1 00:00:08,240 --> 00:00:11,519 Hello, and welcome to the Physics World weekly 2 00:00:11,519 --> 00:00:12,019 podcast. 3 00:00:12,320 --> 00:00:13,619 I'm Hamish Johnston. 4 00:00:14,434 --> 00:00:15,094 This episode 5 00:00:15,474 --> 00:00:15,974 features 6 00:00:16,274 --> 00:00:17,414 Amanda Randalls, 7 00:00:17,954 --> 00:00:20,695 who is a computer scientist and biomedical 8 00:00:21,154 --> 00:00:21,654 engineer 9 00:00:22,035 --> 00:00:23,254 at Duke University 10 00:00:23,554 --> 00:00:24,294 in The 11 00:00:25,234 --> 00:00:28,195 US. Duke is also where she began her 12 00:00:28,195 --> 00:00:29,335 scientific career 13 00:00:29,730 --> 00:00:33,270 by studying for an undergraduate degree in physics. 14 00:00:34,130 --> 00:00:36,950 For the past few years, she's been using 15 00:00:37,009 --> 00:00:37,909 physics based 16 00:00:38,450 --> 00:00:38,950 computationally 17 00:00:39,329 --> 00:00:39,829 intense 18 00:00:40,210 --> 00:00:40,710 simulations 19 00:00:41,329 --> 00:00:42,229 of the circulatory 20 00:00:42,689 --> 00:00:43,189 system 21 00:00:43,664 --> 00:00:46,164 to identify signs of heart disease 22 00:00:46,545 --> 00:00:48,164 before symptoms start. 23 00:00:48,945 --> 00:00:51,424 These are signs that show up not only 24 00:00:51,424 --> 00:00:55,184 in clinical tests, but also in data from 25 00:00:55,184 --> 00:00:56,244 wearable devices, 26 00:00:56,789 --> 00:00:57,929 such as smartwatches 27 00:00:58,789 --> 00:01:00,250 that record information 28 00:01:00,710 --> 00:01:03,850 while patients are going about their daily lives. 29 00:01:04,790 --> 00:01:08,250 More recently, Randalls has also become interested 30 00:01:08,629 --> 00:01:09,450 in understanding 31 00:01:09,750 --> 00:01:11,129 how cancer cells 32 00:01:11,495 --> 00:01:12,795 move through the bloodstream 33 00:01:13,255 --> 00:01:13,995 and initiate 34 00:01:14,534 --> 00:01:15,034 metastasis, 35 00:01:16,055 --> 00:01:19,115 which is the process where cancer that started 36 00:01:19,174 --> 00:01:21,275 in one location within the body 37 00:01:21,734 --> 00:01:23,515 spreads to other organs. 38 00:01:24,579 --> 00:01:26,119 In 2024, 39 00:01:26,340 --> 00:01:29,000 the Association for Computing Machinery 40 00:01:29,539 --> 00:01:33,399 awarded Randalls its ACM prize in computing 41 00:01:33,859 --> 00:01:34,920 for her groundbreaking 42 00:01:35,539 --> 00:01:37,000 contributions to algorithms 43 00:01:37,795 --> 00:01:38,854 and high performance 44 00:01:39,155 --> 00:01:40,215 computing methods 45 00:01:40,515 --> 00:01:41,814 aimed at diagnosing 46 00:01:42,274 --> 00:01:43,015 and treating 47 00:01:43,394 --> 00:01:44,375 human diseases. 48 00:01:45,314 --> 00:01:46,935 As well as scientific 49 00:01:47,314 --> 00:01:47,814 prestige 50 00:01:48,354 --> 00:01:51,155 and $250,000 51 00:01:51,155 --> 00:01:52,054 in cash, 52 00:01:52,599 --> 00:01:54,539 This prize comes with an invitation 53 00:01:55,079 --> 00:01:55,899 to the annual 54 00:01:56,280 --> 00:01:56,780 Heidelberg 55 00:01:57,239 --> 00:01:58,539 Laureate Forum, 56 00:01:58,920 --> 00:01:59,819 or HLF, 57 00:02:00,840 --> 00:02:02,939 which brings prize winning researchers 58 00:02:03,399 --> 00:02:05,179 and early career researchers 59 00:02:05,560 --> 00:02:07,819 in computer science and mathematics 60 00:02:08,495 --> 00:02:09,155 to Heidelberg, 61 00:02:09,455 --> 00:02:09,955 Germany 62 00:02:10,254 --> 00:02:12,675 for a week of talks and networking. 63 00:02:14,495 --> 00:02:17,555 My colleague Margaret Harris caught up with Randalls 64 00:02:17,775 --> 00:02:20,115 at last year's h l f. 65 00:02:20,469 --> 00:02:21,209 But interestingly, 66 00:02:21,669 --> 00:02:24,489 this wasn't Randalls first trip to the forum. 67 00:02:24,949 --> 00:02:26,569 She'd previously attended 68 00:02:26,870 --> 00:02:29,129 when she was an early career researcher 69 00:02:29,509 --> 00:02:30,009 herself. 70 00:02:30,709 --> 00:02:33,349 You'll hear her talk about this experience later 71 00:02:33,349 --> 00:02:34,250 in the podcast, 72 00:02:34,605 --> 00:02:35,105 But 73 00:02:35,485 --> 00:02:38,125 first, let's hear about her early interest in 74 00:02:38,125 --> 00:02:38,625 physics 75 00:02:38,925 --> 00:02:41,504 and how it fed into a concept 76 00:02:41,805 --> 00:02:44,385 that's known as a digital twin. 77 00:02:52,670 --> 00:02:54,670 My first question is, how did you get 78 00:02:54,670 --> 00:02:57,330 interested in physics in the first place? 79 00:02:58,189 --> 00:02:59,870 Yeah. I I think basic things of, like, 80 00:02:59,870 --> 00:03:01,629 you know, I was always a kid, you 81 00:03:01,629 --> 00:03:03,629 know, who was always asking lots of questions. 82 00:03:03,629 --> 00:03:05,365 And my parents ended up buying me, like, 83 00:03:05,365 --> 00:03:06,645 that whole set of the, like, ask me 84 00:03:06,645 --> 00:03:09,365 why books to, like, you know, try to 85 00:03:09,365 --> 00:03:10,264 address that. 86 00:03:10,564 --> 00:03:11,844 In high school, I went to a math 87 00:03:11,844 --> 00:03:13,444 and science center for half the day every 88 00:03:13,444 --> 00:03:15,844 day for, the high school programming. And that 89 00:03:15,844 --> 00:03:17,044 was where I, you know, first took some 90 00:03:17,044 --> 00:03:18,645 physics classes and really kind of fell in 91 00:03:18,645 --> 00:03:20,459 love with it and and really liked it. 92 00:03:20,840 --> 00:03:22,840 I had like, the the physics teacher there 93 00:03:22,840 --> 00:03:24,360 was fantastic and and just, I don't know, 94 00:03:24,360 --> 00:03:25,659 made everything really exciting. 95 00:03:26,120 --> 00:03:27,719 So I knew when I went to Duke, 96 00:03:27,719 --> 00:03:29,400 I really wanted to study some kinda like 97 00:03:29,400 --> 00:03:31,634 biophysics or or something in that space. 98 00:03:32,034 --> 00:03:33,474 And then, yeah, I just I I really 99 00:03:33,474 --> 00:03:34,754 liked it when I when I was there. 100 00:03:34,754 --> 00:03:36,034 I I I like the, you know, the 101 00:03:36,034 --> 00:03:37,394 approach of really, you know, how do you 102 00:03:37,394 --> 00:03:40,114 understand something from first principles and and really 103 00:03:40,114 --> 00:03:42,215 get to the answer. And And you're continuing 104 00:03:42,275 --> 00:03:44,349 to apply that sort of physics based approach 105 00:03:44,349 --> 00:03:46,150 to your work now, which is very much 106 00:03:46,150 --> 00:03:47,210 on the board between 107 00:03:47,510 --> 00:03:49,770 computer science, biomedical engineering, 108 00:03:50,150 --> 00:03:52,629 and particularly this concept of the digital twin. 109 00:03:52,629 --> 00:03:53,750 I wonder if you could just tell our 110 00:03:53,750 --> 00:03:56,010 listeners what a digital twin is. 111 00:03:57,034 --> 00:03:59,594 Yeah. So so, basically, like, we're we're reviewing 112 00:03:59,594 --> 00:04:01,754 a digital twin is where it's a virtual 113 00:04:01,754 --> 00:04:04,395 replica of a physical entity that's gonna be 114 00:04:04,395 --> 00:04:07,675 continuously updated with information driving, like, from that 115 00:04:07,675 --> 00:04:09,835 physical entity. So you have, you know, potentially 116 00:04:09,835 --> 00:04:12,510 sensors are gonna be giving continuous feedback to 117 00:04:12,510 --> 00:04:14,590 that entity and then allowing you to draw 118 00:04:14,590 --> 00:04:16,350 something that may, you know, then influence that 119 00:04:16,350 --> 00:04:17,490 entity in the future. 