Transcript
You know, I always want to be the last speaker at 9:00 on Friday night. This is gonna be the best place to be. I'm gonna try to make this as bearable as possible and as short as possible. Basically, what I wanna try to tell you is a story. Let's see. Next slide. I'm a consultant for Brainlab. I have a research grant from them as well, as well as Baxter and Stryker. I agree with what Dr. Byerson said about teams being able to help us. My mentor was Andy Parsa, and he used to say this phrase that we shouldn't be afraid of failure, but we should be afraid of succeeding at the wrong things. And I think it's really important what your neurosurgeon does, and I think it's just as important what your neurosurgeon doesn't do. And that's the story I hope to tell you about acoustic neuromas. So this is the tumor pressing along the facial nerve, and this is the nerve here, and this is the tumor, and how do we preserve facial nerve? How do we preserve cochlear nerve? And what's the story that we can tell?
Well, we started, and this is... I can't believe that this dates us. This is over a decade ago looking at hearing preservation. That radiosurgery is pretty good for hearing preservation and treating acoustic neuromas. We looked at facial nerve function and radiosurgery's even better for facial nerve preservation than surgery, and this is also proved by many analyses. And so then we want to look at radiosurgery and said, "Well, what kind of radiosurgery should you be doing?" We looked at patients who got treated with radiosurgery in single fraction or radiotherapy in 28 fractions. And we found that the people who were treated in the radiotherapy group, they had better hearing preservation. That something about doing bite-sized doses in these patients had better hearing preservation. And that there's a story there that when you have patients who have good hearing and you look at them with surgical options, they need to know that with radiosurgery now with radiotherapy, their hearing preservation may be better.
And the next story...part of the story that we want to look in, this is the actuarial hearing preservation for that is that once you go past 10 years, it may not matter whether you get radiosurgery or radiotherapy, but in the initial first 10 years, that radiotherapy has a significant improvement in hearing preservation. And we published this in the "Red Journal" I believe, and this was about five or eight years ago. And the next place we wanted to go with this story was why is this so? Why is radiotherapy better for hearing preservation than radiosurgery? And so we went back...and by we, I mean my research team and some of them are here, is that we went back and we contoured the cochlear on all these plans.
So this is how we do it. This is the T2 MRI, and we went back and saw the cochlea here. You can see it's hyperintense. And then we overlaid that with the CT scans that this is the cochlea. What's the radiation dose to the cochlea? And we found a really interesting finding is that when you do single-dose radiosurgery, the cochlear dose matters. That if your cochlear dose stays below 8 gray, you have a very, very significant p-value of 0.03, very significant rate of hearing loss if your cochlear dose goes above this point. But in radiotherapy... Sorry. In radiotherapy, we didn't find that. That something about being in the bite-sized dose of cochlear dose to that matter, but in the radiosurgery, the cochlear dose made a huge difference. And so the part of that story is that we wanna keep the cochlear dose less than 8 gray.
And in the modern area, radiosurgery is an adjunct to surgery. I am not here advocating that we radiate everything. I'm not here advocating that we operate on everything. I'm saying that we should use the best of all worlds. And in the modern era, this is also another paper for my mentor, we found that when you're looking at gross total resection and near-total resection, that the recurrence rate, whether you cut the whole thing out, or you have a near-total resection and you radiate the residual, the recurrence rate is going to be about the same. Philip Theodosopoulos, who was gonna give this initial talk, he has found that same thing at UCSF. Most recently, he published that the recurrence rate, the growth rate was about 10%. And it's very similar to the data that my mentor published. This is about 15 years ago, that this data is about 10% of these acoustic neuromas, they're going to recur no matter what.
Well, how does this tell a better story? What is personalized neurosurgery? Well, at UCLA, one of the things we do is we do DTI. Can we see the facial nerve before surgery? And in this series of 38 patients, we found that in about 95% of cases, we can see the facial nerve before we make skin incision. We can do tractography to know where the facial nerve is. And then about 90% of the times, it's accurate in its positioning. Does it make clinical difference? No, but we can make really nice pictures. This is the Brainlab navigation. Here, you can see the nerve, right? We are doing tractography. We are seeing the facial nerve. Look at this. It's draped. You see the facial nerve is draped, anterior and superior, just where it's supposed to be. This works 95% of the time. It's 90% accurate. But these are all my cases. I still have about 85%, you know, House-Brackmann preservation and about 10% House-Brackmann, not so good. And that's still the case. It didn't make such a big difference. It didn't make a clinical difference, but it makes really nice pictures.
