaging & RNFL Parameter Overview - 7:22 — Optic Nerve and Macula Software Differences - 11:23 — Pattern Recognition and Symmetry Principles - 19:43 — Literature Review and Clinical Parameters - 24:31 — Glaucoma Progression Analysis and Trends - 33:45 — OCT Artifacts, Diseases, and Machine Variations - 54:54 — Acknowledgments and Closing Remarks ## Transcript **[0:02]:** Hi, I'm Theresa Chen. I'm a full time glaucoma specialist at Mass Eye and Ear. Glaucoma imaging has a special place in my heart and is one of my favorite topics because the topic provides a framework for the story of my academic career. And within that framework, I have been blessed to know many colleagues with whom I've done glaucoma imaging research over the past two decades. So I'm going to share you a little bit about the wonderful people that I've met in my career. And then lastly, how glaucoma imaging is not just my story and my colleague's story, but how glaucoma imaging can be your story. Because we as ophthalmologists, all of us, really need to understand how we can fully utilize glaucoma imaging to take best care of our patients in the clinic. So with that, I'd like to start the talk also acknowledge my funding source. **[0:57]:** So this talk will be divided into two segments. The first segment is going to be talking about the good things about oct. Then the second part of the talk is going to be talking about the bad things about oct. **[1:11]:** So first of all, the good things, we can use spectral domain OCT to diagnose glaucoma. And why do we need to use this? Well, glaucoma imaging provides us an objective, quantitative way to examine the structural and nerve tissue that can be lost or damaged with glaucoma. And most people feel that structural changes precede functional changes. So imaging structural changes in glaucoma disease offer us as ophthalmologists an opportunity to not only one, diagnose glaucoma earlier to, but to initiate treatments earlier to prevent irreversible vision loss. So we as ophthalmologists can make a difference if we really understand imaging. So what are we imaging in glaucoma? Well, in glaucoma there are basically three regions of interest and all three regions can be imaged with oct. So the first region is the optic nerve. So we all know that the optic nerve undergoes cupping with glaucoma. This is the old way of looking at the optic nerve with the 78 or 90 Diophara lens in the clinic. **[2:19]:** And here you can see one eye here is normal and the other eye here has classic cupping. However, with oct, we can look at the optic nerve in a more sophisticated way. We can see not only a normal eye with a small cup and a glaucoma eye with a big cup, but we can also quantify a thick normal neuroretinal rim and, and a thin glaucomatous neoretinal rim. So here you can see in yellow, the neural retinal rim is very thick in a normal patient and very thin in yellow in a glaucoma patient. So that's the first region. The second region of interest in glaucoma is the retinal nerve FIBA layer. So this is important because, as we know, when we look at the patient the old way again in the clinic, with our slit lamp, perhaps with red free lighting, we can see in a normal eye a highly reflective, bumpy nerve fiber layer. **[3:10]:** But in glaucoma we can see the classic pattern of arcuate, less reflective, flat nerve fiber layer. But with oct, we can look at the rnfl, the nerve retinal nerve fiber layer in a more sophisticated way. The sophisticated way, of course, is here we can see the black retinal nerve layer on the top, normally very thin. But as we become closer to the optic nerve, the nerve fiber layer normally becomes thicker as it feeds into the optic nerve. So that's the new way. Here you can see a thick retinal nerve fiber layer with the second order blood vessels totally encased in the thick retinal nerve fiber layer. Here you can see the second order blood vessels totally inside the thick black retinal nerve fiber layer. But in a glaucoma patient, these second order blood vessels become visible and are shown in relief, pooching out on top of the very thin nerve fiber layer. **[4:05]:** So the third region, which is probably not as obvious, is the macula. Well, the macula, the old way again is the region in the posterior pole where you have more than one cell layer thick of ganglion cells. So I remember G ganglion cell, g glaucoma, that is the cell that is primarily involved in glaucoma pathophysiology. And again, it makes sense that it imaging the macula is important for us glaucoma patients because the macula is a region where the ganglion cell is more than one cell layer thick. And in the macular region here resides about 50% of the retinal ganglion cells. And therefore the punchline is we know that the macula thins with glaucoma. And most importantly for OCT imaging, this can be quantified ob so what are we imaging in glaucoma? Again, before oct, we did not have such a beautiful view of the macula where we can see the ganglion cell complex with the nerve fiber layer, ganglion cell layer, and in a proxiform layer or we can image the macula, the whole retinal thickness. **[5:11]:** So this is the new way to look at the macula in our glaucoma patients. So that's what we're imaging in glaucoma. We're imaging the optic nerve, the nerve fiber layer and the macular region. Those three regions, the next part of the talk, the good things about OCT, is that we need to understand how software can help us in the clinic. And specifically, this is glaucoma software. So of course we know that OCT has retina software for retinal diseases. But again, this talk is going to focus on glaucoma software for glaucoma disease. So again, the three regions of vitreous are the nerve fiber layer, the optic nerve, the macula, and what are the software differences for the retinal nerve fiber layer? So I start off with a nerve fiber layer because the neurofiber layer thickness is the most commonly used parameter when we evaluate glaucoma patients. **[6:00]:** And it is the parameter that is the common parameter for all spectrallivin OCT machines, of which there are many. So here we can see how a circular skin around the optic nerve head can be cut like a yellow string and displayed like a linear graph. And here is the nerve fiber layer temporal to the optic nerve, superior nasal, inferior temporal. So we get the tisnet retinal nerve fiber layer plot and our patient's nerve fiber layer is mapped in black here. **[6:33]:** So most spectral domain OCT machines use a R peripapillary scan circle of about 3.4 plus millimeters in diameter. And here is an article which we published. It was at American Academy of Ophthalmology Physician paper on Spectral domain OCT and Diagnosing Glaucoma 2018. If you want to look that up. It summarizes the software differences for neurofibril layer thickness. Here in the spectralis there are many different circle sizes. However, this paper showed us that still around 3.4 to 3.5, that