120 00:04:17,949 --> 00:04:19,550 I think this is an idea that actually 121 00:04:19,550 --> 00:04:22,110 has a quite long history in engineering. So, 122 00:04:22,110 --> 00:04:24,430 yeah. This this idea is it's very popular 123 00:04:24,430 --> 00:04:26,214 right now, but it's not it's it's not 124 00:04:26,214 --> 00:04:28,235 anything new. They've been doing it for years. 125 00:04:28,535 --> 00:04:30,774 It's been really popular in civil engineering for 126 00:04:30,774 --> 00:04:32,375 a long time. You see it a lot 127 00:04:32,375 --> 00:04:34,555 in mechanical engineering. So there there's constantly 128 00:04:34,935 --> 00:04:36,774 you know, we see strong examples of using, 129 00:04:36,774 --> 00:04:38,375 you know, digital twins of bridges where you 130 00:04:38,375 --> 00:04:40,134 have sensors all over the bridges to see, 131 00:04:40,134 --> 00:04:41,495 like, what the load is like and getting 132 00:04:41,495 --> 00:04:44,029 that information about, like, that specific bridge. They 133 00:04:44,029 --> 00:04:45,629 had the same thing. You know, many companies 134 00:04:45,629 --> 00:04:47,069 have been doing this for their jet engines 135 00:04:47,069 --> 00:04:47,810 for years. 136 00:04:48,430 --> 00:04:50,029 We're even starting to see it for, like, 137 00:04:50,029 --> 00:04:52,029 supercomputing centers. We have a digital twin at 138 00:04:52,029 --> 00:04:54,189 the supercomputing center to, you know, watch, you 139 00:04:54,189 --> 00:04:56,430 know, the power consumption, cooling, and that that 140 00:04:56,430 --> 00:04:58,725 side of things. So it's pretty pervasive. We're 141 00:04:58,725 --> 00:05:00,405 just we're we're trying to kind of make 142 00:05:00,405 --> 00:05:02,085 use of these advances and use it for 143 00:05:02,085 --> 00:05:03,444 the health side and start trying to make 144 00:05:03,444 --> 00:05:05,205 this digital twin of the human. It it's 145 00:05:05,205 --> 00:05:07,045 been around for a while. Could you give 146 00:05:07,045 --> 00:05:09,125 me an example of what's a digital twin 147 00:05:09,125 --> 00:05:10,884 might how it might work in practice in 148 00:05:10,884 --> 00:05:11,625 your field? 149 00:05:12,350 --> 00:05:14,670 So we're we're really starting with just focusing 150 00:05:14,670 --> 00:05:16,509 on a digital twin of vascular of a 151 00:05:16,589 --> 00:05:18,269 like a vascular digital twin. So we're looking 152 00:05:18,269 --> 00:05:20,829 at your circulatory system, and we're building a 153 00:05:20,829 --> 00:05:23,310 replica. So we need medical imaging. We get 154 00:05:23,310 --> 00:05:25,310 a three d model of your, you know, 155 00:05:25,310 --> 00:05:27,089 coronary arteries, for example. 156 00:05:27,535 --> 00:05:28,035 And 157 00:05:28,574 --> 00:05:31,615 in that case, we then can extract you 158 00:05:31,615 --> 00:05:33,535 know, we get the three d geometry representing 159 00:05:33,535 --> 00:05:36,175 that specific patient's coronary arteries, and then we 160 00:05:36,175 --> 00:05:38,894 get information from, like, wearable devices. So we 161 00:05:38,894 --> 00:05:40,894 can use that to understand, you know, what 162 00:05:40,894 --> 00:05:42,334 is your heart rate, kind of try to 163 00:05:42,334 --> 00:05:43,670 figure out what is what are the flow 164 00:05:43,670 --> 00:05:45,430 properties, and try to assess, you know, what 165 00:05:45,430 --> 00:05:47,430 would be the velocity going into those coronary 166 00:05:47,430 --> 00:05:49,910 arteries and drive the flow simulations from the 167 00:05:49,910 --> 00:05:51,189 data that we're able to get from the 168 00:05:51,189 --> 00:05:53,509 wearables. And then you could imagine as someone's, 169 00:05:53,509 --> 00:05:55,209 you know, going through their daily life, 170 00:05:55,925 --> 00:05:57,285 typically, what a doctor would look at is, 171 00:05:57,285 --> 00:05:58,564 you know, information they can get in the 172 00:05:58,564 --> 00:06:00,805 doctor's office, what they can measure in real 173 00:06:00,805 --> 00:06:02,664 time over just a couple heartbeats. 174 00:06:03,044 --> 00:06:05,144 And what we're trying to do is understand 175 00:06:05,604 --> 00:06:07,865 how is someone's three d blood flow changing 176 00:06:08,004 --> 00:06:09,925 as they go about, you know, a week 177 00:06:09,925 --> 00:06:11,610 or a month to see, you know, not 178 00:06:11,610 --> 00:06:13,610 just, like, did you enter a vulnerable state, 179 00:06:13,610 --> 00:06:15,050 but, you know, how long did you spend 180 00:06:15,050 --> 00:06:16,970 in these vulnerable states, and is that helping 181 00:06:16,970 --> 00:06:18,110 us predict or identify 182 00:06:18,970 --> 00:06:20,970 identify any kind of disease progressions? You could 183 00:06:20,970 --> 00:06:23,209 imagine, you know, someone wearing this for heart 184 00:06:23,209 --> 00:06:25,850 failure and before they ever have symptoms of, 185 00:06:25,850 --> 00:06:28,305 like, you know, shortness of breath or kind 186 00:06:28,305 --> 00:06:30,144 of chest pain, there may be a change 187 00:06:30,144 --> 00:06:30,964 in their hemodynamics 188 00:06:31,264 --> 00:06:32,805 that the digital twin can identify. 189 00:06:33,105 --> 00:06:34,464 And then the doctor might be able to, 190 00:06:34,464 --> 00:06:35,985 like, you know, see that that change has 191 00:06:35,985 --> 00:06:38,084 happened and either, you know, change their medication, 192 00:06:38,545 --> 00:06:40,144 ask them to come in, do something that's 193 00:06:40,144 --> 00:06:41,660 a bit more proactive. So kind of use 194 00:06:41,660 --> 00:06:43,339 it as a way of monitoring the patient 195 00:06:43,339 --> 00:06:45,100 while they're at home. And I guess one 196 00:06:45,100 --> 00:06:46,240 of the things that connects 197 00:06:46,540 --> 00:06:48,939 the examples you gave in civil engineering and 198 00:06:48,939 --> 00:06:51,019 what you're now trying to do with humans 199 00:06:51,019 --> 00:06:51,919 and health care 200 00:06:52,220 --> 00:06:53,600 is that you can't do 201 00:06:54,220 --> 00:06:54,720 infillimen 202 00:06:55,100 --> 00:06:55,600 experiments 203 00:06:56,314 --> 00:06:58,475 on other system, and you can't do things 204 00:06:58,475 --> 