Dan Keller 0:02 Welcome to this episode of Substantial Matters: Life and Science of Parkinson's. I'm your host, Dan Keller. At the Parkinson's Foundation, we want all people with Parkinson's and their families to get the care and support they need. Better care starts with better research and leads to better lives. In this podcast series, we highlight the fruits of that research, the treatments and techniques that can help you live a better life now, as well as research that can bring a better tomorrow. Parkinson's disease today is a clinical diagnosis, as it has always been. It's based on a physician's observation of clinical signs and an individual's report of their symptoms. In general practice, there are no biomarkers for it, such as a blood test or readings from a machine, as there is, for example, with an electrocardiogram for heart disease, but based on advances in computing power to analyze masses of data that may be about to change with the development of artificial intelligence, or AI. AI uses computers to sort through very large streams of data to discern patterns of interest. It's like someone fly fishing and observing lots of ripples and eddies in a stream and being able to pick out the ones indicating where fish will gather to feed. So it was recently when a group of researchers used AI to detect breathing patterns among people with Parkinson's disease to see if those patterns differed from ones seen among healthy control subjects. They published their report of the study in the prestigious journal Nature Medicine. In this episode, co-author of this study, Dr. Aleksandar Videnovic of Harvard Medical School and chief of the Division of Sleep Medicine at Massachusetts General Hospital, describes how the study was done, its findings, and where AI may fit into clinical practice in the diagnosis of PD and monitoring its severity and progression. Let me just ask you, first, what was the purpose of this study?
Dr. Aleksandar Videnovic 2:30 Well, the purpose of this study was to learn whether we can use breathing patterns to determine who may have Parkinson's disease and who may not have it, and also to decide whether analyzing breathing patterns can help us comment or decide on the severity of Parkinson's disease, its progression, and even potentially understanding the risk for developing Parkinson's disease.
Dan Keller 3:04 Can you briefly describe the study method?
Dr. Aleksandar Videnovic 3:08 Yes, the study that we are talking about used this amazing technology that has been developed by Professor Dina Katabi from Massachusetts Institute of Technology, and this methodology basically uses radio waves that are being emitted in an environment, in a home environment, for example, and then as they spread through that environment and bounce off of various structures, we can really capture this radio signal and analyze how does it reflect a person's body. The methodology is using these radio signals, and by analyzing these radio signals, one can really analyze breathing patterns of individuals, and by analysis of these breathing patterns we can see whether there are differences among healthy individuals versus those who have Parkinson's disease. The method uses artificial intelligence and big data analysis, and it has very complicated algorithms how these radio signals are analyzed and detected.
Dan Keller 4:23 What kind of equipment is involved? Can it be done at home? Does it have to be done at a sleep center or laboratory?
Dr. Aleksandar Videnovic 4:30 Well, the beauty of this is that this can be done at home. It is a device that can be positioned in an individual's home, hallway, bedroom. Patients normally just go to sleep, do not have any devices on themselves, and that radio sensor that is really deployed in an individual's bedroom analyzes these radio reflections from the environment while patients are sleeping to extract their breathing pattern.
Dan Keller 5:05 I guess, at this point, what were the study findings so far?
Dr. Aleksandar Videnovic 5:11 Well, the study findings are very exciting. By analysis of this breathing signal, we managed to distinguish between individuals who are healthy individuals and those who have Parkinson's disease. Analysis of breathing signals in individuals with Parkinson's disease was sufficiently different from individuals who do not have the disease. Furthermore, we were also able to predict the severity of Parkinson's disease by analyzing breathing patterns while individuals are asleep, and also to comment on disease progression, as well as the risk of development of Parkinson's disease. One can ask, how was this achieved, right, to comment whether one has Parkinson's disease or not, but also to understand the risk of developing Parkinson's disease, and that was achieved because the study team has access to various types of datasets, and within the various types of datasets, which were composed of individuals who are healthy and those who have Parkinson's disease, and those who did not have Parkinson's disease but subsequently developed it, we managed to ask these different types of questions.
Dan Keller 6:35 A couple of measures that come up in studies often are sensitivity and specificity, sensitivity being how good is the method at detecting something if it's really there. Is there a number for the sensitivity of this method so far?