Well, this is where I do believe that adaptive hybrid surgery can come into play. Is that if you have software that can help you identify, what is a better resection? What's the ideal resection? Can this make us better surgeons? Can this make us better doctors? And this is really a question is, "Is it worthwhile to have this discussion?" I'm not here to argue gross total resection or radiation. I'm here to ask you, "If you had to leave residual, if you had to leave residual, what does your residual look like?" And there's all these experts, and all these great surgeons have a consensus on what that residual should look like. And that's the question we want to ask because what we're trying to find is as you resect more tumor, I believe the risk of facial injury goes up. As the tumor gets smaller, the radiation risk falls down. There must be a point where you can get the best of all worlds, where you can diversify and get all the ultimate lower risk. I believe this point exists. And so the question I wanna ask is, "What is the ideal resection?" And who should answer that question? Should we have neurosurgery and ENT radiation oncology board saying, "Yes, that's the ideal resection?" And so then you go ahead and do that, and that we should agree on what the goal is before every surgery? And what I hope to show you in the story is that that may not be the case. And so these are all my cases.
We did a study where we took my patients, my surgeries, and here's the volumetrics of the initial tumor. And what we said is we asked this surgeon right here, "What's the ideal resection, and what's the residual? If you have to leave residual, what's the residual?" And this doctor here said, "You should leave 24%." And the next one said 16%. This pointer is dying. Next one said, here, "This is the next resection. Do 16%, 23%, 34%." And I did this surgery, and the actual residual I left was 8%. And this is what the computer suggested may be an ideal resection. This is what the software said is something to do. And it's higher. And this is the actual case. This is the actual tumor here. This is my post-operative residual. This is what I left here on the brain stem. This is what the computer said may be ideal, that the surgery could be shorter and maybe the patient outcomes could be better.
And this is what the skull-based surgeons contoured. Here's this contour here, this contour here, this contour here, and what I want you to appreciate is there's no consensus. There's no consensus. When I ask you, "What's the ideal resection?" It's like asking you, "What's your favorite ice cream flavor?" Every neurosurgeon's gonna have a different idea, but there should be, there should be a standard. And what does the standard look like? And so we did this with multiple patients in multiple rounds, and you can see that in every single time, there's a variance from these patients and that the software always said, "You can leave more tumor behind."
And we actually presented this earlier today in the Vestibular Schwannoma section, surgeon plans sub-total resection. And here, here's what we found. This is surgeon one, surgeon two, and surgeon three, and surgeon four. Our variance is in this area. When we said, "What's the resection? What's the residual tumor to leave behind?" We all, myself included, said, "This is the kind of tumor that you should leave behind." There's variability in this. And what we also found is that the surgeons were always more aggressive than the software suggested. That perhaps we are being too much. Like in the words of my mentor, maybe we're being successful at the wrong things. Maybe we're being successful at the wrong things. And if we can prove that surgery is shorter, facial nerve outcomes are better, and patients do better in this group, maybe that's gonna be a game-changer, maybe that's gonna change us.
And so we did this with my patients. All of these are my outcomes, all of these House-Brackmann 3-6, these are mine. So what we found, and we compared in this series of patients, these are the patients in which I concur. I didn't use AHS. I just did the surgery. And I just inadvertently came up to the same conclusion that the computer came up with. And in those patients, these patients got House-Brackmann 1-2. When I deviated further from whatever the computer recommended, my outcomes weren't as good. Now, this was not statistically significant. This cohort was too small. But it suggests that the computer may know better than me.