00:07:00,074 to the bridge that's gonna destroy it, just 205 00:07:00,074 --> 00:07:01,595 like you can't do anything to a human 206 00:07:01,595 --> 00:07:04,074 that's going to in the severely hormone, unless 207 00:07:04,074 --> 00:07:06,074 there's a commensurate benefit, like you're doing in 208 00:07:06,074 --> 00:07:08,875 face of test, for example. Right. So how 209 00:07:08,875 --> 00:07:10,709 does it work to sort of link 210 00:07:11,629 --> 00:07:13,110 what what's the time scales on which you're 211 00:07:13,110 --> 00:07:15,189 linking the information you're gathering from wearables, you're 212 00:07:15,189 --> 00:07:16,709 gathering from scans, and how are you up 213 00:07:16,789 --> 00:07:18,629 using it to update the digital twin so 214 00:07:18,629 --> 00:07:21,110 you get information about the the actual work? 215 00:07:21,110 --> 00:07:22,725 Yeah. I know. That's a great question. So 216 00:07:22,725 --> 00:07:24,084 the way the way we're doing it is 217 00:07:24,084 --> 00:07:25,285 that we really see it as, you know, 218 00:07:25,285 --> 00:07:26,964 a blending of data from things like the 219 00:07:26,964 --> 00:07:29,764 wearable sensors with the physics based modeling. What 220 00:07:29,764 --> 00:07:31,205 we're trying to set up right now is 221 00:07:31,205 --> 00:07:32,884 we're we've developed this method. We call it 222 00:07:32,884 --> 00:07:35,910 longitudinal hemodynamic mapping framework, and it allows you 223 00:07:35,910 --> 00:07:37,589 to really kind of take a parallel on 224 00:07:37,589 --> 00:07:39,689 time approach to simulating the hemodynamics. 225 00:07:40,230 --> 00:07:42,470 And it allowing like, basically, we're we're we're 226 00:07:42,470 --> 00:07:43,990 getting at, you know, a finite number of 227 00:07:43,990 --> 00:07:46,790 hemodynamic units that we can identify that if 228 00:07:46,790 --> 00:07:48,444 you can piece those back together and and, 229 00:07:48,444 --> 00:07:49,805 you know, figure out it's almost like a 230 00:07:49,805 --> 00:07:51,004 lookup table of, you know, you figure out 231 00:07:51,004 --> 00:07:53,004 what these hemodynamic units are. You use your 232 00:07:53,004 --> 00:07:54,925 wearable data and you figure out which hemodynamic 233 00:07:54,925 --> 00:07:57,084 unit you should be pulling from. What we 234 00:07:57,084 --> 00:07:58,365 need to do then is simulate all of 235 00:07:58,365 --> 00:08:00,444 those hemodynamic units when we first get the 236 00:08:00,444 --> 00:08:02,365 medical imaging. So we're trying to minimize the 237 00:08:02,365 --> 00:08:03,699 amount of time it takes, you know, the 238 00:08:03,699 --> 00:08:06,019 number of hemodynamic units and how efficiently we 239 00:08:06,019 --> 00:08:08,180 can calculate each of those. But by doing 240 00:08:08,180 --> 00:08:09,560 that, you know, it might take 241 00:08:10,019 --> 00:08:11,620 somewhere in the order of, like, fifteen minutes 242 00:08:11,620 --> 00:08:13,220 to half an hour. If it's a really 243 00:08:13,220 --> 00:08:15,139 complicated geometry, maybe a little bit longer to 244 00:08:15,139 --> 00:08:16,199 segment the geometry 245 00:08:16,595 --> 00:08:17,795 and get the data, like, you know, get 246 00:08:17,795 --> 00:08:19,875 that geometry from the medical imaging. Once you 247 00:08:19,875 --> 00:08:21,715 have that, then it can take, you know, 248 00:08:21,715 --> 00:08:23,715 a few hours to maybe a day or 249 00:08:23,715 --> 00:08:24,215 two, 250 00:08:24,675 --> 00:08:27,074 to run all of those hemodynamic units. But 251 00:08:27,074 --> 00:08:28,915 then the goal is then in practice is 252 00:08:28,915 --> 00:08:30,215 once you've run the simulations, 253 00:08:30,620 --> 00:08:32,220 it works like a lookup table. You can 254 00:08:32,220 --> 00:08:34,139 then use that. So then, you know, it 255 00:08:34,139 --> 00:08:36,299 could almost run on the on your devices 256 00:08:36,299 --> 00:08:37,360 or on edge computing, 257 00:08:37,660 --> 00:08:38,160 setups 258 00:08:38,540 --> 00:08:40,460 to have, you know, draw driven from your 259 00:08:40,460 --> 00:08:42,540 wearable device to identify what are those hemodynamic 260 00:08:42,540 --> 00:08:44,714 units and then have algorithms on that that 261 00:08:44,714 --> 00:08:46,315 can can kind of track and see what 262 00:08:46,315 --> 00:08:48,315 the the markers are. And that's, you know, 263 00:08:48,315 --> 00:08:50,475 how we would see it for diagnostics or 264 00:08:50,475 --> 00:08:51,134 for tracking. 265 00:08:51,674 --> 00:08:53,274 But then there's, you know, the other side 266 00:08:53,274 --> 00:08:55,194 of things where, you know, that's assuming we 267 00:08:55,194 --> 00:08:56,794 know what the marker is that we're looking 268 00:08:56,794 --> 00:08:58,714 for. There's a whole another piece where it's 269 00:08:58,714 --> 00:09:00,740 trying to just, you know, do a discovery 270 00:09:00,740 --> 00:09:02,899 phase where we wanna create, save all of 271 00:09:02,899 --> 00:09:04,980 the three d blood flow over six months 272 00:09:04,980 --> 00:09:07,059 of time, do a large scale, you know, 273 00:09:07,059 --> 00:09:09,379 analysis to identify what is that marker that 274 00:09:09,379 --> 00:09:12,044 is changing before you have shortness of breath, 275 00:09:12,524 --> 00:09:14,204 and can we identify what that is? So, 276 00:09:14,204 --> 00:09:15,884 you know, that takes much longer and actually 277 00:09:15,884 --> 00:09:17,884 needs, you know, full simulations, all of the 278 00:09:17,884 --> 00:09:18,704 data retained. 279 00:09:19,004 --> 00:09:20,365 But that allows you to then, you know, 280 00:09:20,365 --> 00:09:22,365 create potential AI surrogates to go on the 281 00:09:22,365 --> 00:09:23,804 edge and pieces like that. So there's there's 282 00:09:23,804 --> 00:09:25,485 kinda two pieces of when you would wanna 283 00:09:25,485 --> 00:09:27,004 deploy it and how you'd really use it 284 00:09:27,004 --> 00:09:28,019 to find new markers. 285 00:09:28,660 --> 00:09:30,420 And I can see so you're talking about 286 00:09:30,420 --> 00:09:33,080 simulating these little chunks of a bloodstream effectively. 287 00:09:33,379 --> 00:09:35,220 It's not not quite sort of individual red 288 00:09:35,220 --> 00:09:37,460 blood cells, but are are are chunks of 289 00:09:37,460 --> 00:09:38,040 the bloodstream. 