Dr. Aleksandar Videnovic 6:52 Well, the sensitivity and specificity for the method has been really very good, if not excellent. It was in the high 80s and 90s, and at some point even reaching 100%. The beauty of this technology is that really we can get repetitive data by analyzing breathing signals on repetitive nights, and if one analyzes breathing signals for let's say 10 to 12 days, the sensitivity and specificity of this method is approaching almost 100%. And we kind of need to remember that patients really do not need to do anything, they just go to bed, they sleep, and this device that captures their breathing signals is sitting in their bedroom doing all the work, so I think this is a very promising digital biomarker of Parkinson's disease, and we certainly need to employ it in a larger number of individuals with Parkinson's disease to further develop the potential of this digital biomarker.
Dan Keller 7:58 In terms of its specificity, how specific is it for detecting Parkinson's disease, its severity, its progression versus other similar movement disorders, such as Parkinson's Plus or progressive supranuclear palsy or multiple system atrophy, dementia with Lewy bodies—all sorts of things that have to be sorted out?
Dr. Aleksandar Videnovic 8:21 This is an excellent question. This particular study has been centered specifically on Parkinson's disease, so we do know how this methodology behaves in other atypical Parkinsonian syndromes. In this study, there has been a comparison between patients with Parkinson's disease and Alzheimer's disease, and there has been a good sensitivity and specificity in differentiating Parkinson's disease patients from Alzheimer's disease patients, but about other Parkinsonian conditions we really don't have that data at this point.
Dan Keller 9:02 How well does it compare to the standard traditional measures used in Parkinson's disease, such as the Movement Disorder Society Unified Parkinson's Disease Rating Scale, the MDS-UPDRS?
Dr. Aleksandar Videnovic 9:16 One segment of this study looked at the possibility to monitor the disease progression, and that part of the study revealed that changes in the breathing signal within a six to 12-month period are quite more sensitive to change than scores gathered by performing MDS-UPDRS, which is traditionally used to monitor and measure the burden of Parkinson's disease.
Dan Keller 9:46 In regard to that, I guess we would ask, what are the implications of these findings? What do you think the new tool may provide, or does it solve a problem or facilitate something that needs help?
Dr. Aleksandar Videnovic 10:01 In my mind, as a Parkinson's disease specialist, the main findings of this study are that we have a possibility for a great new digital biomarker for Parkinson's disease, that this biomarker can contribute both to the diagnosis of Parkinson's disease and monitoring its progression, and that it is non-invasive, and it's easy to be implemented in a patient's own home. The other very important implication is that this biomarker can be employed in clinical trials, which need to monitor the effectiveness of an intervention, either in addressing symptoms of Parkinson's disease or the progression of neurodegeneration, and having this tool can quite substantially improve the ability to monitor the effects of certain interventions that we are employing in clinical trials. It can also be widely employed in terms of geography, right, and can be used to monitor individuals who may not have such great access to care, who may have difficulties to visit their providers due to disease severity or geographic barriers to the care, and these are some of the major advantages that this new system can provide our community and our patients.
Dan Keller 11:26 Do you think it's more of a global assessment, sort of Parkinson's overall, or does it correlate with especially motor symptoms, or does it also correlate with some of the autonomic symptoms, like low blood pressure and constipation, and even mood, and things like that?
Dr. Aleksandar Videnovic 11:47 That's a great question. The routinely used instrument, MDS Unified Parkinson's Disease Rating Scale, has four subscales that are measuring various aspects of the disease. For example, Part III of the MDS-UPDRS assesses the burden of motor symptoms. Part II addresses the issues of activities of daily living. Part I is centered on some of the non-motor manifestations of the disease, and interestingly, signals obtained through analysis of breathing patterns really correlated quite, quite well with each of these three subscales of the MDS-UPDRS I mentioned—Part I, Part II, and Part III. Obviously, this needs to be studied even further, but these findings that we have gathered through the study that has been published suggest that various aspects of Parkinson's disease comorbidity may be correlating well with analysis of breathing signals.
Dan Keller 12:48 Now, as I understand it, this study involved people with Parkinson's disease. Do you think it will ever be useful in predicting risk for developing Parkinson's disease?