scan circle size is the best for diagnosing glaucoma. So that's RNFL thickness. Again, that's one of the key parameters we use in OCT for glaucoma patients. Now, what about the optic nerve, the second region of interest? **[7:22]:** Well, this is important because of course we know that the neuroretina rim thins with glaucoma and we can quantify this objectively with spectral immunoctor. Now, there are two ways to quantify optic nerve tissue. You can do this with a reference plane and without a reference plane. So the Cirrus, Rdview and 3D Oct uses a reference plane and the Spectralis machine doesn't. So basically the OCT machine takes a reference plane that is above the RPE. So for example, in RdView the reference plane is 150 microns below, sorry, above the RPE. And this reference point above the RPE is the boundary, below which everything is the cup, above which everything is a red neuroretinal rim. So when you see rim area on the OCT printout, it denotes the rim area along this reference plane right there. Doesn't tell you anything about the nerve above or below, but the rim area is the rim along that reference plane. **[8:21]:** For the Spectralis you have the VMO minimum rim width, which is the shortest distance between the Brooks membrane opening and the cup surface. And that of course is the yellow neuroretinal rim we see here. Our researchers group has evaluated the high density minimum distance band, which is the research correlate of the low density BMO minimum rim. Here you can see differences in software for the macula, the third region of interest. And as we alluded to before, you can visualize different layers of the macula. And it's important to realize that different machines have two differences. Artiview, 3D OCT, Cirrus and Spectralis. They are measuring different layers of the macula. So for example, the RDVIEW has the ganglion cell complex, where it gives us the total thickness of the ganglion cell complex, which is comprised of three layers. The top retinal nerve fibril layer, which is the ganglion cell axons, the gcl, the ganglion cell bodies, the inner plexiform layer are the ganglion cell dendrites. **[9:26]:** So all three portions of the ganglion cell are incorporated in the GCC or ganglion cell complex in the RDB machine. However, for example, in the Spectralis, the macular software for glaucoma has posterior polyasymmetry analysis, which gives us the micron thickness for the whole retina. And of course, if the ganglion cell layer thins with glaucoma, then of course the whole retinal layer is going to thin. But be aware that different machines are imaging different layers of the macula. So that's the first difference in software. Different machines are measuring different layers of macula. Different machines are also imaging different regions of the macula. So for example, the serous looks at an oval area, the RDV looks at a circular area a little temporal to the fovea. And then the 3D OCT is measuring a square macular region. And then the Spectralis machine measures a square region, but is slightly tilted along the optic nerve foveal axis, which of course is about 15 degrees downward. **[10:38]:** So the bottom line is different software for different machines in the macula look at one different layers of the retina and two different regions of the macula retinal region. So be aware. So the bottom line is values. That means the micron thickness values of these three different regions between spectral domain OCT machines are not interchangeable. So neurofibrillar thickness measurements are not interchangeable, neural retinal rim measurements are not interchangeable, and the macular region values are not interchangeable. So once you use a certain machine for your patient, please stick with that machine for the rest of their glaucoma assessments. **[11:23]:** So that's glaucoma OCT software. The wonderful benefits that we have in the clinic to help us focus on the three regions of interest and how they can be measured and how we can tell if they thin with glaucoma disease progression. So this is kind of a fun area. General principles of OCT imaging in glaucoma. So pattern recognition. This is sort of fun because it's just the gestalt impression of the OCT data that we're seeing in the clinic. So pattern recognition of the optic nerve. Well, we all know that the optic nerve looks like a donut. Donut. Just to refresh. Donut. That's the donut. So. So that's the normal norim. It's like a donut. Now, if your donut looks kind of squashed superiorly and inferiorly, that's glaucoma. **[12:14]:** So macula pattern recognition. So, interestingly, if we look at a macular printout for spectral domain oct, and here's a normal printout with glaucomas and normal features. And if we look at the details, we have the top row of heat maps and then we have the bottom row of deviation maps. So to dissect the anatomy of this printout, again, let's focus on the top row. A heat map maps the thickness of the tissue. We're meshing. So here you can see OD thickness map, OS thickness map. Here's the right eye, the left eye, and the macula usually also looks like a donut. Normally. All right, a donut, sort of a yellow orange donut. The macular region in a glaucoma eye does not look like a donut. Looks like someone ate the bottom half of our yellow orange donut. So this is the heat map where it actually gives us the actual thickness value in microns of this macular region. **[13:13]:** And it's a heat map because the hotter values are thicker and the thinner values are colder. So that's the top row. However, on this printout you can see deviation maps. So this is OD deviation map. And here you can see it says OS deviation map. Now, normally the deviation map should be gray. There should be no deviation from age match normals. Everything should be green like a traffic light. Green is good, red is bad, yellow. It's kind of a warning. However, in the right eye, this is the left eye normal. The right eye has glaucoma features because there is thinning of the ganglion cell region and therefore we see red. So red traffic light is bad and there's thinning and deviation from age match normals. So if you see color, that means there's a deviation from age match normals. So again, deviation match if there's color indicates the degree of deviation from expected age match normals. **[14:06]:** If it's red, it's more deviated from normal. If it's yellow, it's just a little deviated from normal. And if there are no colors, that means there's no deviation from normal. Your patient probably has a normal macular thickness. **[14:20]:** So again, the bottom line, the pattern recognition normally is a donut pattern. Anything that doesn't look like a donut is probably glaucomatous. Also, for pattern recognition for glaucoma, if you just look at the printout, just gestalt, your printout should be symmetric because we're symmetric. I mean, I have a right hand. I have a left hand. They're symmetric