290 00:09:38,660 --> 00:09:41,460 How much computing power does that currently take 291 00:09:41,460 --> 00:09:43,544 and what challenges does that 292 00:09:44,004 --> 00:09:45,125 present? Like, I mean, a lot of our 293 00:09:45,125 --> 00:09:48,105 goal is to reduce the computing power. So 294 00:09:48,245 --> 00:09:49,684 a lot of this is an interesting, 295 00:09:50,164 --> 00:09:51,684 we have to do, like, the brute force 296 00:09:51,684 --> 00:09:52,184 simulation 297 00:09:52,804 --> 00:09:54,404 to make sure we're, like, you know, we're 298 00:09:54,404 --> 00:09:56,440 keeping everything in that we need to. So 299 00:09:56,440 --> 00:09:58,200 we're it requires a lot of times, like, 300 00:09:58,200 --> 00:09:59,660 the world's largest supercomputers, 301 00:10:00,279 --> 00:10:02,379 but that's only to run, like, the validation 302 00:10:02,440 --> 00:10:04,360 ground truth simulation that we're then using to 303 00:10:04,360 --> 00:10:06,680 test our algorithms against. So we're we're working 304 00:10:06,840 --> 00:10:07,960 you know, we have a lot of different 305 00:10:07,960 --> 00:10:09,720 models to go, you know, longer in time 306 00:10:09,720 --> 00:10:10,360 or deeper, 307 00:10:10,920 --> 00:10:11,420 spatially 308 00:10:12,084 --> 00:10:13,845 that, you know, we run the huge simulation 309 00:10:13,845 --> 00:10:15,924 on the large supercomputer, and then we'll run, 310 00:10:15,924 --> 00:10:17,684 you know, on several nodes in the cloud 311 00:10:17,684 --> 00:10:19,304 and compare against that simulation. 312 00:10:19,684 --> 00:10:21,044 So to to really, you know, put it 313 00:10:21,044 --> 00:10:22,404 into use, you would only need the few 314 00:10:22,404 --> 00:10:24,245 nodes on the cloud. But it's, you know, 315 00:10:24,245 --> 00:10:26,940 critically important to do the testing. And should 316 00:10:26,940 --> 00:10:28,620 note, like, we're pairing that testing with, you 317 00:10:28,620 --> 00:10:30,779 know, testing with Doppler ultrasound and testing with, 318 00:10:30,779 --> 00:10:33,500 you know, implantable devices. So also trying to 319 00:10:33,500 --> 00:10:35,980 make sure it's actually matching the patients as 320 00:10:35,980 --> 00:10:38,220 well in the experimental side, but I think 321 00:10:38,220 --> 00:10:40,620 having that verification with, like, the ground truth 322 00:10:40,620 --> 00:10:42,985 has been important. So, yeah, we we we 323 00:10:42,985 --> 00:10:44,584 do use a lot of the, yeah, like, 324 00:10:44,584 --> 00:10:46,424 the worst biggest supercomputers to run some of 325 00:10:46,424 --> 00:10:48,024 these simulations. And our goal is to try 326 00:10:48,024 --> 00:10:49,704 to, you know, put absolutely everything into the 327 00:10:49,704 --> 00:10:51,225 model so we can figure out what we 328 00:10:51,225 --> 00:10:52,424 can take out and how to pair it 329 00:10:52,424 --> 00:10:53,625 down. And so you do that book out 330 00:10:53,625 --> 00:10:54,840 table every time. Yes. 331 00:10:55,800 --> 00:10:57,820 Yeah. What are some principles that's 332 00:10:58,200 --> 00:11:00,360 that you're using in your work you're looking 333 00:11:00,360 --> 00:11:02,059 for in when you're simulating, 334 00:11:02,600 --> 00:11:05,559 the circulatory system? Yeah. So I guess, fundamentally, 335 00:11:05,559 --> 00:11:06,920 I think the the key piece is, like, 336 00:11:06,920 --> 00:11:08,804 it's not just we're not just looking at, 337 00:11:08,804 --> 00:11:10,644 like, random data points. There's not, like, just 338 00:11:10,644 --> 00:11:11,544 a data driven, 339 00:11:11,924 --> 00:11:14,084 approach. Like, this really truly is a physics 340 00:11:14,084 --> 00:11:14,584 informed, 341 00:11:15,284 --> 00:11:17,304 like, we are running physics based modeling, 342 00:11:17,764 --> 00:11:19,444 where you are getting less, like, three d 343 00:11:19,444 --> 00:11:20,424 blood flow model. 344 00:11:20,889 --> 00:11:22,490 We are taking that three d model that 345 00:11:22,490 --> 00:11:24,089 you're getting out of the the medical imaging 346 00:11:24,089 --> 00:11:24,750 from segmentation. 347 00:11:25,209 --> 00:11:27,049 We use a discretized approach. So we put, 348 00:11:27,049 --> 00:11:29,149 like, a regular like, just a regular Cartesian 349 00:11:29,209 --> 00:11:30,669 grid across that mesh. 350 00:11:31,049 --> 00:11:32,409 We figure out, like, what's an inlet node, 351 00:11:32,409 --> 00:11:33,845 what's an outlet node, and then you're just 352 00:11:33,845 --> 00:11:36,644 solving, you know, very basic fluid dynamics equations 353 00:11:36,644 --> 00:11:38,024 at each of those grid points. 354 00:11:38,964 --> 00:11:41,284 It's really, really, you know, very fine mesh, 355 00:11:41,284 --> 00:11:42,725 so you need, like, you know, so many 356 00:11:42,725 --> 00:11:44,324 grid points you end up needing large scale 357 00:11:44,324 --> 00:11:44,824 supercomputers. 358 00:11:45,204 --> 00:11:46,259 But at the end of the day, you're 359 00:11:46,259 --> 00:11:47,379 just solving you know, we use a lot 360 00:11:47,379 --> 00:11:48,740 of Boltzmann, but it's a way of solving 361 00:11:48,740 --> 00:11:50,659 Navier Stokes. So we're really just solving, you 362 00:11:50,659 --> 00:11:52,820 know, basic fluid equations at each of these 363 00:11:52,820 --> 00:11:54,659 at each of these grid points. And that 364 00:11:54,659 --> 00:11:55,959 allows us to recover, 365 00:11:56,259 --> 00:11:57,959 you know, the three d float profile. 366 00:11:58,464 --> 00:12:00,784 So we can get quantities like vorticity, velocity, 367 00:12:00,784 --> 00:12:02,304 we need pressure, and we can get that 368 00:12:02,304 --> 00:12:04,225 in a three d format. And then we're 369 00:12:04,225 --> 00:12:06,884 combining that with cellular models. So we have 370 00:12:07,264 --> 00:12:09,504 finite element models of, you know, red blood 371 00:12:09,504 --> 00:12:11,745 cells as they're moving through the geometry, how 372 00:12:11,745 --> 00:12:13,365 they're interacting with that fluid. 373 00:12:13,679 --> 00:12:15,679 But it is it's important to have, like, 374 00:12:15,679 --> 00:12:17,759 a physics understanding and, like, a physics based 375 00:12:17,759 --> 00:12:19,279 model of, like, how is that red blood 376 00:12:19,279 --> 00:12:22,159 cell deforming as it's moving through? Where is 377 00:12:22,159 --> 00:12:24,080 it going to adhere? How is it interacting? 378 00:12:24,080 --> 00:12:25,440 And it it it's all coming from the 379 00:12:25,440 --> 00:12:25,940 physics. 380 00:12:26,754 --> 00:12:28,754 I think that's been your your focus for 381 00:12:28,754 --> 00:12:31,154 for several years now. Yeah. And but you're 382 00:12:31,154 --> 00:12:33,315 now actually starting to shift as well while 383 00:12:33,315 --> 00:12:34,855 still working on that to 384 00:12:35,315 --> 00:12:38,195 understanding, like, the biology and the physics impact 385 00:12:38,195 --> 00:12:40,034 of of cancer and what that looks like 386 00:12:40,034 --> 00:12:41,815 in a digital twin sort of simulation. 