[00:00:05]
>> So I guess sort of thank you for coming thank you so much thank you it's a pleasure to be here so as a start there will be a lot of clinical stuff so it might be somewhat boring to the group here and there are none clinical yet I would like you to ask questions because I think that will get you more intrigued to work with us to help us also in taking care of these patients.

[00:00:27]
I have no disclosures and my overview here is that we're going to talk about a little bit of clinical medicine and data in the c i.c.u. c.n.n. which I think probably a lot of you guys are really familiar was in our experience with c n n and c.s.u. data.

[00:00:43]
This is an example of patient that we would be taken care off so a lot of things that are talked to this kid there are a lot of action monitors that you don't see this is a and you need to is actually after heart surgery open heart surgery have an open chest with a patch that is only covering part of his chest and he's hooked through many monitors many tubes and drains so a lot to process actually and that's not showing even the data that we will see so that is a segue to tell you what we deal with in the pediatric cardiac i.c.u. mostly the patients that I deal with are kids with congenital heart disease it accounts for about one percent of all newborns have some type of hard defects about $40000.00 births per year will have heart disease congenital heart disease remains the 1st cause for mortality and death in the 1st year of life we called here and in pediatric i.c.u. that it's a problem with plumbing it's a lot more than that actually the timing of development of these heart issues could start really early on it starts from the 1st almost 3rd week of just Station so it's a lot harder to detect.

[00:01:52]
Mostly And we detect them usually the weeks of 20 weeks of just station at that time and it's hard to do anything by then yet there is some developments that we're trying to work on how to prevent the worsening these congenital heart disease issues in utero so that's part of what we do as well the normal heart so much a little bit how the normal heart pops you'll see what we always say blue blood red blood the blue blood when you consider is the one that is deoxygenated it doesn't have enough oxygen so the blood usually returns back from the whole body that has been utilized through the big vessels and tried to see if they can actually mark from these big vessels and returns to the 1st chamber that is called the right to treat him and the right atrium pumps it to the r.v. which is a pumping chamber to go to the lungs this is through a big vessel that called only artery in the long it becomes oxygenated and returns back to the left atrium where it pumps it out to the whole bodies through the order through the left ventricle which is to very important pumping chamber called the left ventricle.

[00:03:03]
Now there are different types of congenital heart disease there are holes in the heart that a lot of people might have even in this room in a population of about 15 percent of us might have a benign hole in the heart between the upper 2 chambers I might even happen because I didn't check my heart it obstructed ordinary the valves and blood vessels those are going to be symptomatic of normal connections that could happen a normal number of pumping chambers and a failure of this pump to work or a rhythm ears which abnormal heart rhythm when these babies are born especially if they have something called not a car disease they could look pink and look fine initially or their blue as you see in the baby on the left here he's blue he doesn't look normal so that could be said on the card baby.

[00:03:51]
How did we start to recognize these infants or new units actually some of them they could be discharged home without knowing what happened and then they come back to us in a state of what we call shock and the way they might present they look blue they might have a heart murmur which is a noise that we can detect with a stethoscope they might have issues with breathing they might be able not able to feed Well they have increased work of breathing their fussy their color is poor and they don't have good pulses and what we call profusion is really poor How do we diagnose them do physical exam and then we can do it echocardiogram which is an ultrasound of the heart and we do lab test and among other things.

[00:04:33]
The pediatric heart surgery actually started in 1950 s. So it's not too long ago quite honestly initially the survival was really the small at that time and over time with the research and improvement in quality techniques this survival rate had improved tremendously in the hospital where I work at Children's Healthcare of Atlanta we are actually one of the largest cardiac programs in the country and we do about $900.00 cardiac surgical procedures per year $15000.00 cardio catheterizations but a year and that's only to Kate it to the pediatric Cardia population of Georgia among sometimes some other areas in the state we do have a dedicated pediatric cardigan sense of care as well.

[00:05:17]
Overall the survival now had increased to 97 percent among all the survivals of the cardiac heart surgeries with amazingly low morbidity and mortality I'm not sure if you guys understand the word morbidity so morbidity it means like having other complications and having other diseases along the side would original cardiac disease because a sicker kid with a sicker heart will have affected other organs as well too a simple b.s.d. which is a hole in the between the lower chambers of the heart could last about 22 days in the CAC you in the cardiac i.c.u. and maybe one to 2 days post that in the step down or the floor and you can go home within the 1st week and just remember actually doing the Vs dear repair it means you have to stop the heart with the heart and heart lung machine we call arrest the heart for 20 to 40 minutes and then restart the heart after actually regaining the circulation so it's not a trivial repair but compared to how it used to be even 10 to 10 years ago now the story is no more than probably $4.00 to $7.00 days in whole and that is actually not a trivial surgery.