butterflies. Everything in nature mostly is symmetric. However, in this posterior poly symmetry analysis on the Spectralis machine here, you can see it doesn't look like a normal macular donut. Somebody ate part of the donut out more so in the right eye than the left. And here you can see more of an arcuate pattern or a defect on the right side more so than the left. So again, this is not symmetric. Someone ate more of the doughnut on the right than the left. And someone has more arcuate. **[15:11]:** Again, pattern recognition. Arcuate is a classic feature of glaucoma. Then arcuate defect is more pronounced on the right eye than the left eye. Then the bottom line is the right eye has more glaucoma, so it's not symmetric with the left eye. **[15:27]:** And this asymmetry principle can be carried on with all features of ocular tissues. So again, 5, 6% of the normal population has some optic nerve cuppedis ratio Asymmetry, but much of the time, cupcake asymmetry can be at early sign of glaucoma. Retinal nerve fibrillary asymmetry can be at early sign of glaucoma. And macular asymmetry can be early sign of glaucoma. So the bottom line is asymmetry should be a red flag. It should send warning signs in your head saying that your patient may have early glaucoma. Pattern recognition. So what are some pattern recognition of the nerve fiber layer? Well, the nerve fiber layer when it's mapped out on the tisnid graph, usually here, our patient's nerve fibrillator thickness in this dark green line is usually a double hump as we see in nature with this camel. Age matched normal is a shaded green area here. **[16:23]:** So pattern recognition. The neurofiber layer tisnet graft should be normally a double hump where the superior and inferior nerve fiber layer is thicker. So hopefully everybody has a thicker superior and infraneurifibal layer in this room. However, with glaucoma, again, we know there's preferential thinning of the superior and inferior nerve fibral layer. So normally, instead of this green double, high double hump with progressive glaucoma, that double hump flattens. And with more advanced glaucoma, as seen in red here, the double hump flattens, which is not normal. Normally you should have high double humps, but in glaucoma, the double humps flatten. So that's pattern recognition. **[17:04]:** Pattern recognition. So on the heat maps for your nerve fiber layer thickness maps, normally if you were to cut out the nerve fiber layer pattern, the inset around the disc, again, this rectangular region blown up to this rectangular region that normally on the heat map that the nerve fiber layer for all of us normally should be thicker superiorly and inferiorly. And our nerve fiber layer normally is thinner temporally and nasally. So this looks like a bowtie. So if your nerve fiber layer normally doesn't look like a bow tie, something is wrong. So again, just gestalt you look at your printout for your patient. If you see a bow tie pattern such as in his right eye, it's probably good. Deviation maps again tell us how much our patients nerve fiber layer thickness values deviate from expected age match normal, meaning a normal patient of similar age to our patient. **[18:03]:** Here on the deviation map, there should be no deviation from age matched normals. In a normal nerve fiber layer thickness map, if you have early glaucoma, there's some deviation Superiorly and inferiorly. As there's some thinning superiorly and inferiorly, moderate glaucoma, there's more thinning superiorly and inferiorly. In advanced glaucoma, you have a lot of superior inferior thinning. So the bottom line is if you see color on your deviation map, then the machine is screaming to you and telling you your patient has some deviation from age match normals. So again, this is a normal patient, no deviation, no colors. In an advanced glaucoma patient, there is a lot of thinning superiorly and inferiorly. So that's pattern recognition. We should see a normal doughnut pattern for our optic nerve neurore retinal rhyme. We should see a normal donut pattern for the macular scan. For oct, we should focus on the heat maps where, for example, the retinal nerve fiber layer thickness looks like a bow tie. **[19:01]:** Normally, deviation maps, nothing should deviate from normal. But of course, when glaucoma, if you see color on your deviation map, that's a bad sign. If you see anything asymmetric between the right eye and the left eye, that could be glaucoma. If you see any asymmetry from the top hemiphil to the bottom hemiphil, that could be abnormal. And another general principle is location, location, location makes a difference. I'm trying to sell my parents condo and I'm hopeful that location makes a difference because it's in a great area of the city. But again, location makes a difference. The same thing when you're practicing glaucoma, as in real estate, location makes a difference. **[19:43]:** So here you can see the paper that we did. I'm going to summarize the results of this paper. Again, this is a review paper of the literature for the past decade on two things, diagnosing glaucoma and spectrodominoct. And believe it or not, for the past decade there have been over 700 papers on glaucoma and spectrodominoct. And the happy thing is, I'm gonna summarize all those papers for you. The first following slides. So again, to reiterate the three regions of interest in glaucoma, the nerve fibrillator, the optic nerve and macula for the neurofiber layer. Different machines, serous, rdview and spectralis all conclude the same thing. So that review paper that we did for the academy shows us that the nerve fiba layer location, location, location preferentially thins in glaucoma globally inferiorly or superiorly. Or if you had to look at Sectors and not quadrants, it preferentially thins temporal inferiorly and temporal superiorly. **[20:49]:** Which makes sense because we know that dischemorrhages usually occur infratemporally and supratemporally. So this all makes sense. So what we know about glaucoma physiology, dishemorrhages and neurofibrillari neofibrillar thinning. All machines, regardless of which spectral domino CT machine you have in your office, all of them are associated with neurofibril layers thinning superiorly, inferiorly, superior temporally and infratemporally. So that's neurofibril layer thinning with glaucoma. So for the optic nerve, happily all machines also show neuroretinal rim thinning. So for example, for the serous machine, what is the best parameter of the many parameters we get on the printout? The best parameters of course are the rim area and the vertical cup dis ratio according to the review paper of all of those papers in the past decade. And again, if you don't have time and you're very busy of all those parameters on your serous spectrodomain oct printout, you want to focus on rim area which thins with glaucoma, and the vertical cup to ratio which increases with glaucoma. **[21:56]:** If you have an