 Good morning, thank you very much for you know joining today So it's gonna be a very up sorry You know very exciting Workshop to bring in people from AI and also people from mental health practice and also research So we are you know, literally in the same room, you know open a discussion on How AI can be utilized for the you know mental health? So for your information, you know, we're gonna Try this a side or calm. So if you have any questions to the you know invited the speakers or Later doing the you know panel discussion if you have any questions then Go to slido.com slash air air 2019 and then you know type in your you know cash on that, you know, we can actually see from our laptop Okay, so somehow we'll you know project and then we try to answer to your question Okay, so I want to open this workshop with a By introducing this recent, you know the report. So this one came out CCC and triple AI recently actually the last month and it shows, you know, 20 year Community roadmap for the AI research in the US and they mention about, you know, six societal You know areas that will be transformed by AI, right? As you can see, you know, health care education and business innovation national security social opportunity and Scientific, you know discovery. So, you know, this seems not really surprising to me But one thing I found a very interesting thing is that They are talking about the prevention of in this and also Elderly care and mental health and also behavioral, you know, health, right? So, you know, in a way the discussion that we are doing today is very well-aligned With you know, this community they are looking at, you know, 20 year down the road from now. Okay So I want to give you a little bit of a background over our team. So today we're gonna have three invited speakers and and also many You know, other speakers most of them are came from UC San Diego. So UC San Diego is one of our IBM AI Horizon Network University, so we are doing research together on AI for healthy aging So we are looking at physical and cognitive and, you know, mental health changes of all the others all the others by definition, you know, 65 plus years old and You know, we are collecting multimodal data set and try to utilize AI To figure out the early signs of functional decline So we've been working almost like two years, you know, right for now So we we try to take a lot of challenges and some of them I actually listed here So we want to see what is the objective way we can actually assess the behavior and thoughts of these older people And also what is the early signs of you know functional decline and what are the risk factors and what are the protective in the factors And also what's the innovative strategy for prevention and intervention and then also, of course, you know Data is a big big challenge, right? So, you know data availability and ethics and success and so on So these are the challenges we try to tackle. I Want to introduce another report recently came out. So this one two months ago from our world Economic forum. So this report is about the ethical adoption of, you know, technology So already mobile internet and AI technology is widely used for the mental health, especially Self-care of, you know, mental health. So these technology Basically provide any time, you know, anywhere and any way kind of You know way right to provide a service and also it is a scalable and also very, you know, low cost But still early stage, but certainly you know, rapidly Technologies are rapidly adapting to this domain and These are the couple of startups out there. So you might know, you know, some of these companies already So Mindstrong and Seven Cubs and, you know, Inuka and then Spring Health. So there are lots of, you know, technology based mental health self-care You know businesses going on and people have a lot of ideas, new ideas, right? Try to adapt, you know, this domain So I think that's why this is a very timely Right, you know to have this kind of open discussion We're opening the discussion about the AI for mental health But as you can imagine, you know, this is not a small challenge This is we are talking about really really, you know, big challenge, right? So we need to consider many many things and these are some of You know, things that I can think of. So, you know, in terms of setting, I mean, do we want to Apply AI for clinical setting or as I mentioned earlier, you know, self-care setting So what's the best practice, right? And then what's the barrier preventing effective care? Who are the target, right? Are we targeting on the young people or, you know, all the others? And is there any way we can unify the metrics, you know, for assessment and of course the data, you know This is a big big challenge, especially for the, you know, data seed over here. So you guys Especially, you know, mental health related, you know, data set, you know, there are lots of stigma issue and privacy issue So that's a big, big, you know, challenge And again, right, what can be right technology and tools in terms of intervention and, you know, prevention and also Interesting topic is like, you know, how we can train mental health, you know, workers, right? So I don't think we can touch all this topic today We'll very likely focus on, you know, clinical setting and maybe the, you know, all the other things and so on But hopefully, you know, we can think about these, all these topics and, you know, we can certainly have further discussions, okay? So this is today's agenda. So we have a very nice setting of, you know, invited speakers and so on But the talk is about 20 minutes So I think, you know, 15-minute talk, then maybe we can take one or two questions for each speakers And then we'll have, you know, open discussion, Q&A session, you know, 10 to 10, 30 minutes and then at the end of the session And later, today we'll have a panel discussion on the data ethics So we'll have our, you know, speakers coming out and then discuss with you guys What is the challenges for dealing with the special data related to mental health? So because we have like 20 minutes, so we're not going to introduce all, you know, speakers, their background in all the details But these are a list of our