387 00:12:42,700 --> 00:12:44,940 What's that look like for you? Yeah. So 388 00:12:44,940 --> 00:12:46,700 everything we've done on, like, the heart disease 389 00:12:46,700 --> 00:12:49,340 side, you really don't need to have individual 390 00:12:49,340 --> 00:12:51,519 cells in that case. Like, having bulk fluid, 391 00:12:51,659 --> 00:12:53,899 you know, is enough to recover, like, the 392 00:12:53,899 --> 00:12:55,580 values that we're interested in that are known 393 00:12:55,580 --> 00:12:58,134 biomarkers for heart disease. But by having this 394 00:12:58,134 --> 00:12:58,875 large scale, 395 00:12:59,735 --> 00:13:01,575 kind of framework, you know, we started adding 396 00:13:01,575 --> 00:13:03,455 in red blood cells probably I don't know. 397 00:13:03,455 --> 00:13:05,674 I think, like, 2,009, 2,010. 398 00:13:05,975 --> 00:13:08,455 And, we realized you could use this model 399 00:13:08,455 --> 00:13:10,634 for studying cancer and trying to understand 400 00:13:11,290 --> 00:13:12,970 we we don't look at tumors themselves. We 401 00:13:12,970 --> 00:13:14,889 tend to look at when a cancer cell 402 00:13:14,889 --> 00:13:16,730 has ripped off a primary tumor cell and 403 00:13:16,730 --> 00:13:18,029 is moving through the bloodstream. 404 00:13:18,490 --> 00:13:20,410 We wanna understand what it is about that 405 00:13:20,410 --> 00:13:22,330 cancer cell that is making it choose one 406 00:13:22,330 --> 00:13:24,225 path or the other path. Why is it 407 00:13:24,225 --> 00:13:26,304 spending so much time at certain locations on 408 00:13:26,304 --> 00:13:28,544 the wall? What's gonna make it more likely 409 00:13:28,544 --> 00:13:30,485 to adhere to that location on the wall? 410 00:13:31,024 --> 00:13:32,464 So from that end, it's, you know, it 411 00:13:32,544 --> 00:13:35,264 it's an incredibly computationally intense but very physics 412 00:13:35,264 --> 00:13:37,504 based question. So we, you know, we're trying 413 00:13:37,504 --> 00:13:38,085 to build 414 00:13:38,730 --> 00:13:40,649 accurate models of the cell membrane so we 415 00:13:40,649 --> 00:13:43,210 can, you know, compare to microfluidic devices and 416 00:13:43,210 --> 00:13:44,809 make sure, you know, we're getting those specific 417 00:13:44,809 --> 00:13:45,950 cancer cells accurately. 418 00:13:46,570 --> 00:13:48,409 We're trying to add we have models of, 419 00:13:48,409 --> 00:13:50,570 like, ligand receptor interactions. So we have you 420 00:13:50,570 --> 00:13:51,929 know, one one of the projects we have 421 00:13:51,929 --> 00:13:52,730 going right now is, 422 00:13:53,384 --> 00:13:55,545 having the large scale fluid model to understand 423 00:13:55,545 --> 00:13:57,785 the wall shear stress on, like, the vascular 424 00:13:57,865 --> 00:13:59,804 like, on the endothelium and on the walls, 425 00:13:59,945 --> 00:14:01,865 and then using that to drive these the 426 00:14:01,865 --> 00:14:04,105 receptor expression and understand how that's, you know, 427 00:14:04,105 --> 00:14:07,220 influencing the endothelium and causing different receptors. Like, 428 00:14:07,220 --> 00:14:09,220 trying to, like, predict what that pattern is 429 00:14:09,220 --> 00:14:10,740 going to be based on that wall shear 430 00:14:10,740 --> 00:14:11,559 stress interaction. 431 00:14:12,180 --> 00:14:14,019 And then you have the computational model of 432 00:14:14,019 --> 00:14:15,700 the cancer cell moving through to see where 433 00:14:15,700 --> 00:14:17,860 is it going to here and really try 434 00:14:17,860 --> 00:14:19,779 to, you know, find a way of parsing 435 00:14:19,779 --> 00:14:22,455 the dynamics of, you know, the fluid combination 436 00:14:22,455 --> 00:14:24,054 and how it's affecting the cells and then 437 00:14:24,054 --> 00:14:26,315 how the cells are interacting in that space 438 00:14:26,535 --> 00:14:29,174 and giving you this way. The computational tools 439 00:14:29,174 --> 00:14:31,095 are fantastic and that, you know, we can 440 00:14:31,095 --> 00:14:34,154 control absolutely everything. You can leave everything exactly 441 00:14:34,215 --> 00:14:36,470 the same and tweak just, you know, the 442 00:14:36,470 --> 00:14:39,429 velocity slightly or the bond strength slightly. And 443 00:14:39,429 --> 00:14:41,269 then you can really isolate, you know, how 444 00:14:41,269 --> 00:14:42,970 did that bond strength change 445 00:14:43,350 --> 00:14:44,629 the amount of time you spent at the 446 00:14:44,629 --> 00:14:46,629 wall or your residence time for that cancer 447 00:14:46,629 --> 00:14:48,345 cell. And it really lets us 448 00:14:48,824 --> 00:14:50,824 fine tune test these items that it's harder 449 00:14:50,824 --> 00:14:52,584 to do in an experiment when, you know, 450 00:14:52,584 --> 00:14:54,184 humidity might be changing. Or, like, you know, 451 00:14:54,184 --> 00:14:56,184 there are many other quantities that are changing 452 00:14:56,184 --> 00:14:58,024 that you can't control. So it offers, you 453 00:14:58,024 --> 00:14:59,784 know, a nice compliment to the experiments in 454 00:14:59,784 --> 00:15:02,470 that way. And are you exclusively interested in 455 00:15:02,470 --> 00:15:02,970 this 456 00:15:03,590 --> 00:15:06,309 metastasis process whereby a a cell or a 457 00:15:06,309 --> 00:15:08,470 cancer cell breaks off from the initial tumor 458 00:15:08,470 --> 00:15:10,250 then travels to somewhere else in the bloodstream 459 00:15:11,029 --> 00:15:13,014 to produce a cancer somewhere else? Is that 460 00:15:13,014 --> 00:15:14,695 is that the main how they're focusing it? 461 00:15:14,695 --> 00:15:16,454 Yeah. That that's really what we've looked at 462 00:15:16,454 --> 00:15:18,214 at this point. We're we're starting to branch 463 00:15:18,214 --> 00:15:19,414 a little bit into like, you know, we 464 00:15:19,414 --> 00:15:20,934 we model all the red blood cells. We're 465 00:15:20,934 --> 00:15:22,214 kind of branching a little bit into, you 466 00:15:22,214 --> 00:15:23,815 know, how to what changes with sickle cell 467 00:15:23,815 --> 00:15:24,315 anemia, 468 00:15:24,934 --> 00:15:26,454 like, you know, changes to the red blood 469 00:15:26,454 --> 00:15:27,274 cells themselves. 470 00:15:27,654 --> 00:15:29,860 And even just as you age, how do 471 00:15:29,860 --> 00:15:31,779 the mechanical properties of the red blood cells 472 00:15:31,779 --> 00:15:33,539 change, and how does that affect, you know, 473 00:15:33,539 --> 00:15:35,940 your response to diseases? So we're kind of 474 00:15:35,940 --> 00:15:37,620 branching into, like, the red blood cell side 475 00:15:37,620 --> 00:15:39,299 as well, but a a lot of the 476 00:15:39,299 --> 00:15:40,820 focus has been just on the can we 477 00:15:40,820 --> 00:15:42,705 understand what it is about the cancer cell 478 00:15:42,785 --> 00:15:44,945 that's driving like, the mechanics of the cancer 479 00:15:44,945 --> 00:15:47,585 cell that's driving cancer metastasis or the likelihood 480 00:15:47,585 --> 00:15:48,644 of disease progression. 