ARGU machine when you're stymied by all the parameters that the optic nerve can give you. Again, for best parameters, where do you want to focus on the printout? You want to focus on the rim area, specifically the inferior rim area and the vertical cuppedis ratio. So again, where to look? Rim area and vertical cup to ratio. So if you have a spectralis machine in your office, well, where are the best parameters to look? Well, it's basically here you can see a radial scan over the optic nerve head. You can have thinning of the Brooks membrane opening minimum rim width, which is that thinning of that yellow band 360 around the optic nerve head. So thinning of that BMO minimum rim width is associated with glaucoma. So neuroretinal rhythm thinning. So that's the nerve viva layer and the optic nerve with different machines for the macula, it doesn't matter which machine you have in clinic, the macula also location, location, location tends to thin superiorly and inferiorly and infratemporally and supratemporally regardless of which machine you have in your clinic. **[23:09]:** So to summarize this entire review article, the most important parameters in the clinic that you should focus on are neurofiber layer thickness or thinning with glaucoma, macular thinning, either of the GCC complex or the GCIPL complex. Or if you have the Spectralis machine, the total retinal thickness. When you look at the disc, you want to focus on rim area for the spectral luminosity machine. However, if you have the Spectralis machine, focus on BMO minimum wind width thinning. And if you have a non Spectralis machine, look at vertical cuppedness ratio. So those are the three regions of interest that you should focus on when you're in the glaucoma clinic with your patient. The bottom line, all of these regions thin with glaucoma. And again, location, location, location. If you are busy, don't focus on the nasal and temporal parameters. Focus on the inferior and superior micron values and deviation maps. **[24:12]:** So general principles. Again, very fun. Pattern recognition, heat maps, deviation maps, look for asymmetry and again focus on the locations that matter, which is our superior temporally and infratemporally. So those are general principles. So those are the happy things about spectrum and Oct and glaucoma. **[24:31]:** Now, on a higher plane, how can these values be used for glaucoma treatment? So when we're in clinic, we basically have to decide two things. Is my patient stable or is my patient progressing? There are two ways for us ophthalmologists to determine if our patient is progressing. We can use event based analysis, which is different than the prior visit, or we can use progression analysis. So I'm going to focus on event based analysis. So for event based analysis, what is a clinically significant change from today compared to last year? So for example, if I do yearly octs, what is a clinically significant change? Well, a significant change is a change that's greater than normal test variability or normal test noise and a change that's greater than normal aging change. So for example, our nerve fiber layer thins with aging, but that's not glaucoma. Glaucoma is thinning on the nerve fiber layer that's greater than normal age change. **[25:33]:** So to determine this, as a physician, we need to know, well, what is the normal noise for each of these three regions of interest and what is normal aging change? So to start out with the most common parameter, the nerve fiber layer thickness, when we look at our user manual, it has really important data. It tells us that the normal test retest variability for a normal patient is around 4 microns, and in the glaucoma patient, 4 plus microns. So therefore if the normal testing noise is 4 microns, meaning if you test your patient three times on the same day, you can get measurements that are within 4 microns of each other. That's just measurement noise. However, the rule of fives tell us if you have a change of 5 microns or more, that's probably not testing noise, that's probably not test retest variable. That's probably or could more likely be real glaucoma disease progression. **[26:30]:** So what is normal aging change? Is this even significant? Well, the bottom line is we don't really need to think about this because happily for all of us, our nerve fiber layer is not thinning very much while we're sitting here. It's about 1 to 2 microns per decade and perhaps even more so as we get older. But the bottom line is this age related thinning is much, much less than testing noise. So from year to year, we don't even have to think about normal aging change or normal aging decline. We need to focus on the rule of five. So five microns or more change is probably glaucoma nerve fibrillar thinning. Now, what are the progression for the other regions of interest? So for example, the optic nerve, 13 microns or more is significant. **[27:12]:** This makes sense because the neural retinal rim is usually thicker than the nerve fiber layer. So normal test variability is more. And it's easy to remember 13 because that's an unlucky number. Right. Four to five microns is true for the total macula and GCC layer according to the literature. **[27:30]:** So here's an example determining change by numbers. Now that we know the rule of fives, let's use it. Here is a patient with neurofib, a layer global average thickness measurement of 80 microns. So 360 degrees around the optic nerves, 80 microns. However, two years later, from 2010 to 2012, the nerve fiber layer drops 4 microns 80 microns to 76. Well, we know the rule of 5 is that 4 micron changes probably normal test variability. So I'm not going to escalate treatment or add drops or anything. However, the patient comes back in November and now has a nerve fiber layer thickness of 72. So red flags go off in my brain because I realize, well, between 80 microns and 72, that's 8 micron difference. So 8 microns. Again, rule of 5, 8 microns is more likely real glaucoma thinning. So let's say if the patient also has, let's say, visual field progression. **[28:26]:** And we know that this is probably real RNFL thinning, then I probably need to get a lower intraocular pressure and either do surgery, add drops or do laser or something to stop the process. Here's another example where RNFL thickness is not that useful. So structural testing, which is what OCT is, is not as useful when you've hit the floor. So the bottom line is we all have about 50 microns plus minus of non neuronal tissue in the nerve fiber layer. So this is glial cells or blood vessels. So even if we have a 0.9 nerve, we still have 50 microns of nerve fiber layer thickness. So once we hit the floor, once we have a 0.9 nerve, we could be totally stable at 0.9 and we can be totally stable at 50 microns on the nerve fibrillator thickness. But our visual field could be progressing. **[29:17]:** So the bottom line of the take home point is OCT imaging is not that