And that's really hard to accept. That's really hard to accept, but that this ideal resection, it may be out there. It may be some kind of volume that we can get. Then it may be something that the software and the computer, and as we get more data points, as imaging gets better, like Dr. Byerson says, as our imaging gets better, as there's more data points, as there's more computing power to process all those data points, we may come up with a better conclusion. And I believe this is gonna happen. And this isn't only being validated on the surgical side, what I just showed you was my surgery, this is the radiosurgical side. They're validating the exact same software on the other side. Can they do this with the radiation? And here, looking here at the Karolinska Study results, they essentially proved that this adaptive hybrid software accurately is verifiable and validated on the radiosurgery side. Just like I'm validating it on the surgery side, they're validating it on the other side, that the computer may be able to tell us what the ideal resection rate is. And this is the future.
I took this off social media. This is off the Brainlab Twitter. This is Stefan Vilsmeier, the CEO for Brainlab, and this is with Satya Nadella, who's the CEO of Microsoft. They're working together. They're looking at AI. They're looking at these data points because when you do registration, when you're doing these computer registrations, it's immense data, there made be an answer. It's coming. This is from the Stanford website last year, November of 2017. The radiologists have to watch out why the Stanford algorithm is better than radiologists at diagnosing pneumonia. The computer is better than the human eyes. And there's another paper that I didn't note here, where they looked at breast cancer markers, and they found that the computer was better at recognizing the markers of invasiveness and staging the cancer than human pathologists, almost two times. It's a TEDx talk if you wanna Google it. The computers are getting better.
And if you don't think that, and I know I wanna bristle, I wanna say, "Look, I will always know better than the computers. I'm a surgeon. The computer can't do a residency. Computer can't do hemicolectomy." But I want you to pause. And if you don't think this is happening, I want you to think of an era, back when I learned to drive, a long time ago now, when you got into a troublesome area and you got into an area where you had to hit the brakes, they taught me to pump the brakes. Do you remember this? For those of you who are older than I, you remember, they didn't say you just slam the brakes, what? They said do this, right? Pump the brakes. They said, "Pump the brakes."
But now, today in 2018, nobody says, "Pump the brakes." Why? Because you have something called anti-lock brakes. It sounds really safe and benign, but you know what anti-lock brakes are? It's computer-assisted braking. The computer can brake faster than you can. The faster than you can. My wife, she drives an Acura, when it drifts out of the lane, it starts blinking. And my colleagues who drive... I drive a Honda, but my colleagues who drive Teslas, they don't even have to drive the cars on the freeway. And you don't think this is gonna affect neurosurgery? It's coming, and this is generation 1.0. And this is the generation that shows it's coming and talking for us. I think that's the story that I wanna tell.
I have to thank my lab who's helpful, and the brain tumor program that funds me and all the funding resources. And that's all. Thank you.
Well, we started, and this is... I can't believe that this dates us. This is over a decade ago looking at hearing preservation. That radiosurgery is pretty good for hearing preservation and treating acoustic neuromas. We looked at facial nerve function and radiosurgery's even better for facial nerve preservation than surgery, and this is also proved by many analyses. And so then we want to look at radiosurgery and said, "Well, what kind of radiosurgery should you be doing?" We looked at patients who got treated with radiosurgery in single fraction or radiotherapy in 28 fractions. And we found that the people who were treated in the radiotherapy group, they had better hearing preservation. That something about doing bite-sized doses in these patients had better hearing preservation. And that there's a story there that when you have patients who have good hearing and you look at them with surgical options, they need to know that with radiosurgery now with radiotherapy, their hearing preservation may be better.
And the next story...part of the story that we want to look in, this is the actuarial hearing preservation for that is that once you go past 10 years, it may not matter whether you get radiosurgery or radiotherapy, but in the initial first 10 years, that radiotherapy has a significant improvement in hearing preservation. And we published this in the "Red Journal" I believe, and this was about five or eight years ago. And the next place we wanted to go with this story was why is this so? Why is radiotherapy better for hearing preservation than radiosurgery? And so we went back...and by we, I mean my research team and some of them are here, is that we went back and we contoured the cochlear on all these plans.