481 00:15:49,585 --> 00:15:52,065 Thinking about challenges now as you try to 482 00:15:52,065 --> 00:15:54,465 move forward in research, what would you say 483 00:15:54,465 --> 00:15:55,125 the biggest 484 00:15:55,504 --> 00:15:56,884 physics related challenges 485 00:15:57,345 --> 00:15:57,845 are? 486 00:15:58,759 --> 00:16:00,120 Yeah. I think a lot of it has 487 00:16:00,120 --> 00:16:03,000 been, trying to make stable simulations that can 488 00:16:03,000 --> 00:16:05,500 cross spatial and temporal boundaries. So 489 00:16:06,200 --> 00:16:07,500 for the cancer project, 490 00:16:07,879 --> 00:16:09,019 we want to understand 491 00:16:09,399 --> 00:16:11,320 it's not just the deformation of the cancer 492 00:16:11,320 --> 00:16:13,080 cell or the nucleus which requires a high 493 00:16:13,080 --> 00:16:13,580 resolution. 494 00:16:14,034 --> 00:16:15,875 But then we really want these ligand receptor 495 00:16:15,875 --> 00:16:18,355 pairings, but we wanna understand how it how 496 00:16:18,355 --> 00:16:20,695 those, you know, adhesion is changing the behavior 497 00:16:21,075 --> 00:16:22,995 over a centimeter scale or, you know, it'd 498 00:16:22,995 --> 00:16:24,514 be great if we get a mill like, 499 00:16:24,514 --> 00:16:26,754 a a meter scale. But by doing that, 500 00:16:26,754 --> 00:16:28,649 you you know, we've been using, like, this 501 00:16:28,649 --> 00:16:30,569 approach we call, like, adaptive physics refinement. It's 502 00:16:30,569 --> 00:16:32,569 very similar in spirit to, like, adaptive mesh 503 00:16:32,569 --> 00:16:33,069 refinement. 504 00:16:33,529 --> 00:16:34,569 But as soon as you go in that 505 00:16:34,569 --> 00:16:35,690 direction, it's like, you know, how do you 506 00:16:35,690 --> 00:16:38,009 stably couple these different levels of different mesh 507 00:16:38,009 --> 00:16:39,769 refinements and how do you get you know, 508 00:16:39,769 --> 00:16:41,610 in our case, we have one physics model 509 00:16:41,610 --> 00:16:43,529 being used at the the fine grain model, 510 00:16:43,529 --> 00:16:45,605 the coarse grain model, a different physics unit. 511 00:16:45,985 --> 00:16:47,665 And how do you have that couple in 512 00:16:47,665 --> 00:16:49,585 a stable way that is, you know, accurate 513 00:16:49,585 --> 00:16:50,325 and continuing? 514 00:16:50,785 --> 00:16:52,384 We have, like, in the in the coarse 515 00:16:52,384 --> 00:16:54,304 grain model, it's the bulk fluid, which has 516 00:16:54,304 --> 00:16:56,465 a very viscous fluid. So from the even 517 00:16:56,465 --> 00:16:58,710 just the fluid side, it's coupling to, like, 518 00:16:58,710 --> 00:17:00,570 the plasma, which is more like a water. 519 00:17:00,629 --> 00:17:02,070 So you're you're introducing it just a lot 520 00:17:02,070 --> 00:17:04,730 of points of potential instability, and it's, it's 521 00:17:04,789 --> 00:17:06,230 requiring a lot more on the physics and 522 00:17:06,230 --> 00:17:08,150 the algorithm side versus just, you know, we 523 00:17:08,150 --> 00:17:09,589 know the equation. Let's try to make it 524 00:17:09,589 --> 00:17:12,765 faster. And there's there are current limitations in 525 00:17:12,825 --> 00:17:15,305 what the resolution gap between that fine grain 526 00:17:15,305 --> 00:17:17,705 and the coarse grain model can be. And 527 00:17:17,705 --> 00:17:19,465 the more we can push that boundary and 528 00:17:19,465 --> 00:17:21,805 allow that to widen would really help us. 529 00:17:22,184 --> 00:17:24,505 And in terms of computational challenges, what are 530 00:17:24,505 --> 00:17:26,210 the big issues there? I think at the 531 00:17:26,210 --> 00:17:27,930 point at this point, like, data is a 532 00:17:27,930 --> 00:17:29,529 huge issue of, you know, we I keep 533 00:17:29,529 --> 00:17:31,369 saying, like, we're really great at, you know, 534 00:17:31,369 --> 00:17:33,130 the input, the STL file that you're getting 535 00:17:33,130 --> 00:17:34,509 at the triangulated mesh is 536 00:17:34,809 --> 00:17:35,950 probably a couple of kilobytes. 537 00:17:36,410 --> 00:17:38,250 You're getting, you know, a few your input 538 00:17:38,250 --> 00:17:40,009 file is, like, 10 numbers in it. It's, 539 00:17:40,009 --> 00:17:41,984 you know, it's very small data. And then 540 00:17:41,984 --> 00:17:44,164 we are running on the world's biggest supercomputers 541 00:17:44,384 --> 00:17:46,464 and using all of the data, which can 542 00:17:46,464 --> 00:17:47,904 be on the order of a petabyte of 543 00:17:47,904 --> 00:17:49,904 data for one time step. And then you're 544 00:17:49,904 --> 00:17:51,904 trying to run for millions and millions of 545 00:17:51,904 --> 00:17:53,684 time steps. So, you know, 546 00:17:54,065 --> 00:17:54,805 we're generating 547 00:17:55,184 --> 00:17:58,089 petabytes and petabytes of data. Obviously, we can't 548 00:17:58,089 --> 00:17:59,450 store all of that data. So it's, you 549 00:17:59,450 --> 00:18:01,049 know, how do you analyze that data? How 550 00:18:01,049 --> 00:18:02,409 do you make use of it? What do 551 00:18:02,409 --> 00:18:03,150 you store? 552 00:18:03,529 --> 00:18:05,230 It's like data has really become 553 00:18:05,609 --> 00:18:07,049 a a huge part of it. Like, we're 554 00:18:07,049 --> 00:18:08,250 we're doing a lot of work on in 555 00:18:08,250 --> 00:18:10,409 situ machine learning and in situ visualization and 556 00:18:10,409 --> 00:18:11,690 what can we do while the data is 557 00:18:11,690 --> 00:18:14,075 still in memory on the super computer. But 558 00:18:14,075 --> 00:18:15,835 it just trying to figure out, you know, 559 00:18:15,835 --> 00:18:17,115 and that's, you know, just for one patient, 560 00:18:17,115 --> 00:18:19,035 if you wanna start looking across multiple patients, 561 00:18:19,035 --> 00:18:20,795 how do you bring all this together and 562 00:18:20,795 --> 00:18:22,475 actually analyze it in a way that that's 563 00:18:22,475 --> 00:18:23,615 tractable and useful? 564 00:18:23,994 --> 00:18:25,195 It's I think we we've had a lot 565 00:18:25,195 --> 00:18:26,700 of success scaling the code, which is 566 00:18:31,500 --> 00:18:31,525 is great, but that has then led to 567 00:18:31,525 --> 00:18:31,551 this problem of now we have tons of 568 00:18:31,551 --> 00:18:31,576 data, which is, you know, a good problem 569 00:18:31,576 --> 00:18:33,359 to have, but it's it's definitely an issue. 570 00:18:33,419 --> 00:18:35,500 Touch the part more businesses. I know. Yeah. 571 00:18:35,500 --> 00:18:35,980 It's 572 00:18:37,259 --> 00:18:39,579 So part about this, what is something that 573 00:18:39,579 --> 00:18:41,119 you can't do now 574 00:18:41,924 --> 00:18:44,085 that you see coming in the future and 575 00:18:44,085 --> 00:18:45,625 you're really excited about? 