useful in end stage glaucoma because again, the nerve fibrillator thickness can be rock stable, our cuppedis ratio can be rock stable, 0.9. But our patient can still have disease progression on visual fields. So if you see the visual field progressing in an advanced glaucoma patient, we need to treat that patient to stop the process or to slow it down. So again, OCT can be stable, test of structure can be stable, but the test of function, the visual field is progressing. Again, OCT is just one of many tests in clinic and we need to look at the patient and all their testing as the total composite. The last region of interest, the macula by pattern recognition. We should not see arcuate changes or arcuate defects on our OCT printout. **[30:12]:** So again, aside from pattern recognition, remember the numbers if you see rule of PHI's five or more that might be significant for RNFL thickness or other regions for the optic nerve, 13 is an unlucky number and that portends possibly glaucoma disease worsening. So that's determining glaucoma progression by event based analysis. **[30:34]:** Glaucoma disease progression is change greater than test noise and normal aging change. So that's 5 microns or more for RNFL thickness, that's rule of 5s or more than 13 for neuroretinal rim. So when we're in the clinic, our printouts also give us trend based analysis. So basically that's GPA glaucoma progression analysis. So meaning we're in clinic, if we have more than 4, 5 data points or 5 data points or more, our software can map out, let's say neurofiber layer thickness with a linear regression line. So this is a fascinating paper. Here we can see a patient at one year and then right around two and a half years our patient's trend analysis shows that the slope worsening. So this patient is thought to have glaucoma progression. So by trend analysis the patient progress if the slope is statistically different than a stable slope of zero this green line. **[31:39]:** So again our patient has progression at two and a half years. However, the problem with trend analysis for the clinical scenario is that it can be associated with a high falls progression rate, high falls positive. So here the same patient imaged with more RNFL OCT Data points by 5 years is progressing but by 4 years was stable. So to really determine progression in this patient we need five years of data. But again, most patients can't wait for their doctor to have five years of data before we need to treat. So again, I personally think that trend based analysis not as useful clinically unless you have years worth of data. But again, if you only have two points of data, let's say last year and today's visit, an event based analysis is better. But again trend based analysis can be useful if you have many years of data points. **[32:39]:** However, be aware that can be associated with a false positive over progression over detection of glaucoma progression. So be aware of the pros and cons of trend based analysis when you're in the clinic. So the big picture. So when is OCT good? So this is a summary is good in earlier disease when there's more structural change on OCT than visual field functional changes. **[33:09]:** You can use OCT for pattern recognition. It provides objective data, actual micron values of nerve tissue thickness. It has great reproducibility and we can evaluate in 3D. So when is OCT bed? Well, obviously in late disease, when you have reached the floor for the nerve fiber layer thickness at 50 microns or we have a 0.9 nerve, the OCT can be stable and not very useful. Then you really need to focus on the visual field data. And then this is the cliffhanger. **[33:45]:** OCT is bad when you have artifacts. So this is the second segment of the OCT talk, But before going to that again, I told you that glaucoma imaging has been a wonderful story of my career because I met wonderful people on the way, such as my imaging collaborators. Here is not one of my imaging collaborators, but this is Johnny Busak who I had the chance to meet. This is not a HIPAA violation because he has publicly said that his grandson has congenital glaucoma which was treated by one of my mentors, Dr. David Walton. So we did a fundraiser with the Boston Bruins because he was a captain of the Boston Bruins when it won the Stanley cup two years, and here he is holding the Stanley Cup. But I had a chance to meet him as a glaucoma specialist. So the bottom line is our career has many wonderful things along the way and it really boils down to, to caring for our patients and the wonderful people that we can meet. **[34:38]:** So here's a wonderful gentleman who is trying to raise money to further glaucoma research. So with that note, we're gonna go into the bad things about OCT when OCT test fails us in the clinic. Now, this is called OCT artifacts. So whether you have a time domain OCT machine or everybody now has spectral domain OCT and swept source machines, all OCT machines have artifacts. And all OCT machines, regardless of which scan you're using, the rnfl, the optic nerve and the macula, all glaucoma scans have artifacts and it's about 50%. So this is really bad. That means 50% of the time, plus, minus. We as clinicians need to know when our OCT scan has an artifact and it is giving us bad data because obviously if the data is unreliable or bad, then we need to throw it out. It's like a visual field test with 100% fixation losses. **[35:38]:** It's a bad test. We're not going to rely on that visual field test. Similarly, if your OCT has artifacts or is giving you bad data, then just don't use that in your decision making. And again, to reiterate, all testing in the clinic has artifacts. So sometimes we have blurry disk photos we can't use. Sometimes we have visual fields with RIM artifacts that are not that useful. And sometimes the OCT has bad data. So the reality of it is we need to have the strength in the clinic to throw out the OCT scan if we notice these three buckets of OCT problems. I'm going to go over the top 10 artifacts in Oct in glaucoma, I'm going to go over OCT diseases and I'm going to go over the issues when your patient switches from OCT machine to OCT machine. So the happy thing about OCT imaging artifacts is that most of the time, over 80% of the time, we can tell if there's an artifact if we take the time in the clinic to look at the printout. **[36:39]:** So in Letterman style, we're going to go over the top 10 imaging artifacts in glaucoma. And these are the papers on which these slides are based on so Snoween Liu, who is a medical student at Harvard, did some research with me in the summer and she went to UCSF afterwards for residency and is doing a fellowship now. But she basically went over a thousand spectralis OCT scans and found out for us and reported this in ajo, that one of the least common artifacts in glaucoma from 1 to 10 is a cut edge artifact. So less than 1% of the time, sometimes the OCT scan has just missing