[00:06:29]
The history of the 1st surgical treatment of a heart defect was something called Pedialyte geisha and the p.t.a. is actually a small vessel that connects between the great vessels the artery in the order and all of us are actually born with it or at least 97 percent of all of us are born with it it should normally close spontaneously within the 1st few hours to a few weeks of life the kids who have at present they might have heart failure symptoms and they needed to be closed or like you did the 1st surgery was in 1989 but that was without having to stop the heart because everything was actually outside the heart itself the next treatment was for a blue baby and that used to be for it to follow patient that was in 1944 in Hopkins the 1st heart lung machine used was 9 $153.00 and that's when truly the heart surgery the pediatric heart surgery started to take off and this is some of the 1st attempts on using the heart lung machine or bypass because the idea of doing that is you have to empty the heart of the blood so you can work on the heart you have to stop the heart but then how is the rest of the body's going to survive you have to take over by having the machines circulate the blood outside the heart and the Longs and then go from there that was one of the 1st generations of the heart lung machine and this is a new compact truly amazing machine to use to bypass the heart and the long.

[00:08:03]
Now I'm going to take you to one of the most common actually congenital heart disease defects that is very unique and those patients. Compromise a lot of the kids stick used to maybe 20 years ago or actually start years ago die and when we counseled these moms or their families we said they might not live more than a few days of life or they might not live more than a month of life now they have a higher survival rate compared to what it used to be this is called hyper plastic left heart syndrome as you can see compared to the previous normal heart diagram the left ventricle is really diminutive and very small and you can see that the order is looking really small and tiny This is the connection that we called the p.d. a that we discussed before in the 1st heart surgery done before and that is a key instance here in this patient because if you didn't have this p.t.a. he would not live either in utero or he would not live in the 1st few hours of life because having a vessel that is really tiny and not able to pump out of the left ventricle makes you dependent on this p.t.a. to maintain blood flow between the heart the punny arteries that goes to the long and to the whole body that goes down and you have to have what we call Anstey which is an open like an atrial septal defect and Oho old between the 1st opportunity Bers to be able to mix the blood that comes back from the long to the left atrium and able to mix and then go down here to the r.v. that is now acting for boards 2 chambers instead of having one chamber or 2 chambers You have to act with one chamber to pump to both the heart to actually the heart 1st and then to pump to the lung and the rest of the body.

[00:09:56]
I have a plastic left heart syndrome requires multiple stages of palliation and these stages of palliation start with a 1st stage called a Norwood from all the congenital heart disease the hypo plus the counts for 1.4 to 3.8 percent of congenital heart disease it requires a lot of resources including invasive until ation mechanical circular support and when I say resources also have to think about cost as well too so they're heavy cost utilization significant morbidity those kids are at risk of having kidney problems in their logical complications and infections and the mortality is quite high and among all the congenital heart procedures it ranges from 7 to 19 percent varies according to different papers.

[00:10:44]
Studies but it is quite high compared to all the congenital heart disease and diet what makes these babies with a plastic left heart syndrome very fragile yet they're very also unique so we need to utilize a lot of what we do to help them survive through this and decrease this more tell t. as well as the morbidity actually so surgical targets as we have discussed earlier you have to have.

[00:11:10]
An obstructed systemic blood flow to make sure where you have these narrowings because the vessel is really narrow you can have enough that can supply the rest of the body you have also to have an unrestrictive blood flow from the Palm reveres that came back was also oxygenated blood to come back and mix in the right to go down to the right ventricle to be pumped out and then you have also to have a control of how much blood goes to the lungs and now the blood that goes to the lungs is going to be tremendous amount of blood that goes along what could happen with that you would like engineer in layman words you could flood the long because a lot of blood True Blood of flow will go to the long and less will go to the body and the kid will start to breed too fast and then they're not going to have enough blood flow that close to the rest of them so it's a lot of plumbing physics and actually balancing the circulation not always procedure could be different types you have something called modified bt shunt or plot of contests that the sick show and this is actually the names of the 1st surgeons started this procedure and r.v. to puke kind of it which is a shunt that comes from the right ventricle to the primary artery which called this and according to the based on the surgeon that invented it it's actually has a life.

[00:12:32]
So there is only one chance actually to get it right to do this surgery you have to do it perfectly from the 1st time as are worries you'll have a lot of complications you have to preserve the ventricle or function and you have to maintain the peace in c. of the valve that is between the 2 upper and lower chamber and you have to encourage the growth of the pony arteries and also allow deploy vessels pressures to go down so you can have enough blood flow but not too much blood flow and hope also for growth for the baby over the time after the surgery.

[00:13:07]
The 1st surgery of the 1st Norwood surgery happened in 1903 Dr Norwood his actual life and his counselor in Albuquerque and he did the 1st surgery at this stage and this kid a baby actually the 1st case survived and Hughes was discharged home and both. 18 so that was an amazing.