576 00:18:46,484 --> 00:18:47,384 Right now, 577 00:18:48,085 --> 00:18:49,524 for the for the right now, for the 578 00:18:49,524 --> 00:18:51,444 digital twins, especially even just on the heart 579 00:18:51,444 --> 00:18:51,944 side, 580 00:18:52,325 --> 00:18:53,704 we've done a lot of work 581 00:18:54,079 --> 00:18:56,079 validating that, like, at the we can match 582 00:18:56,079 --> 00:18:57,759 single time point measures of what you get 583 00:18:57,759 --> 00:18:59,440 in the clinic. And then that's, you know, 584 00:18:59,440 --> 00:19:01,619 a critical first step, really important. 585 00:19:01,920 --> 00:19:03,519 But what we want to see happen is 586 00:19:03,519 --> 00:19:05,039 that we are getting the right data as 587 00:19:05,039 --> 00:19:06,320 you go out into the world and you're 588 00:19:06,320 --> 00:19:08,240 wearing your wearable and we're getting the right 589 00:19:08,240 --> 00:19:10,455 data. And it it looks promising, but we 590 00:19:10,455 --> 00:19:11,894 have a lot of studies going on to, 591 00:19:11,894 --> 00:19:13,974 like, validate that data. You know, we've been 592 00:19:13,974 --> 00:19:15,654 doing a lot of comparisons against him for 593 00:19:15,654 --> 00:19:18,295 heart failure for the an implantable device that 594 00:19:18,295 --> 00:19:19,894 they it measures once a day when you're 595 00:19:19,894 --> 00:19:21,275 laying down, when you're sedentary. 596 00:19:21,710 --> 00:19:23,630 And we can match that very strongly, and 597 00:19:23,630 --> 00:19:25,710 it looks really promising. But now we need 598 00:19:25,710 --> 00:19:27,470 to do it while you're exercising, while you're 599 00:19:27,470 --> 00:19:29,069 moving around and see can we get heart 600 00:19:29,069 --> 00:19:31,650 rate recovery. Are we really capturing what's happening 601 00:19:32,029 --> 00:19:34,130 throughout the day and dynamically? And I think 602 00:19:34,345 --> 00:19:35,725 once we can do that, 603 00:19:36,184 --> 00:19:37,705 it really opens up that's where you're gonna 604 00:19:37,705 --> 00:19:39,144 find the new biomarkers, and that's where you're 605 00:19:39,144 --> 00:19:40,904 gonna be able to identify something that goes 606 00:19:40,904 --> 00:19:42,745 further than what we have now. And it's 607 00:19:42,904 --> 00:19:45,144 you know, we have tons of collaborations in 608 00:19:45,144 --> 00:19:46,745 place to get the data we need from 609 00:19:46,745 --> 00:19:48,924 wearables to be able to make that jump, 610 00:19:49,150 --> 00:19:50,829 But it's still an open question of, like, 611 00:19:50,829 --> 00:19:52,269 what's gonna be the right way of measuring? 612 00:19:52,269 --> 00:19:53,789 Like, you know, for instance, we need stroke 613 00:19:53,789 --> 00:19:56,589 volume. That is an incredibly hard measurement to 614 00:19:56,589 --> 00:19:59,390 get with wearables, and we have many projects 615 00:19:59,390 --> 00:20:00,670 trying to figure out how to do that 616 00:20:00,670 --> 00:20:03,674 accurately. But it's until you really have that 617 00:20:03,674 --> 00:20:06,234 estimation in there, it it's incredibly promising, but 618 00:20:06,234 --> 00:20:08,075 there's still work to be done. So stroke 619 00:20:08,075 --> 00:20:10,315 volume is how much blood gets pulse when 620 00:20:10,315 --> 00:20:12,474 this with a single heartbeat. Yes. Yep. So 621 00:20:12,474 --> 00:20:14,154 we can get heart rate easily, but we 622 00:20:14,154 --> 00:20:15,994 really need cardiac output. And for that, you 623 00:20:15,994 --> 00:20:18,179 need heart rate and stroke volume. And, Usually, 624 00:20:18,179 --> 00:20:20,259 like, a tiny robot inside of the body, 625 00:20:20,259 --> 00:20:22,660 kind of measuring things inside the the artery 626 00:20:22,660 --> 00:20:24,599 or something. Yes. Yep. Yeah. 627 00:20:25,220 --> 00:20:26,599 That would be really helpful. 628 00:20:27,539 --> 00:20:29,234 And talk a little bit about how the 629 00:20:29,234 --> 00:20:30,914 physics that you use in your work. Which 630 00:20:30,914 --> 00:20:32,674 aspect of physics would you say that you 631 00:20:32,674 --> 00:20:35,335 use most? What lesson from your physics education 632 00:20:35,394 --> 00:20:37,654 do you do to keep coming back to? 633 00:20:38,835 --> 00:20:40,115 I feel like it's more just like the 634 00:20:40,115 --> 00:20:41,015 approach of 635 00:20:41,419 --> 00:20:43,259 really, you know, trying to under like, trying 636 00:20:43,259 --> 00:20:45,519 to understand what's going on from the fundamental. 637 00:20:45,660 --> 00:20:47,679 Like, how do you describe the problem? 638 00:20:48,059 --> 00:20:49,259 I don't even know how to describe I 639 00:20:49,259 --> 00:20:50,700 know I've had many conversations where I I 640 00:20:50,700 --> 00:20:52,380 think everybody approaches a problem in that way, 641 00:20:52,380 --> 00:20:53,660 and then you're like, oh, no. No. This 642 00:20:53,660 --> 00:20:56,240 is definitely a physicist. Like, think of mindset. 643 00:20:56,775 --> 00:20:59,115 And it's, I think it's a very systematic 644 00:20:59,174 --> 00:21:01,335 way of breaking down problems and how you 645 00:21:01,335 --> 00:21:02,394 approach it that 646 00:21:02,695 --> 00:21:04,455 is there are so many things that is 647 00:21:04,535 --> 00:21:05,894 like, in biology, there are so many things 648 00:21:05,894 --> 00:21:07,335 that we can't account for or we don't 649 00:21:07,335 --> 00:21:08,855 know. And I know you kind of always 650 00:21:08,855 --> 00:21:10,375 think of the spherical cow of, like, okay. 651 00:21:10,375 --> 00:21:11,789 How close can we get? What do we 652 00:21:11,789 --> 00:21:13,470 need to take into our model? And having 653 00:21:13,470 --> 00:21:14,210 that approach, 654 00:21:14,829 --> 00:21:16,210 really helps us understand, 655 00:21:16,509 --> 00:21:17,789 you know, we're not gonna be able to 656 00:21:17,789 --> 00:21:18,929 take into account everything 657 00:21:19,230 --> 00:21:20,669 for the human body and where can we 658 00:21:20,669 --> 00:21:22,269 make those assumptions, but how do we do 659 00:21:22,269 --> 00:21:23,789 that in an appropriate way? And I think 660 00:21:23,789 --> 00:21:24,990 you kind of learn a lot of that 661 00:21:24,990 --> 00:21:26,289 from the physics education. 662 00:21:27,125 --> 00:21:29,625 So we're talking at the Heidelberg Law Firm 663 00:21:29,684 --> 00:21:32,085 where you were initially a student. What do 664 00:21:32,085 --> 00:21:34,664 you know now that you wish you'd known 665 00:21:34,964 --> 00:21:37,304 back then? Not about the Heidelberg Law Firm, 666 00:21:37,365 --> 00:21:39,284 specifically, but just about what would have helped 667 00:21:39,284 --> 00:21:40,264 you in your career. 668 00:21:40,724 --> 00:21:41,704 Oh, that's interesting. 