data. We just throw out that missing data. We can also have motion artifacts maneuver. Our patient moves so much that the OCT scan is not even in the display box. So that's garbage data. We can't use that in our clinical decision making. **[37:32]:** So sometimes we have incomplete Scans Less than 1% of the time, so we have no RNFL thickness values that are meaningful. Here's a normal example where the nerve fiber layer in white on the top, that's the top layer of the retina segmented normally in this red anterior border and the red posterior RNFL border. However, in incline segmentation, that red segmentation line doesn't go all the way to the end. And here you can see our patient's black RNFL thickness graph is truncated PPA associated error. Now, Joel Schumann and the greats before us determined that the optimal scan size is about 3.4 millimeters in diameter because that usually means goes around most areas of peripapillary atrophy. But sometimes that scan circle size is too small and we can be scanning over areas of peripapillary atrophy that are really large and that will give us garbage data. So again, we need to disregard those retinal nerve fibrillar thickness measures. **[38:34]:** Sometimes we get missing scan parts due to if that green circle scan for the RNFL printout is scanning over a floater and there's missing OCT data. **[38:46]:** Sometimes you get missing data, as in the serous OCT printout because the patient blinked. So this is not red deviation on the deviation map from RNFL thinning. This is quote unquote artifactual thinning from the patient blinking. So there's no data here. You can see the patient has sort of this epiretinal membrane or PVD pulling on the nerve fiber layer. But the machine incorrectly is drawing the anterior nerve fib layer border way up here where that white line is and not down here where the real anterior RNFL border is. So this gives us artifactually thick RNFL readings of around the 4,500 range. So again, don't use this measurement for your clinical decision making. Here you can see another example of this artifact where the machine incorrectly drew the RNFL anterior border. Here, in 2015, my patient had an accurate anterior RNFL border in red. However, a year later, the patient's RNFL thickness anterior border was incorrectly drawn and segmented in red. **[39:51]:** As way down here dips down here, where we can clearly see the RNFL is up here. So that's artifactual thinning. So inferiorly the patient's RNFL thinned and is now red. But again, that's just an artifact. So the bottom line is this is a computer error and my patient really had a stable rnfl value. Number four, more common. We can about 5% of the time have poor signals. So here's a Spectralis scan where the quality Q index is signal strength is 9, and for the Spectralis machine, 9 is low because the maximum value is 40. So I need to view that OCT scan as being suspect. It's like a blurry disk photo. How much can I really rely on that blurry disk photo? Or like the visual field that has 50% fixation losses, how reliable is that visual field test? So similarly, a quality score of 9 for an Oct scan is not good. **[40:47]:** However, for the Cirrus machine, which has a different signal strength out of 10, nine is fabulous. So the bottom line, or the take home point is you need to be aware of the range of signal strength for your machine because every signal strength range is different for different machines. Now, number three, more common is posterior RFL misidentification about almost 8% of the time. Now this makes sense that the computer has a harder time segmenting the posterior border of the rnfl because in a normal neurifiable layer, the RNFL is very reflective and easy to segment. However, with glaucoma, the glaucomatis RNFL is less reflective and hard to segment. So here a normal RNFL very easy to segment in white. Here, however, in a glaucoma, it's not so white and it's kind of grayish. So here you can see the machine incorrectly delineating the red posterior NFL segment as near where the inner plexiform layer is. **[41:44]:** So this is one of my favorite OCT artifacts. It's very easy to spot in the clinic when you're rushed. But basically take the time to look at the TISNET graph. If you see an RFL thickness of 0, red flags should go off in your brain because you know that this is an artifact because it is impossible physiologically to have a zero RNFL thickness, because again, we know that there's a floor effect. Everybody's neurofibril layer has 50 microns of non neuronal tissue. So let's say if your patient has a 0.9 cup, he'll still have a 50 micron neurofibrillator thickness because he has 50 microns of non neuronal tissue. So again, here's an OTC scan where the RNFL thickness is zero microns. And that's not normals, that's garbage data. Here is a different patient where you can see this is the micron scale. Patient has zero micron RNFL thickness. **[42:33]:** That's an artifact. That's garbage data. Don't use that printout. Here you can see the scan is terrible. And really you shouldn't look at this data because here the RNFL thickness is zero, which makes no sense because it should be at least 50 microns. So you know the data is bad. So number two, so getting to the more common artifacts. So this is something I call the Goldilocks artifact because as you know, Goldilocks, this little girl stumbled into the Three Bears house where she had porridge that was too hot, too cold and just right. So similarly, the PVD or posterior vitreous detachment associated error can be seen. So about 14, 15% of the time our OCT scans can see a PVD. Sometimes that PVD can cause the RNFL thickness to be segmented incorrectly, too thick. Sometimes that PVD can cause the RNFL to be incorrectly segmented too thin. **[43:29]:** And sometimes the PVD doesn't affect the segmentation by the computer. And the RNFL thickness can be just right or can be accurate. Meaning if it's just right, we can still use that RNFL thickness data despite that pvd. So number one, what is the most common artifact in some machines? Well, in the Spectralis machine, where the technician has to center the green circle scan around the optic nerve, that happens commonly in other machines like the Cirrus where the machine automatically centers the optic nerve RNFL thickness scan, that's less of an issue. But for the Spectralis machine or other machines where you have to have the technician manually center the RNFL scan circle over the optic nerve head, this is a problem. And the reason why is we know that the nerve fiber layer normally becomes thicker as you're closer to the optic nerve. Here you can see the thick black nerve fiber layer closer to the optic nerve and the thin nerve fiber layer further away from the optic nerve. **[44:27]:** So in this paper they took a normal patient and purposely decenter the RNFL thickness Scan in many different directions. And this is what they found in the well centered RNFL