[00:13:32]
While stone in pediatric heart surgery quite honestly we have something called s.p.i. trial this is a trial that we have looked to see a single ventricle lesions and what are the best actually surgical options for them comparing the bt shunted r.v. to peer kind do it and I know this is too much for you guys because this is actually not a clinical science but it will help you in using these data eventually and I'm hoping that you'll start to find these intriguing so you can actually use them to help in decision making as we go along so to compare these kids in this s.v.r. trial we saw the most just happened with bt shunt initially and then they kind of became the same between the 1st stage and the 2nd stage but I had the end of this we discovered the r.v. to be a conduit patient did not have like superiority from being transplant free compared to bt shunt because some of these kids will eventually need heart transplant and this is how we counseled these babies and their I'm sorry their parents'.

[00:14:38]
R.v. to be a kind of a child had lower rate of c.p.r. and having cardiac arrest and bt shunt patients required to have more. Less Palmeri artery interventions was r.v. to be a kind of a treated more interventions over time starting from after their surgery until even after the 2nd surgery who decides on which went to choose quite honestly it's about based on the center and the surgeon preference but if you can convince the surgeon or the center that we have a better option by using one versus the other that might make a difference to the group because then you're going to try to decide to use that better.

[00:15:15]
Some of the risk factors for the mortality on these kids depends on their Actually heart disease to start with Do they have much else to know says Erickson they're high risk their aorta is really small and it's less than like 2 millimeters to $2.00 millimeters if they have restrictive atrial septum if they have actually obstructed venous return if they have other congenital anomalies if they're premature and if their weight is less than 2.5 that makes them at high risk and if you did the surgery after 14 days of life they're allowed to take in account from preoperative acidosis how much on the troops that are needed medications to support them if they have a dysfunction if they have a valve that is not working all of these plays into their morbidity and mortality afterwards.

[00:16:03]
Skip that and then into or you have to think about how much time they're actually on the heart lung machine and how low we have to push their temperature down and if they have any lesions that are not supposed to be there anymore and if their arch is not working well and if they have bleeding if the rhythm is normal so you have to take all of this in account when you take care of them post-operatively So this is an example of a baby who comes immediately out of or you see some of the things that are attached to him so you have mirrors which is a monitor to assess perfusion or circulation that is a.

[00:16:40]
Monitor probes that are put on the forehead and also on the back you have a ventilator on the side with in total which assess the c o 2 release you have an open chest at times you have a pulse oximetry you have a piece in wires these are wires that attach to the heart to us to help with pacing the heart if it's not facing a normal rhythm you have a feeding tube that is not for feeding but for decompression you have lines that are inside the heart to assess the pressures of the heart you have a continuous blood pressure monitor through another line in the small artery you have a fully to.

[00:17:14]
Continuously monitor their urine output you have a temperature core monitor that is continuously wanting their temperature something called totem as well too and you have dreams that are in the plural space which is the space between the lungs and the lining of the Long as well as in the abdomen and some are in the chest as well too and then you have this monitor that you have to look ahead and make sure that you analyze it continuously and always you have a heart rate which is the one in the green $128.00 you have to look also at the waveform of this heart monitor you have to look at the blood pressure which is the one in the red you see you have to analyze the blood pressure and numerical numbers so 74 over 46 that's an important number but also I have to look at the waveform and see is it narrow is it dampened is it variable with each breast all of these give me information as a clinician and then you have something called t v p which is central venous pressure so I have to look at that numbers only per se but also have to look at the waveform to analyze it and make sure is this waveform telling me that there he might have been.

[00:18:22]
A valve leak or does he have a normal rhythm and it's waveform is abnormal and then you have an atrial pressure as a continuous monitor again I have to analyze this as well too you have oxygen level and you have a story number with a plus for that I have to look at and you have something called temperature skin skin temperature I'm sorry and in other places you will have also a quarter as well 2 and a totem So all of these you have to analyze continuously all the time.

[00:18:53]
Now how do you monitor them as I showed you these screen and then I have to examine the patient frequently have to look also divide medical data that includes blood gases like sick acid him a globe in electrolyte car dilution profile makes Venus which is the oxygen that comes back after the blood every organ has used it and then imaging will do x. rays ultrasound echo cardiogram an e.k.g..

[00:19:20]
I have to look at all the medications are back on most of the time this is the center medication they're back on their infusions it's so Miller known which is one of the visa active medication that helps with relaxing the heart as well as squeezing the heart of an effort it helps with the squeeze of the heart the opening also helps with the squeeze of the heart and increasing the blood pressure fentanyl which is a sedation medication and then this is their addict medication are they going to come back all on the same medication No you'll have a variation of different drips coming back less standard other medications some of them are sedition and some of them are visa violation medications most of them will have something called Low cardiac output and this is where the money comes how did that all cardiac output before it happens and how to react to it in a timely fashion because of the low credit output happened and that we fail to recognize it early the kid will have a of cardiac arrest and their heart will stop so one of those times will have an adequate blood pressure rising feeling pressures and lactate increasing mix to be in a set is less a poor profusion or long compliance that is trending down lower heart treat heart stops but I'm told him 3rd is going down.