So this is how we do it. This is the T2 MRI, and we went back and saw the cochlea here. You can see it's hyperintense. And then we overlaid that with the CT scans that this is the cochlea. What's the radiation dose to the cochlea? And we found a really interesting finding is that when you do single-dose radiosurgery, the cochlear dose matters. That if your cochlear dose stays below 8 gray, you have a very, very significant p-value of 0.03, very significant rate of hearing loss if your cochlear dose goes above this point. But in radiotherapy... Sorry. In radiotherapy, we didn't find that. That something about being in the bite-sized dose of cochlear dose to that matter, but in the radiosurgery, the cochlear dose made a huge difference. And so the part of that story is that we wanna keep the cochlear dose less than 8 gray.
And in the modern area, radiosurgery is an adjunct to surgery. I am not here advocating that we radiate everything. I'm not here advocating that we operate on everything. I'm saying that we should use the best of all worlds. And in the modern era, this is also another paper for my mentor, we found that when you're looking at gross total resection and near-total resection, that the recurrence rate, whether you cut the whole thing out, or you have a near-total resection and you radiate the residual, the recurrence rate is going to be about the same. Philip Theodosopoulos, who was gonna give this initial talk, he has found that same thing at UCSF. Most recently, he published that the recurrence rate, the growth rate was about 10%. And it's very similar to the data that my mentor published. This is about 15 years ago, that this data is about 10% of these acoustic neuromas, they're going to recur no matter what.
Well, how does this tell a better story? What is personalized neurosurgery? Well, at UCLA, one of the things we do is we do DTI. Can we see the facial nerve before surgery? And in this series of 38 patients, we found that in about 95% of cases, we can see the facial nerve before we make skin incision. We can do tractography to know where the facial nerve is. And then about 90% of the times, it's accurate in its positioning. Does it make clinical difference? No, but we can make really nice pictures. This is the Brainlab navigation. Here, you can see the nerve, right? We are doing tractography. We are seeing the facial nerve. Look at this. It's draped. You see the facial nerve is draped, anterior and superior, just where it's supposed to be. This works 95% of the time. It's 90% accurate. But these are all my cases. I still have about 85%, you know, House-Brackmann preservation and about 10% House-Brackmann, not so good. And that's still the case. It didn't make such a big difference. It didn't make a clinical difference, but it makes really nice pictures.
Well, this is where I do believe that adaptive hybrid surgery can come into play. Is that if you have software that can help you identify, what is a better resection? What's the ideal resection? Can this make us better surgeons? Can this make us better doctors? And this is really a question is, "Is it worthwhile to have this discussion?" I'm not here to argue gross total resection or radiation. I'm here to ask you, "If you had to leave residual, if you had to leave residual, what does your residual look like?" And there's all these experts, and all these great surgeons have a consensus on what that residual should look like. And that's the question we want to ask because what we're trying to find is as you resect more tumor, I believe the risk of facial injury goes up. As the tumor gets smaller, the radiation risk falls down. There must be a point where you can get the best of all worlds, where you can diversify and get all the ultimate lower risk. I believe this point exists. And so the question I wanna ask is, "What is the ideal resection?" And who should answer that question? Should we have neurosurgery and ENT radiation oncology board saying, "Yes, that's the ideal resection?" And so then you go ahead and do that, and that we should agree on what the goal is before every surgery? And what I hope to show you in the story is that that may not be the case. And so these are all my cases.
We did a study where we took my patients, my surgeries, and here's the volumetrics of the initial tumor. And what we said is we asked this surgeon right here, "What's the ideal resection, and what's the residual? If you have to leave residual, what's the residual?" And this doctor here said, "You should leave 24%." And the next one said 16%. This pointer is dying. Next one said, here, "This is the next resection. Do 16%, 23%, 34%." And I did this surgery, and the actual residual I left was 8%. And this is what the computer suggested may be an ideal resection. This is what the software said is something to do. And it's higher. And this is the actual case. This is the actual tumor here. This is my post-operative residual. This is what I left here on the brain stem. This is what the computer said may be ideal, that the surgery could be shorter and maybe the patient outcomes could be better.
And this is what the skull-based surgeons contoured. Here's this contour here, this contour here, this contour here, and what I want you to appreciate is there's no consensus. There's no consensus. When I ask you, "What's the ideal resection?" It's like asking you, "What's your favorite ice cream flavor?" Every neurosurgeon's gonna have a different idea, but there should be, there should be a standard. And what does the standard look like? And so we did this with multiple patients in multiple rounds, and you can see that in every single time, there's a variance from these patients and that the software always said, "You can leave more tumor behind."