speakers So quickly, Dr. Brazil from MIT Dr. Deb from UC San Diego Dr. Ferante from NIMH And Dr. Zest from UC San Diego And Dr. Zest from UC San Diego And myself from IBM And Dr. Lee from UC San Diego And Dr. Nebeka from UC San Diego And Dr. Park from MIT And Dr. Torres from, you know, Harvard So they are the invited speakers taking lead discussion today Okay, thank you very much And with that, I'd like to introduce Dr. Dilip Zest from UC San Diego Hi, good morning And thank you, Hochal Hochal and I want to welcome you all to this session It's really an exciting session And I especially want to thank Michelle Ferrente, Cynthia Brazil John Torres who will be here And he won Park for coming here from Outside UC, SV and IBM So I'm going to talk about AI for mental illnesses and mental health With the focus on aging So this is our San Diego team And you'll see number of the people here from our group In this picture The Center for Healthy Aging at UC San Diego Is really unique in several ways It brings together school of medicine School of pharmacy, school of engineering And school of management These are the four schools UCSD has Along with the general campus So this includes social sciences, biological sciences Arts and humanities Technology industry as well as senior housing Welcome to John Torres So it's really nice that we bring together the whole group Which is essential for studying aging Aging doesn't belong to any one specialty So there are two parts to my talk Initially I'm going to talk about mental illnesses And the opportunities for AI Second half I will talk about mental health And opportunities for AI And talk about how these are not exactly the same Mental illnesses and mental health And I will give you an example of the approach to dementia In the first part and loneliness and wisdom In the second one So in terms of mental illnesses The DSM-5 Is called the Bible of psychiatry That's actually it's a big fat book That includes all the psychiatric disorders there are There are 22 different categories of illnesses Which includes more than a couple hundred illnesses So this is published by the American Psychiatric Association It was published in 2013 And I actually was the president of the APA At that time so I was actively involved In designing this And one thing we realized Actually that one of the shortcomings of DSM-5 Is that the criteria are mostly subjective They are based on what the clinician diagnoses We don't have in psychiatry at this time For most illnesses any biomarkers Or any technology based assessment That's the great opportunity for AI to make an impact So the diseases there are 22 different categories Including psychosis, depression, anxiety, PTSD Sleep disorders and sexual function disorders And so on what I view The one I am going to focus on a little later Is dementia That's what is applicable to older people So there is both need and opportunity For AI research in this field Why? 20% of the people in the US As well as all over the world Have a mental illness So one in five So that actually counts for a large population And there are high healthcare costs Associated with conditions like dementia And depression And yet there is a severe shortage Of psychiatric clinician Severe shortage of psychiatry, psychologists Nurses, social workers Who specialize in mental health So the future of medicine needs technology The current system of individual clinicians Taking care of individual patients Is not sustainable Because the gap between the supply and demand Is huge and it is growing At the same time there are issues In applying AI to mental health field Unlike most of the other medicine In psychiatry, the clinical practitioners Are hands on and patient centered Relying on soft skills Skills like forming relationship with patients Observing patient behaviors and emotions So these are hard to measure Using technology And the data are subjective So the data include qualitative statements By patients who say that I have these symptoms But not others And written notes by physicians, nurses And other clinicians So it is mostly subjective Unlike other branches of medicine Such as surgery or imaging Where the skills that are taught Are hard skills When it comes to aging It becomes even more complicated We all know that the population is aging The average lifespan in the US In 1900 was 45 years 45 years Today it is 80 In 2050 it will be 90 So from 1900 to 2050 The average lifespan would have doubled From 45 to 90 This has never happened in the history of the world In the history of humanity So there is something radically changing And that is increasing the number of older people And also aging is heterogeneous We think that as people get older They become more similar Not true at all As people get older They become more different from one another So you can be a highly functional older person Or you can be somewhat disabled Or you can be totally disabled in a wheelchair Also within the same body Different organs age at different rates So my liver may be 50 years old Kidney may be 20 And spleen may be 100 years old Multi-morbidity There are multiple body systems That could be affected Something called stochastic events Unexpected events like stroke Heart attack All which suddenly change the equation rapidly And high rates of dropouts in studies For the various reasons And importantly there is a dearth of data On technology and older adults Especially the old people People over age 55 That is the fastest growing segment of the population The number of people over 85 Are going to quadruple in just a couple of decades And yet there are practically no data Most of the technology companies Don't study people over 85 They are harder to study And there are other issues Such as they are not being used to the technology So they are somewhat resistant to use of that So for all these reasons It is both a challenge and opportunity for AI So the potential for transforming Neurosacatic research and practice So some of the data are quantitative Such as EMR data Or imaging, genetics So there one can use machine learning But most of the data are qualitative Subjective statements, clinical notes And there NLP would be appropriate So there is as Hochal said