[00:20:41]
And how we do work it up I was in the vital signs that we looked out and all the numbers will look at the echocardiogram maybe we'll take them to the cast lab but we'll get other imaging and will discuss with the surgeon what are the next steps that we should do you will manipulate some medications and sometimes you have to take them back to the or and sometimes you have to put discarded on a heart lung machine to rest this hard because it's failing before it stops.

[00:21:07]
Oxygen level could come down and become really adequate and when you have an adequate oxygen level the lactate will come up and increase because in your tissues are having low oxygen levels and that fact it will increase and produce that might lead actually to abnormal rhythms as well too so you can use medications you might have bleeding issues as well too and you have to use medication as well through all of this happens in the 1st almost $2.00 to $3.00 days immediately after surgery.

[00:21:39]
At times you come back with an open chest as you guys saw the picture but you might have to actually work on closing the chest and doing any surgery for the baby in this critical time is really crucial because they might actually do compensate during that time yet you cannot leave the chest open all the time now the next step while taking the breathing tube out and then starting to feed them with possible complications for all of it failing the breathing tube coming out because they might have a vocal cord that is paralyzed which is one possible risk but the surgery 12 percent could happen a death from that is paralyzed because also you we could injure the Sir Nick nerve that supplies to there from that could happen 5 percent of the times you might have actually intestinal infection and low blood flow to the intestine a 10 percent of the time and we might have something called Kyle authorities which is the lymph lymphatic will drain through the plural spaces but 9 percent of the time all these things could happen including renal dysfunction or kidney dysfunction they might have to have continuous renal replacement therapy which is there also is another world and 6 percent of the times they might have stroke and 6 percent also of the groups of hyper bust and seizure and 6 percent of them infection up to 10 percent so these Their funding for their life actually if you think about it immediately after surgery the survival in this paper from 2002 to 2009 if you look at this institution actually from the S.T.'s which is like looking at all the numbers of stage one palliation which is the Norwood that happened between 20022009 you'll notice that we have we started with only $300.00 cases per year so that's not a lot actually but the desperately discharge was 90 percent 95 percent so only $68.00 of them had survived that is really poor while by 2009 we had a higher number 2320 cases performed in why.

[00:23:38]
That year 432 died which is better than it used to be yet it's still almost 20 percent mortality rate as I mentioned earlier was 7 to 19 percent mortality risk so there's still we're not doing great and most of this is actually in hospital mortality and in hospital actually complications so we're trying to avoid that.

[00:24:04]
They're almost trying to reach the home this is how they're doing most of the time and some kids will stay in the hospital and 3rd they get to the next surgery at Asia almost 4 months of life for each discharge post nor would they still have risk of mortality as well too up to 4 to 15 percent they need monitoring they need close monitoring they need medications and they need continuous therapies as well to.

[00:24:32]
Out of the hospital so this is after post discharge from Norwood So this group is really fragile and this is a group that has been studied extensively from clinical standpoint and we've been trying to do a lot of measures that including Best Practices having core groups that only deal with this type of plastic of heart syndrome or single ventricle population adjusting their shuns their Barents even doing stem cells for them and having devices that continuously measure their oxygen levels and their.

[00:25:05]
Oxygen levels now the next big thing I think where I find Hugo's going to be very helpful and it hopefully is organizing interpret in these data as you can imagine this is a lot of data for a clinician to stop there and look at the quality of these decisions that they make every day and actually sometimes every few minutes for one patient.

[00:25:28]
You're not always doing great with it we hope that we're doing the best and we're actually at the positive. Side but a lot of times we are not actually. The patient if you think of her them between the health and illness side you would hope as they get admitted they will eventually get discharged home you'll have many pumps on the way but you would hope that eventually we'll get them home some of them there will exceed our expectations and they will go home very quickly without all these bumps but unfortunately up to 20 percent could go to death and we're trying to avoid that and actually not only thinking about death we're trying to think about their morbidity is how to avoid complications while in the hospital so to make all the medical decisions during taking care of these kids you have to interpret all of these there that I showed you including study results and had to with your experience as well the knowledge you have to stay up to date and what's coming up from literature and you have to keep up with the experiences of the providers around you so this is a patient that you look at and you see this is the heart rate similar different than the one that we looked at earlier this is a piece tartrate So what we're doing is using the piecing wires that actually were attached to the lower and upper chamber of the heart to fire at a set rate of $130.00 so he's basing both.