thickness scan. **[44:44]:** The patient's black nerve fiber layer thickness is fine. However, that same normal patient purposely had their artification scan decentered superiorly. And this causes artifactual thinning superiorly where there's red areas. So again, this is a traffic light. Red is a warning or bad. Yellow is sort of a warning, green is good. But here you can see superior thinning. And that's just artifactual thinning of the superior nerve fiber layer because the skin was decentered. So be aware, take the time in clinic that before you call your normal patient as worsening superiorly, make sure that the scan is centered properly. The same patient, normal patient had their scan inferiorly de centered, which caused inferior thinning on the tisnet graft. When the graft was nasally, dissentered causes artifactual nasal thinning. And when the patient scan was de centered temporally, there was artificial thinning temporarily. So top 10 reasons for artifacts decentration is number one. **[45:42]:** Pvd. Sometimes goldilocks artifact can cause artifactually thickening, thinning or no change. Posterior NFL misidentification almost 78% of scans and all these other ones are pretty rare. Poor signal strength in about 5% of scans. So you need to know the frequency because you need to know how likely are these artifacts seen in your clinic. So those are the top 10 artifacts. Now this is an interesting group of diseases. These are OCT diseases, meaning your patient can only be afflicted with these diseases if you have an OCT machine, which is the pathogen. So again, these are diseases that your patient can have if you have an OCT machine. **[46:30]:** So these are artifactual changes due to the OCT machine giving you incorrect measurements. Now, usually our OCT machines, artificial intelligence shows us red colors when the patient might have glaucoma changes, shows us green colors when the patient is normal. So if you see red on the color printout but your patient's normal, that's called red disease. That means your OCT machine is incorrectly giving your patient a diagnosis of glaucoma or a false positive when your patient is actually normal. So when you look at the nerve, it's 0.1. When you look at the field, it's normal. The vision is 2020. The patient only has an abnormality because they have red on the OCT printout. Now, there are many etiologies for red disease, meaning your patient's totally normal without glaucoma, but there's red on the printout. Some etiologies are the top 10 artifacts, which we just went over the most common one being decentration. **[47:27]:** So here's a real patient of mine with a right eye and a left eye. And I said, oh my goodness, my patient suddenly has a new superior area thinning. And especially Teresa knows, or you know, that supratemporal is the preferential location for glaucoma. This must be real. But I took the time to look at the scan and realize that the technician had decentered the scan. So I told the technician, can you reimage the patient? So she reimaged the patient and more carefully centered the scan. And now my normal patient is normal, not only by clinical exam, but by Oct. Everything is green. **[48:05]:** Aside from the top 10 artifacts, you can also have non glaucomatous causes of RNFL thinning. I call out this one because aside from other diseases like optic nerve diseases or retinal diseases that cause RFL thinning, one of the most common referrals that I get for RNFL thinning is myopia. Assuming your patient is Otherwise clinically normal, 20, 20, normal visual field, normal pressure, but they only have myopia, meaning there's a minus seven myope, so they have RNFL thinning. And the reason why is myopia can independently cause RNFL thinning. So again, the normative database in the OCT machines are non myopes. So again, if you have a myopic patient, you can RNFL thinning on the printout even though their disc is normal, their field is normal except for a rim artifact. And if you throw at that OCT data point, the patient's clinically normal without glaucoma. So bottom line is red disease or red on the OCT printout is a false positive diagnosis by the machine of glaucoma. **[49:09]:** So again, the take home point is don't use the OCT brains, use yours. Which is better because you have all the data and can tell your patient you are normal. You don't have glaucoma, but sorry, the OCT machine has some red because you're a myope or you have an artifact. Yellow disease is a false borderline value. So meaning if you look at your printout, there's yellow on the printout, meaning the machine saying, well wait, there's some nerve fiber layer thinning. However, Steven Spanberg published this paper showing that if you manually resegment or correct segmentation errors at RNFL, that 25% of the yellow color will go away and become green. So the bottom line is that segmentation error or OCT artificial intelligence can give you a false impression that Your patient may have RNFL thinning where if you just manually redraw the borders of the rnfl, then that patient actually has normal RNFL thickness. **[50:11]:** So the bottom line is yellow disease is a force borderline warning by the machine. And then of course, green disease is something not to be missed because this is the printout OCT printout telling you patient is normal. But again, you don't want to miss it because your patient really has glaucoma. So here's an example of a patient with green disease. You're busy and clinic. You only look at the OCT and say, oh, the patient has green colors. I know this machine. OCT is telling me the patient's normal because green means normal, green means normal. But again, don't use the OCT brains, use yours. Because you can look at the optic nerve and see a disc hemorrhage. You can look at the visual field and see the paracentral defect in this patient with classic focal RNFL thinning from normal tension glaucoma. So again, some focal RNFL thinning can be masked by the global RNFL summary numbers. **[51:03]:** But again, this patient has green disease. So the moral story is look at all the clinical data. Do not miss a diagnosis of glaucoma and do not miss the opportunity to treat this patient. Despite seeing green on the OCT printout. Like all diseases, green disease can progress. Here you can see a scan from2013 where the machine incorrectly caused this patient to have green areas on the nerve fiber layer even though they have a 0.9 nerve. However, three years later, the patient now has quote unquote totally normal nerve fiber layer thickness. However you look at the scan makes no sense. It's just incorrectly segmented where the RNFL thickness is artifactually high. But my patient still has a 0.9 nerve and they still have glaucoma. So the bottom line is green disease is a false negative and incorrect normal diagnosis by the OCT machine. So variations between machines. **[52:06]:** This is interesting because different measurements on different machines can