And we actually presented this earlier today in the Vestibular Schwannoma section, surgeon plans sub-total resection. And here, here's what we found. This is surgeon one, surgeon two, and surgeon three, and surgeon four. Our variance is in this area. When we said, "What's the resection? What's the residual tumor to leave behind?" We all, myself included, said, "This is the kind of tumor that you should leave behind." There's variability in this. And what we also found is that the surgeons were always more aggressive than the software suggested. That perhaps we are being too much. Like in the words of my mentor, maybe we're being successful at the wrong things. Maybe we're being successful at the wrong things. And if we can prove that surgery is shorter, facial nerve outcomes are better, and patients do better in this group, maybe that's gonna be a game-changer, maybe that's gonna change us.
And so we did this with my patients. All of these are my outcomes, all of these House-Brackmann 3-6, these are mine. So what we found, and we compared in this series of patients, these are the patients in which I concur. I didn't use AHS. I just did the surgery. And I just inadvertently came up to the same conclusion that the computer came up with. And in those patients, these patients got House-Brackmann 1-2. When I deviated further from whatever the computer recommended, my outcomes weren't as good. Now, this was not statistically significant. This cohort was too small. But it suggests that the computer may know better than me.
And that's really hard to accept. That's really hard to accept, but that this ideal resection, it may be out there. It may be some kind of volume that we can get. Then it may be something that the software and the computer, and as we get more data points, as imaging gets better, like Dr. Byerson says, as our imaging gets better, as there's more data points, as there's more computing power to process all those data points, we may come up with a better conclusion. And I believe this is gonna happen. And this isn't only being validated on the surgical side, what I just showed you was my surgery, this is the radiosurgical side. They're validating the exact same software on the other side. Can they do this with the radiation? And here, looking here at the Karolinska Study results, they essentially proved that this adaptive hybrid software accurately is verifiable and validated on the radiosurgery side. Just like I'm validating it on the surgery side, they're validating it on the other side, that the computer may be able to tell us what the ideal resection rate is. And this is the future.
I took this off social media. This is off the Brainlab Twitter. This is Stefan Vilsmeier, the CEO for Brainlab, and this is with Satya Nadella, who's the CEO of Microsoft. They're working together. They're looking at AI. They're looking at these data points because when you do registration, when you're doing these computer registrations, it's immense data, there made be an answer. It's coming. This is from the Stanford website last year, November of 2017. The radiologists have to watch out why the Stanford algorithm is better than radiologists at diagnosing pneumonia. The computer is better than the human eyes. And there's another paper that I didn't note here, where they looked at breast cancer markers, and they found that the computer was better at recognizing the markers of invasiveness and staging the cancer than human pathologists, almost two times. It's a TEDx talk if you wanna Google it. The computers are getting better.
And if you don't think that, and I know I wanna bristle, I wanna say, "Look, I will always know better than the computers. I'm a surgeon. The computer can't do a residency. Computer can't do hemicolectomy." But I want you to pause. And if you don't think this is happening, I want you to think of an era, back when I learned to drive, a long time ago now, when you got into a troublesome area and you got into an area where you had to hit the brakes, they taught me to pump the brakes. Do you remember this? For those of you who are older than I, you remember, they didn't say you just slam the brakes, what? They said do this, right? Pump the brakes. They said, "Pump the brakes."
But now, today in 2018, nobody says, "Pump the brakes." Why? Because you have something called anti-lock brakes. It sounds really safe and benign, but you know what anti-lock brakes are? It's computer-assisted braking. The computer can brake faster than you can. The faster than you can. My wife, she drives an Acura, when it drifts out of the lane, it starts blinking. And my colleagues who drive... I drive a Honda, but my colleagues who drive Teslas, they don't even have to drive the cars on the freeway. And you don't think this is gonna affect neurosurgery? It's coming, and this is generation 1.0. And this is the generation that shows it's coming and talking for us. I think that's the story that I wanna tell.
I have to thank my lab who's helpful, and the brain tumor program that funds me and all the funding resources. And that's all. Thank you.