[00:27:00]
Upper and lower chambers at a rate of 430 his blood pressure is 79 over 38 I'm going to be satisfied for a new unit with blood pressure yet what I look here and see is concerning is his feeling pressure $1021.00 this is the central way this pressure and the eternal pressure these are high and plus the way form appears normal for this baby which leads me to having to investigate this baby more but I have to notice this change if it happens if it happens like between 1 pm and 2 pm having the baby to sit with the same findings for an hour without me noticing that could be determined so for him it might not be immediately yet if we can detect these changes quickly that will be a key while this baby suddenly changed his heart rhythm if you see that that looks different even for non-clinical person he will see these waveforms looks abnormal is different than it used to be now he's not pierced but this something happened I'm going to rely on the monitor to detect some of these other maladies and alarm for me yet this is of in trickle arrhythmia which means it's in a bromo rhythm coming from the lower chamber of the heart if it persists it could be fatal and the kid could arrest the same kid went back to normal heart rate not needing to be paced blood pressures are reasonable and his phone pressures are down and his oxygen level is more appropriate for him so all these changes have been probably within a few minutes because you did some interventions if we were able to do it in a timely fashion.

[00:28:39]
So critical care data about $5150000.00 waves of each type per day of each patient close to a 1000000 individual waives per patient per day in addition to labs studies clinical diagnosis radiological genetics and historical and interventional data that we have to process looking at this kid this is a kid who is a heart lung machine or he is attached to many tubes many medications here if you see all of these palms are delivering some sort of medications This is New Year's This is the ventilator the mechanical.

[00:29:16]
Ventilation this is the echo circuit with the troops coming out of it and returning back to the baby and then you look and see this is the monitor that I have to see all the way forms that I told you about it becomes buried in the background of all of this.

[00:29:32]
And multiply that by $25.00 to $30.00 patients to have to do this continuously even if you have a one provider per patient it's going to be an overwhelming number of dead a point that you have to process and make a decision continuously on a lot so too much data most of these are ignored actually know way too little for sure lies all the data and understand or utilize them the subtleties of these where we forms that we simply don't see while it might make a huge difference in the care of these babies.

[00:30:05]
I'm going to share with you this something that as a clinician a clinician it hurts to mention yet it is its reality this is a paper from 2003 from a university hospital in Israel they looked into the nature and causes of human errors in the intensive care unit they actually surveyed observed for months the physicians and the nurses continuously and looked at all the reported ers by physicians and nurses they noticed that they had in the form and span of $554.00 errors that's a lot of human errors to happen in only 4 months in an average of $1.00 error per patient per day at an average of 2 severe errors per day per i.c.u. day that is a lot some of them could be fatal actually and they could to graze them into organ systems if you find mostly they're into the input and output which is like calculating how much urine output and how much intake there were yes it's not as crucial but it could make me decide to do something different with this patient if I find his Ensign outs are different than I expected while 5 percent of disease conflict errors happen cardiovascular 5 percent is not as much as 57 percent yet it could be actually the term mental for this baby or this intensive care could be the ones that are shaded are actually the physicians and these are the nursing partly I would say because they're having to do up to $175.00 actions per day for a patient so yes they're more going to be more prone to these ares than the physician that have to make some of these decisions but it's a lot of risk of errors.

[00:31:51]
So in the i.c.u. we started to try to identify the risks of the patients attempting to and to fight changes of his logic physiological changes or patient the compensation before it happens so we can actually have these studies before it becomes too obvious for them some of the most like known and well known and well used course something called Pugh's score which is a pediatric early warning score but this mostly is looking for and the need for urgent admission to the i.c.u. So this is going to be utilized in the actually in the regular floor looking at some of the 7 signs that a nurse will enter and say he qualifies for needing to go to the i.c.u. So Dad call requires us to have to assess the patient and decide if you need to go to the i.c.u. a reasonable sister beauty of 82 percent and specificity as well too but yet it's not going to help me in the pediatric i.c.u. itself some of the other ones that help with the cardiac arrest the problem with them they're still retrospective and a lot of it is actually based on nursing and administered scores so human error slow remains an issue and dirt aspect.

[00:33:05]
So we don't really have a lot of these scores that will help us in the pediatric cardiac i.c.u. specifically to help detect and predict worsening of these kids I'm sorry we have monitors that we use that pulls most of these data that you can look at trend so like this one we have something called Also bed master that we're using and I'm going to tell you about that one in our study that looks at how things are changing and in some sick.