happen in the same patient. So for example, same patient, different machines. Here is an actual example of a patient in the glaucoma service who initially had a stratus OCT. About 10 years ago, the global RNFOS was 104 microns. We decided to do the right thing and upgrade to spectral domain OCT technology. So this is a time domain OCT machine with two dimensional images. Serous is a spectral domain OCT machine and the patient's RNFL fin. So it's like oh my goodness, rule of five, this is more than five microns. Maybe this patient has RNFL thinning. Maybe this is progression in the green. Maybe the patient is going from, let's say a 0.1 cup to a 0.3 cup. Maybe this is clinically significant. However, the good thing is the patients switched doctors to one that preferred the Spectralis machine and they got back to their baseline of 105, 104 microns. **[53:04]:** So the bottom line is the take home point is serous machines give the lowest RNFL thickness values. So the bottom line is this stable normal patient had artifactual thinning because they simply used a cirrus machine. So here is a paper where an investigator in Colorado imaged 48 normal patients on a Saturday with four different machines same day. And he found that the lowest micron values were seen in the cirrus machine, just like our clinical example. So the bottom line is if your patient, let's say, is scanned by a different doctor in a different state with one of these machines and you're a cirrus lover, you need to be aware that that your first serous measurement is going to be lower than the patient's previous RW or Spectralis measurements. And this is not glaucoma progression, but simply using a new machine. And then the other take home point is that different machines have different normative databases. **[54:04]:** So a patient with an RNFL thickness value that might produce a green color on one machine may produce a yellow color on a different machine. So in summary, be aware that OCT can give you bad data, but the good thing is that you are armed with the knowledge that you can detect artifacts in the clinic. You know that OCT can give you the wrong diagnosis. So don't follow the red, yellow and green color scheme. And you know that machines that give the same regions of interest that these values, for example the RNFL thickness values are not interchangeable between machines. The RIM values are not interchangeable machines and the macular data is not interchangeable machines. **[54:54]:** So here is along my story of glaucoma imaging, another interesting picture. So I was privileged to meet and know Bruce Shields. He is a alumni of Mass Eye and Ear as is Janice Reed. And this is myself and Robin Cook who is a Mass Eye and Ear residency alumni. He wrote the book Coma and basically was the father of medical thrillers. I met him at an alumni meeting. I gave him my glaucoma surgery textbook and he wrote me this wonderful note. He said he read my glaucoma text and loved the character development and didn't want to wait for the movie. But here's this lovely note by one of our fellow ophthalmologists. So my academic career has enabled to meet wonderful people and then take home Point here is on the lines of clinical testing, if your OCT test does not match the other clinical data such as the visual field test, you really need to pause and think why? **[55:56]:** Why does the October not match the visual field? Maybe I need to look at the vision, check the pressure, talk to the patient and look at the fundus and then I can better discern should I be using that OCT test to make clinical decisions. So the bottom line, look at the big picture and make the right clinical decision for your patient. And then lastly, I want to have a shout out to some of the wonderful people in my academic career I worked with Katherine Mirando who is our current Mass Eye and Ear Glaucoma Fellow. She basically was first author on a glaucoma text on OCT artifacts. Ellie park was one of the best medical students that I ever worked with. She is now a resident at Bascombe Palmer and she is middle author on this chapter that we wrote for Julia Rossdale who is the editor of one of the best textbooks on OCT called OCT for Glaucoma. **[56:56]:** So get this if you feel that my lecture was not enough, another wonderful colleague is Don Boudenz. If you feel this lecture is not enough, get this Atlas of OCT for Glaucoma also co authored by Ellie park, amazing medical student who worked with me again. Now she is a rock star ophthalmology resident basketball and then lastly for myself, I want to give a special thanks to my research collaborators. I feel totally blessed that I was able to do glaucoma imaging research for the top for the past two decades. But I feel like the proverbial turtle on the fence post because I was buoyed up there by my research collaborators. Now, just as a tangent and a detour, these are not my research collaborators, but a special shout out to the fathers of oct, Joel Schumann, Carmen Poliofito and James Fujimoto, who I had the pleasure of meet at the World Glaucoma Congress in Boston. **[57:55]:** But they developed this technology time domain oct, which we've been using in the 1990s and that gave us 2D pictures like this one of the eye. But 3D imaging came about at the turn of the century around 2000 with my research collaborator Johannes Deboer. He got a new job at Mass General Wellman lab in 2001 and that was around the same time I was hired at Mass Eye and Ear. He was looking for a clinical collaborator. And he tells me in retrospect that he did not seek me out because of my towering intellect and my excellent research acumen. But he said, you know, Teresa, I work at Mass General and you work at Mass Eye and Ear and you're next door so you can recruit patients for my OCT studies. So we started our wonderful collaboration in 2001 and two years later, he and the research group there, these smart PhD people, Johannes DeBoer, Brett Bohm and all these other wonderful people, developed spectral domain OCT and swept source oct, which we know now enables us to image the eye in 3D in video rate imaging. **[58:59]:** So now video imaging of the back of the eye is what we do. And again, my career has been blessed by these people. We actually presented this video at the first American Glaucoma Society meeting where we presented as first author. This was back in the early 2003, 2004 time. **[59:23]:** And that was Johannes left. **[59:32]:** And then lastly, I'd like to give a special shout out to all the wonderful, brilliant students who have worked with me. Thank you for publishing all those papers with me. So bottom line, thank you for spending this hour with me. If you have any questions at all about oct, feel free to email me at teresachen, eei, Harvard. Edu. And certainly if you have any interest in glaucoma imaging research, please feel free to reach out. Thank you again and I hope this has been helpful. Best wishes on your career.