Dr. Aleksandar Videnovic 13:01 It may be, but that would really require more investigation, specifically getting a larger number of patients who would be studied using this technology. In a small subgroup of patients who at some point had acquired this breathing analysis and, at that point, did not have the disease, and then subsequently moved on to develop Parkinson's disease, and have the analysis of breathing signals at that point as well, it seems that this technology was able to pick up signals that would predict conversion to Parkinson's disease. The groups here were quite small. Larger studies will be needed, and certainly, what's going to be very exciting is to employ this technology in individuals who are at risk of developing Parkinson's disease, such as individuals with, for example, REM sleep behavior disorder, and see whether their breathing patterns are different than healthy individuals, and whether within those people who have REM sleep behavior disorder, we may identify those who are at more imminent risk for progression to Parkinson's disease as we diagnose it today, having all of these motor manifestations present at the time of diagnosis.
Dan Keller 14:22 And I take it REM sleep behavior disorder is a very early symptom, maybe even 10 years before the actual development of Parkinson's.
Dr. Aleksandar Videnovic 14:32 Yeah, this is a basic sleep disorder, which is known to represent a very early stage of a neurodegenerative process relevant to accumulation of pathological synuclein, and it can take years before these individuals go on to developing tremors and slowness and stiffness, which are the cardinal features of Parkinson's disease and necessary for the diagnosis of the disease at the present time.
Dan Keller 15:00 If this becomes a useful clinical tool, and even in routine practice, if or when do you think that might happen?
Dr. Aleksandar Videnovic 15:11 Well, while this is really very exciting work, and an amazing technology, yet a very simple technology, we would need to do quite a bit of more work employing a larger number of cohorts within the populations affected with Parkinson's disease before this transitions to either clinical care or some research implications. So this is all very promising, and a lot of additional work in conducting larger experiments and clinical investigations will be needed before we transition it to routine use.
Dan Keller 15:46 I really appreciate it. Thank you. This is quite an amazing technology. It seems that artificial intelligence is bursting out everywhere, really empowering tests that never could be done before.
Dr. Aleksandar Videnovic 15:58 Yes, definitely. Thank you so much, Dan. I really appreciate your time and the opportunity. Thank you. Be well.
Dan Keller 16:14 In the publication of the study, the authors note that the AI method they used has the potential to become both a diagnostic tool for PD and one to monitor progression. An advantage it could offer is that it's objective, so it would not have the subjective aspect of perceiving signs and symptoms that a clinician or a person with PD would have. Thus, it may be more sensitive to changes in their condition, as well as to make an initial diagnosis earlier. The technique can use wireless radio signals to monitor nocturnal breathing in a person's own home, so measurements could be collected every night without even being in contact with the person. Because PD progresses relatively slowly, current clinical methods may be relatively insensitive to small changes, whereas the AI method should be able to readily detect disease progression. Remote monitoring in a person's own home also offers the possibility of bringing expertise usually found in large medical centers out to more rural areas, reducing clinic visits and even offering clinical trial participation more widely. And given the sensitivity of the AI method to small changes in one's condition, it may help speed up and shorten the duration of clinical trials of drugs to slow down progression, reducing the cost of the trials and ultimately of drug development. The study investigators did note some current limitations of their work. First, they tested the method in people with Parkinson's, so they don't know if it can distinguish PD from certain other neurological conditions, especially those that are called Parkinson's Plus, or if it can differentiate among different subtypes of PD, but with more development and validation, this artificial intelligence method, based on breathing patterns, may become the first objective tool for diagnosing PD, monitoring changes in the disease, and helping in studies of drugs to slow its progression. To learn more about this study, search Artificial Intelligence on parkinson.org/blogs and you'll find the article titled "Artificial Intelligence Study Detects Parkinson's from Breathing Patterns." April is Parkinson's Awareness Month, and this year we're going to Take Six for PD. New research shows that someone is diagnosed with Parkinson's disease every six minutes. Take six minutes this month to spread awareness and help create a world without Parkinson's. Visit parkinson.org/awareness to find out more news and updates. Future events and resources are available by joining our email list at the bottom of our website's homepage. If you want to leave feedback on this podcast or any other subject, you can do it at parkinson.org/feedback. If you enjoyed this podcast, be sure to subscribe and rate and review the series on Apple Podcasts or wherever you get your podcasts. At the Parkinson's Foundation, our mission is to help every person diagnosed with Parkinson's live the best possible life today. To that end, we'll be bringing you a new episode in this podcast series every other week. Till next time, for more information and resources, visit parkinson.org or call our toll-free helpline at 1-800-4PD-INFO, that's 1-800-473-4636. Thank you for listening.