[00:33:34]
Scale you can start to be concerned about their outcomes yet again it's not well validated in the pediatric cardiac i.c.u.. This is a summary of what's out there from that analytics into cardiac patients or in general an i.c.u. patients most of them are pediatric intensive care patients looking at the probabilities of adequate oxygen delivery or risk of infection or one of them is actually looking at the compensation and need for c.p.r. but the rest is not as high that helps me again the protector cardiac i.c.u. but none of them have looked at the Sports Group of Norwood's or the head of plastic left Hearts and.

[00:34:16]
Now we've finished the discussion about what's clinical medicine is there and the data in the c.r.c. And how much is overwhelming and the risk of ignoring some of these that even if you don't want to is actually reality c.n.n. which I think you can mostly all know about it is a deep to confront confusion of neural network is a deep learning algorithm that works well and image analysis the computer vision it's looks at the image evaluated categorize it and it had to fight it does well with large data sets and improves as the dataset increases in size so this seems very helpful for me because we have a lot of large that is that we have a lot of waveforms It's considered the waveform as an image to be analyzed and then it's extracted features from the image using filters or kernels and pulls it reducing the demonstrations of the dead of.

[00:35:08]
So I'm going to tell you about our experience with c n n the c.s.u. data show where along with the group here from Georgia Tech true and Dr Sun Hong as well myself and Dr Mar from ca see you we've worked are trying to see conventional neural networks for clinical assessment for infants following Garwood procedure.

[00:35:31]
And I want to highlight something here we cannot use it and generalize it for all pediatric patients because in pediatric you start from being a new an 8 so almost an adult size your vital size your heart treat your train and everything else isn't going to be unique to each age group you cannot generalize it to the whole group The 2nd thing is being a Norwood or have plastic left heart syndrome you truly do have a different also parameters and norms that's why we're using it only on the subset of patients that are really highly fragile so we looked at consecutive new units undergoing the Norwood procedure at our institution between 2016 until September 2019 we looked at what we've formed out of collected by the bed Master which is the.

[00:36:22]
It's a data pooling system that looks at all way forms as well as the. Head to clusters from Cho It was i.r.b. approved boards at Cho as well as Georgia Tech. In this we have conducted a binary classification for differentiating whether a 32nd e.k.g. clips were from a critical postoperative a patient or a healthy ready to transfer to the step down baby so this is looking at these patients concerning those are 4 patients you have you know what surgery you look at all the e.k.g. clips here for the 1st 2 days well they're critical and also we looked at it before just going after the step down here with infinitely also looked at the patients who were deceased or needed to transfer to other areas in the hospital we had at that point 47 children into training sites and then we used that 10 more on the 10 most recent children into the tests that we obtained about 2 153000 valid training e.k.g. clips and $74000.00 test clips per each of this really and we're using 3 e.k.g. leads at the moment this is the part that I think it's all cloning computer science and it's not my field yet my understanding is we have done a study for little convenient neural network was designed based on the rest of that architecture architecture and they have actually my understanding the have done their own training and they did not use a pre-trained model an algorithm to train the c.n.n. on each of the 8 waveforms lead wind 2 and 3 from the e.k.g. the pressure respiratory in place and the c v p or central venous pressure and oxygen level each of these wave forms is segmented to 30 seconds and it was discarded segment that we're having missed values this could in per each of these reach up to $2000000000.00 euro values $2.00 to $3000000000.00 actually values.

[00:38:35]
Day for the prelim to be looked at 4 groups separating lead $12.00 or 3 from e.k.g. or having it with heart rate variability the other one trained on vital sides with lab test and or both and. Group 4 had combined all of them including the lab test and the vital signs that are the comparison models for them.

[00:39:09]
Now this is where it's intriguing for a clinician to look at if I have this for continuously for a patient every day all the time that will be amazing because looking at this starting with him being critically ill having a very low score in the combined scores we started very low and as he's progressing in his post operative days from the 0 until the 6 or 9 or however he's going up and up and his score is improving so we have looked at to see which score is going to indicate wellness overall we're still yet to decide but between point $42.00 as the score we can detect that is putting the kids are closer to this to transfer to the floor or to stability here they look at all the combined per hour and also person hours now the next big thing is have to validate that as well too because as a clinician we're skeptical I'm I going to trust a machine to tell me that this patient is doing fine Probably not but there is also I want to show you this portion before we go to the clinicians skepticism is that an exchange of the patient and putting into the projection spaces so looking at these blue dots is a test patient.

[00:40:21]
That should be changing over time and the red dots weren't aggregation of all the patients were critical immediately postop the green ones were the ones that are ready to transfer to the floor so this isn't the 1st post-operative day immediately after dinner would looking at the blue diet it's still staying even below the red critical illness here it's started to kind of move a little bit above the red dots about 24 hours into it and we agreed and that clinically actually usually by the end of 24 to 48 hours this is when we start to see some stability and some improvement here at 6 hours in the post-operative day he started to actually reach to the stability and even specifically and post operative 3 this is when there.