There is a lot of potential for AI To develop behavior based digital phenotypes This really is something badly needed We are hoping that the DSM6 The 6th version when it comes out Would be much more based on such digital phenotypes Identify biomarker based subtypes Redefine Neurosacatic diagnosis Facilitated early detection of disease Enable better monitoring, personalized treatment Pre-treatment response And offer scalable interventions The AI can do that The current healthcare system cannot So it really has a potential For transforming what we do in this field today So I am going to focus on dementia There are 50 million people in the world today With dementia In the US there are 15 million Rest of the world there are another 35 million That number will rise to 130 million by 2050 Think about the 130 million That is more than the populations Of number of countries combined And yet there are no ways to prevent Or cure dementia So at this stage our main focus has to be On very early detection And yet it is difficult Dementia comes in different sizes and shapes There are different types of dementia There is Alzheimer's dementia Fronto-temporal dementia Another something called Louis-body dementia You all know Robin Williams Robin Williams actually had Louis-body dementia And there are a number of famous people Who have had Alzheimer's disease So what do we do? IBM UCSD center at UC San Diego Our focus, so we So given the challenges that I described What we think is needed Is a team that includes experts And trainees from various areas Working together So we have experts in mental health Experts in geriatric healthcare Experts in technology, ethics Health assessments And we follow a community partnership Approach These are the studies that we do In one particular continuing care Senior housing community I think some of you are familiar with these communities So they are called continuing care Community in the sense They have independent living So you typically enter when you are Functioning okay Then as you become disabled You move into assisted living sector Then as you get more cognitive impaired Move into memory care And nursing home The assessments are comprehensive So they include physical health Cognitive function Psychosocial functioning Variable sensors Audio tapes of qualitative interviews Video tapes of structured activities And Colin Depp will talk about this Get up and go is one of the tests That is used And these are the sensors Sleep EMR Electronic medical records As well as biomarkers Including microbiome And blood based biomarkers So it is really critical to have All of those data And that is where again AI is critical So the long term goal Is to identify the earliest signs of decline So we can do something By the time the patients develop symptoms It is too late to do anything With the pathology that advance a lot So that is about mental illnesses So going to mental health now So 20% people have mental illnesses As I said But 100% people have mental health Alright I mean all of us Have physical health Cognitive health And mental health So what is mental health? So mental health includes The states like happiness Well-being On the other hand It could be sadness and anxiety There are traits Such as resilience, optimism Social engagement That affect not just mental health But also physical health And cognitive functioning And why is this different From illnesses There are no criteria There are no books like DSM-5 That describe their definitions And how you assess them Why should we bother about these traits? There is a lot of literature Showing the importance of this For example there is a meta-analysis Of 83 studies of optimism And the conclusion Was that optimism is associated With better cardiovascular outcomes Better physiologic or markers Like immune function Better cancer outcome And lower mortality All of these findings Significant at PLS and 0.0001 Whatever Calculate So very highly significant Social engagement Even stronger database Meta-analysis of 148 studies From across the world With more than 300,000 people These are studies including men, women People of all ages From different countries Different races, ethnicities, diseases Look at this 50% increased likelihood of survival During the study period Among socially engaged people Compared to non-socially engaged people The effect sizes for optimism, social engagement And resilience For increasing longevity Are equal to or greater than Those with stopping smoking Doing exercise Taking statins I just want to show you One study of optimism This was a Dutch study Of about 500 men Over the age of 65 Who did not have cancer Hard disease to start with So at baseline They divided people into three groups Those who were most optimistic Those who were most pessimistic And people in between Based on the ratings scale And then they followed these people Over a 15 year period And looked at how many people Died from hard disease They found that The optimism lived 8 years Longer than the pessimistic So there's significant difference Between optimism And pessimism in terms of longevity And this is after controlling For past history Family history Exercise Smoking Use of statins So optimism seems to contribute Something above and over All these factors We did a study A few years ago of People All the adults In the sense From age 21 to 100 The entire adult lifespan About 1500 people Who were randomly selected So this is a randomly selected Community based sample Over the entire lifespan And then we looked at Both physical health and mental health The physical health As you would expect Declines with age So in the 20s and 30s People are at the top of their physical health Fountain of health And then it starts declining But this fountain of health Doesn't apply to mental wellbeing Because mental wellbeing Goes exactly in the opposite direction 20s and 30s People have the worst level Of stress Anxiety And depression But the good news for people in 20s and 30s Is that Things will get better Progressively And as you get older This mental And this is now This finding has been replicated By multiple studies Including one that was just Published last month In Journal of American Abnormal Psychology 600,000 