[00:41:09]
Consistently staying above the green dots which tell me if we have this continuously happening that you can see the patient as they progress over time that is a grade and if you see a lag back toward the red you know something is happening because the kids are not going to follow one track and they're going to change over to a.

[00:41:31]
Post of day 4 which again goes with all our data that we know that the kids usually between us operative day 3 or 4 they're going to be stable and almost always ready to the floor some of them they needed to stay in the car tick i.c.u. for 30 days so this is a different as well to so we have to follow them throughout.

[00:41:51]
Now this is the part that I was telling you that we're skeptic we're going to want to know that this is going to align with a clinical evaluation because there is clinical. Understanding there is knowledge and there is experience that you want to implement into all of this I don't know how to explain it to you that we how we you actually process all this data and decide the species is Ok or not Ok It requires this many years of training for example I had to do 9 years post my medical school just to reach to the point that I take care of these patients so to put it out there is one plus one equals 2 is not going to ever happen yet I would appreciate the help from algorithms like this to tell me if it's improving or worsening so it direct my attention toward this kid so in here we compared between elders in prediction and the clinical evaluation these are 3 different patients.

[00:42:44]
So here in this patient what the blue line is how the progressed over the post-operative period and this is the prediction score as we mentioned earlier probably both like between point 5 to point 6 we discuss or point 4.5 mostly point 5 We thought that the patient will be above that are in wellness or predicted score being well so while the green line is the clinical scoring we did keep it a little bit broad by having them critically ill while the critically ill severely critically ill and actually stable or healthy and this is how the green line has been aligning if you see most of the time we kind of were in the same range as the algorithm was detecting where they are here in the 4.

[00:43:31]
As they're almost about to go to the floor we both agreed the group the prediction score as well as the clinical scored that they're ready to go to the floor but you can see the stepwise improvement in his care the other one that is very interesting is this one actually the 3rd patient if you see yes please wife.

[00:43:55]
So actually now these are the pink lions were some clinical changes that happened to the baby or the neonate so. To some change in the reason we. Wouldn't change so what we try to do is to have the score of the critic score by the algorithm that was done by the group here at Georgia Tech and then we had our own assessment clinical assessment as well as putting the timeline changes of the baby during the charter of you so I started he came back from the or on several medication and a breathing tube and then at one step here he came off one of the medications and then another step he we took the breeding to about trying to say.

[00:44:43]
Yes under clinical course to see if there is any change that was predicted by this predicting score if there was any changes as we're doing that and if we so far we couldn't tell 100 percent there is a huge change yet some of these you actually do see some changes if you can see these like going up and down and clinical score as well in the predictors course which in this 3rd patient and it was somewhat aligned with some of the clinical changes or the milestones as you mentioned for example this patient had some normal heart rhythm during that time and he was going backwards a little bit and needed to implement new medications as we're going forward so yes that actually followed the algorithm really nicely and told me I could have predicted that and if anything I think it predicted it but before we did as clinician because you see my improvement here was after it started to improve while the worsening was before I detected it was in worse inning so if anything it might have been an earlier warning sign that I could tell and some of these e.k.g. that will.

[00:45:51]
It was I couldn't tell but only visualization all the time what is the difference but I know the computer had detected something changed with these e.k.g. waveforms those was very interesting and I hope as we continue to work on it to finesse it so we can have it continuous actually and life to work with us this is for example one patient only one of the test patients the desperation that we looked at so looking at it I'm sorry this is a really busy slide but I just wanted to show you how we did it this is the combined predicting score using all the data here with lead one lead to lead 3 blood pressure s.p.o. 2 and labs we found that the lab was a little bit hard to use all the time because it wasn't always predictive because it's one point to time and you cannot have it continuous lab monitor so it doesn't tell you it might tell me about something that happened an hour ago quite honestly yet the e.k.g. was the most telling of all of them as they were going up and down in clinical scoring as well as an overall predictive score and here we're trying to take it without the lab and here with no labs and weighing the lead 3 because most of the time we found at least 3 was a lot of times had a lot of noise or didn't have good way forms so Deller didn't reach about 96 plus or minus 6.5 percent accuracy and when using single leave e.k.g. distinguishing the critically ill versus the able patients the final model that combined all input signals obtained a produce r.c. or c. of the $98.00 plus or minus 2 percent.

[00:47:34]
So yes. Or they're critically ill this is I think is that I think I know actually it is unique it hasn't been done yet specifically in this population and again as a clinician and find it very. Promising because then we can help more we can predict it quicker and be real on top of it before changes happen and I'll tell you a little bit more of ideas that hopefully one point some of you guys will work with us on that that our knowledge exist in the critical care waveforms apply cation of the sea and into the critical care data may enhance the patient monitoring and patient management the initial work presented here isn't going still extensive clinical correlation validation so could hopefully presented we're almost there for his function with the matter script with the group here our additional goals is to include the real time barometer actually for clinical wilderness to see them changing from the yellow the green and the red an opportunity for real time feedback for all medical decisions being made as I implement a medication or I change a ventilator I'm hoping to change to see the changes real time yet and when I say yes the patients are at risk of cardiac arrest and clinical to compensation to early intervention in their course likely to improve their outcomes they're more ready to use and hopefully their organ dysfunctions the Currently we don't have enough and we're not effective as we should be hopefully Husing the residential models will help us continuously monitor but again I know that it's not going to take away the importance of the clinical decision in the clinical management so we need to work together we have not going to be able to only depend on these algorithms that we also cannot only dependent on our own in this era of a lot of data that we cannot utilize and we cannot comprehend that truth innovation and collaboration.

[00:49:33]
Leads to advancing the science and treatment of congenital heart disease and this is our goal and I'm very passionate about it so there's a lot of ideas that we can work on and this amazing group actually is led by both Dr Maher and Dr Sun from Georgia Tech here and Dr Moore from what Emory young true as well as this wall and the Hong from Georgia Tech and our 82 you did amazing work for months actually and we still continue these Could this is one of the kids I'm just telling you so you can see what's the interest out of those is to have this baby looking almost normal Actually you wouldn't even tell that he was undergoing this open chest or anything like this and having those kids after having.

[00:50:21]
Year in the hospital this kid on the side here spent almost a year and a cardiac i.c.u. and had a lot of complications I can name them but there are too many to count and this kid spent months in the i.c.u. as well too and to see them outpatient or going home is the prize that all of us are looking for thank you if you have those any questions.

[00:50:48]
Well. I. At least all of. Us. This is. A critical time for the 1st. Operation So what are the warning signs. So I see it as a. Yes yes yes yes this is. The monitors themselves orders. Have program words yes. It is a very good question so each kid or each so in the high plastic left hard to grow babies specifically they have their own parameters that we're going to consider normal and abnormal so we set in and on the monitors to decide when to the alarm should drink if you may say so for oxygen level that is less than $75.00 it should alarm if it is above $85.00 it should alarm if it is a heart rate that is above let's say $160.00 or less than $128.00 times it depends each kid it's different and that's why I.

[00:52:06]
Said these alarms Yes. There are different ages for different groups and different actually surgeries that's why you have to set it for each kid differently at the time when they come back or they get admitted to the i.c.u. some of them becomes like easy and pattern for cognition because those kids you know they're norms are going to be between this and that and we have actually something called Critical rest prediction sheets that we have like norms that we're right at the bedside and what is the action plan in case something happened and when the nurse should alert the physician or the clinical provider to come bedside so we do have all of these things but it still does not prevent enough the cardiac arrest or the complications and the compensation they do have has been rough.

[00:53:01]
You mean the right side nurses. It is part of how the admission process so as a physician if I'm admitting this baby immediately after surgery I'm going to tell them what are the signs to look for and what are the like numbers that we're going to accept in the beginning these numbers could change based on the clinical scenario of the baby and how they're actually progressing over their course.

[00:53:26]
This thing that I'm saying like that could be subtle and night detected is the change in the way form is this curious becoming wider and wider This is the slight abnormal rhythm that eventually become a bad rhythm that is can be recognized quickly but it. In the heart rate.

[00:53:46]
Yes. Correct it's like over time a number as well to. Absolutely yeah so all of these if we can detected a bit earlier before it becomes obvious but I think if it's obvious yes all of us we're going to recognize that this thing is one to recognize it before it's obvious when it's subtle when the cure is becoming a little wider than it used to be when the heart rate trend became 20 points higher than it used to be all of these things if I have a set range between $120.16 or even $110.16 that's a wide range and I have to allow it because he might cry and increase his heart rate but he might be calm how to give this knowledge to a nurse or provider you cannot all translate all the knowledge you have it's one setting it's hard to do that you have to allow for changes.

[00:54:40]
Yes will correct absolutely. That it is a very difficult task and I think we're overall doing reasonably well with what we got but we can do better that's where we are with this. Well look no. One. Is correct. You. Have to be more. I agree and that's if we can actually which is something that we're already working on and they have submitted to the paper took a t.v. was how to work on the initial like.

[00:55:28]
Real time monitoring and processing of this data which is a nice thing to have if we can do that and hopefully will do a pilot study of that on this subset of patients and we can see if we can finesse that over time to train these models even better on this group and then you can actually to load it on different subset of patients the babies who underwent the r.t.l. switch or the patient to either went another type of heart surgery you can then train your model to each different subset of patients to become eventually Hellboy the whole cardiac i.c.u. will have something like that that will be the eventual dream.

[00:56:14]
Thank you so much because I think.