Americans They studied From 18 to 70s And they found Exactly the same thing That the anxiety stress Were highest in the 18, 20 And then Progressively declined So there is something that happens With aging that's actually positive We usually think about aging As they're all bare Not true There are things that actually Get better with aging So now switching gears slightly In terms of the trends I want to start with something negative That is loneliness Loneliness is a major public Important area It's also a business area The IBM Institute for business value Published a report Two years ago On the impact of loneliness On businesses Come back to that Next slide Loneliness And social isolation They're related Loneliness is subjective I feel lonely For example Whereas social isolation Is objective How many friends I have So that is social isolation So loneliness And social isolation Are called silent killers They have been shown to be As dangerous to health Smoking and obesity In the US This is by the way From the US agency Of health care research And quality 162,000 people per year Die from social isolation That's greater than the number of people Who die from lung cancer or stroke In the UK A new ministry of loneliness Was established last year The main reason was business They found that The country was losing billions of dollars Because of loneliness of the workers That loneliness was associated With a bunch of diseases That reduce their productivity And that's why they decided To have this ministry of loneliness So we did a study Ellen Lee was the first author Of this paper This was a collaboration Between IBM and UCSD That researched actually Wide publicity Including CNN and BBC And what have you So we found that loneliness Was quite common Even in San Diego Which is supposed to be You know, paradise And it was common Across the age groups But it was most common In three age periods Late 20s Mid 50s And late 80s Why should we care about loneliness Because it has health consequences Loneliness is associated With increased risk of cardiovascular disease Metabolic diseases like Diabetes, obesity, depression So why is loneliness associated With these illnesses Several reasons One of which is biological Very nice genetic study Of loneliness that was done in UK Including about half a million people They found that the genes They found that loneliness Is modestly irritable About 50% And that's true for most traits Like resilience, optimism There are about 50% inherited Which means 50% They're affected by the environment And behavior So we can modify them The genes responsible Genes associated with loneliness Often were the ones That were also associated With cardiovascular disease Metabolic disease Depression Triglycerides and HDL So there's a biological reason Why loneliness may be associated With illnesses And there's of course psychological reason If you're lonely You feel depressed You're not likely to go out So you're not likely to have physical activity You will be sedentary Not eat a healthy diet And all of those things combine The good news The best news actually From this study of loneliness Is that loneliness Is inversely correlated with wisdom We'll talk about what wisdom is In the next couple of slides But this was a highly significant finding The correlation was minus 0.51 In this M-Turk study This is the study of 3,400 people And we have replicated that finding In two other studies Including one from Italy So this is I think a real finding With a high effect size So what is wisdom? So wisdom is another personality trait Like resilience, optimism, loneliness But it has several components in it They include self-reflection Ability to look inwards And try to understand ourselves Emotional regulation Control over the emotions But with some happiness Contentedness, positivity Pro-social behaviors Like empathy, compassion, altruism Think that we do for other people Rather than for ourselves Decisiveness amid uncertainty So a wise person accepts the fact That there are different perspectives And he or she is not sure exactly What his or her perspective is the right one There may be other one So you accept uncertainty At the same time You can't be uncertain all the time You don't get anything done So you have So it's a balance between Decisiveness and uncertainty And lastly, spirituality So based on that We actually have developed a scale For measuring wisdom And wisdom, like other traits Is biologically based We have published paper on The neurobiological basis of wisdom The neurocircuitry of wisdom If you will Which includes prefrontal cortex Dorsolateral, ventromedial and antisingural And limbic striatum Especially amygdala So this is my next to last slide So these are important things Resilience, optimism, loneliness, wisdom It's a great opportunity for AI Because these are subjective Harder to measure right now Objectively So we are doing some of these early studies For example, Alan Lee is doing The literature review Following the systematic method Of literature review And Kaoru and Yasu from Japan Are looking at this From the NLP perspective So we will see how the two Methods compare with each other Self-rated scale So these are scales for all of this And so that's where Machine learning can help Qualitative interviews Which are audio types So that's where NLP can help And then to find out predictors And intervention The long term goal is With sensors, biomarkers and AI We hope to develop behavior And biomarker based digital phenotypes For loneliness and wisdom This is really critical And these are just loneliness, wisdom Or examples Same thing would apply to resilience Optimism, social engagement These things have significant impact On health and we should be able to Measure them objectively That's a challenge It will happen But again, that's something I appreciate any suggestions You may have And this is my last slide So the pillars for health And longevity are not statins And other drugs that we talk about Healthy lifestyle Social engagement And wisdom Actually if people had all these things The Incidence and prevalence of heart disease Strokes and some cancer Will plummet And the healthcare cause will be saved So thank you for your attention