669 00:21:43,500 --> 00:21:45,019 Things have changed a lot. Yeah. It's like 670 00:21:45,019 --> 00:21:46,299 the things have changed so much, and there's 671 00:21:46,299 --> 00:21:47,340 been I I I think, 672 00:21:48,619 --> 00:21:50,720 I I was always really excited about, 673 00:21:51,100 --> 00:21:52,080 yeah, like, interdisciplinary 674 00:21:52,539 --> 00:21:53,820 research and that side at the time. And 675 00:21:53,820 --> 00:21:55,784 I think just seeing I don't know how 676 00:21:55,784 --> 00:21:57,744 comfortable I would have been going at like, 677 00:21:57,744 --> 00:21:59,144 you know, now we're doing a virtual reality 678 00:21:59,144 --> 00:22:00,424 work. We're doing the wearable work, which I 679 00:22:00,424 --> 00:22:02,504 had no experience with at that stage. So 680 00:22:02,504 --> 00:22:04,284 I think just kind of knowing 681 00:22:04,585 --> 00:22:06,345 to pay more attention to even, like, some 682 00:22:06,345 --> 00:22:08,024 of the experimental data and some of the 683 00:22:08,024 --> 00:22:09,789 talks on, you know, how do you how 684 00:22:09,950 --> 00:22:11,390 like, you know, what where is where are 685 00:22:11,390 --> 00:22:13,670 the drawbacks of these medical imaging platforms? And 686 00:22:13,670 --> 00:22:14,369 I like, 687 00:22:14,910 --> 00:22:17,069 I I I liked going to talks outside 688 00:22:17,069 --> 00:22:19,069 of my field, but really understanding, like, how 689 00:22:19,069 --> 00:22:20,829 useful it was to kind of embrace those 690 00:22:20,829 --> 00:22:22,464 talks that might not be you might not 691 00:22:22,464 --> 00:22:24,464 immediately see the connection to your research, but 692 00:22:24,464 --> 00:22:26,384 it's worth going to and and can really 693 00:22:26,384 --> 00:22:27,845 spark ideas. I think, 694 00:22:28,625 --> 00:22:30,704 yeah, knowing that would have been helpful. And 695 00:22:30,704 --> 00:22:32,464 even just being more comfortable, like, even like, 696 00:22:32,464 --> 00:22:33,444 the students here, 697 00:22:33,744 --> 00:22:35,825 I like, are so comfortable raising their hand 698 00:22:35,825 --> 00:22:37,940 and asking questions. I think I I was 699 00:22:37,940 --> 00:22:39,859 a lot shyer as the person at that 700 00:22:39,859 --> 00:22:41,460 stage. I don't I don't know if I 701 00:22:41,460 --> 00:22:42,980 I really ever like, I I would not 702 00:22:42,980 --> 00:22:44,340 have been comfortable asking that. So I think 703 00:22:44,340 --> 00:22:46,580 just, you know, knowing it's okay if you 704 00:22:46,580 --> 00:22:48,259 don't know the perfect phrase or how to 705 00:22:48,259 --> 00:22:49,534 do it or, like, you know, you don't 706 00:22:49,534 --> 00:22:50,494 have to think through it. It doesn't have 707 00:22:50,494 --> 00:22:51,775 to be perfect to ask the question. I 708 00:22:51,775 --> 00:22:53,054 think it would have been useful to to 709 00:22:53,054 --> 00:22:53,875 realize them. 710 00:22:54,654 --> 00:22:56,734 And thinking about your job at the moment, 711 00:22:56,734 --> 00:22:58,894 again, this is in the context of giving 712 00:22:58,894 --> 00:23:01,054 careers advice to people who are starting out 713 00:23:01,054 --> 00:23:01,794 in their careers, 714 00:23:02,095 --> 00:23:04,519 what do you like best and least about 715 00:23:04,519 --> 00:23:06,539 your your current work, your current job? 716 00:23:07,799 --> 00:23:09,160 Like, I love that I I get to 717 00:23:09,160 --> 00:23:10,920 work on interdisciplinary work. I get to work 718 00:23:10,920 --> 00:23:13,660 with incredibly smart people. The students are amazing. 719 00:23:14,039 --> 00:23:15,960 We're in an engineering building where the hospital 720 00:23:15,960 --> 00:23:17,714 is across the street, so it is easy 721 00:23:17,714 --> 00:23:19,474 to walk over and ask for help. And, 722 00:23:19,474 --> 00:23:21,714 you know, as a physicist, like, going in 723 00:23:21,714 --> 00:23:22,914 and saying, like, this is what we think 724 00:23:22,914 --> 00:23:24,434 would be really useful and having them be 725 00:23:24,434 --> 00:23:26,195 like sometimes they're like, you know, that's not 726 00:23:26,195 --> 00:23:29,174 implementable, and it's cute, but never gonna happen. 727 00:23:29,795 --> 00:23:31,795 So it's it's kind of fun where you 728 00:23:31,795 --> 00:23:33,329 know? And I think a lot of people 729 00:23:33,329 --> 00:23:34,609 complain about writing grants and things sort of 730 00:23:34,609 --> 00:23:35,410 at the same time. It's like, you know, 731 00:23:35,410 --> 00:23:36,609 by writing the grants and doing this, you 732 00:23:36,609 --> 00:23:37,730 get to spend most of your time kind 733 00:23:37,730 --> 00:23:39,650 of daydreaming about what could we do and 734 00:23:39,650 --> 00:23:41,490 what are you know, dreaming big and thinking 735 00:23:41,490 --> 00:23:43,089 about, you know, high risk projects and then 736 00:23:43,089 --> 00:23:44,535 trying to figure out how to get there. 737 00:23:44,615 --> 00:23:46,455 We now have a really big team, so 738 00:23:46,455 --> 00:23:48,695 it's you know, we have enough people to 739 00:23:48,695 --> 00:23:49,755 really try out. 740 00:23:50,134 --> 00:23:51,734 We have a couple of really complicated problems 741 00:23:51,734 --> 00:23:53,015 that you break down and you have, like, 742 00:23:53,015 --> 00:23:54,695 five different people working on different pieces of 743 00:23:54,695 --> 00:23:55,734 it at the same time, and you can 744 00:23:55,734 --> 00:23:57,414 kinda see how that's all gonna come together 745 00:23:57,414 --> 00:23:59,600 when they're done. And that's it's really fun 746 00:23:59,600 --> 00:24:01,360 to kinda put that puzzle together and see 747 00:24:01,360 --> 00:24:02,640 if we can make progress in each of 748 00:24:02,640 --> 00:24:03,380 these pieces. 749 00:24:04,000 --> 00:24:05,440 I I yeah. I I really like the 750 00:24:05,600 --> 00:24:07,840 just getting to brainstorm the research projects and 751 00:24:07,840 --> 00:24:08,880 get to see where we can go, and 752 00:24:08,880 --> 00:24:10,740 they they kind of afford you that opportunity 753 00:24:10,799 --> 00:24:13,085 there, which is is great. And, like, now 754 00:24:13,085 --> 00:24:14,625 we have a new center where we have 755 00:24:14,924 --> 00:24:16,605 we have a whole section on wearables. We 756 00:24:16,605 --> 00:24:18,285 have another section of virtual reality and one 757 00:24:18,285 --> 00:24:19,964 on AI and one on high performance computing. 758 00:24:19,964 --> 00:24:22,204 We have other researchers and other faculty members 759 00:24:22,204 --> 00:24:24,605 kind of bringing that experience in to all 760 00:24:24,605 --> 00:24:26,365 really help push this idea of the digital 761 00:24:26,365 --> 00:24:27,990 twin and kind of having it be bigger 762 00:24:27,990 --> 00:24:29,990 than my lab is a lot more fun 763 00:24:29,990 --> 00:24:31,210 where it's like, let's bring 764 00:24:31,509 --> 00:24:32,869 in other people who have, you know, whole 765 00:24:32,869 --> 00:24:34,630 new ideas and and can go and, like 766 00:24:34,710 --> 00:24:36,309 and there there's people who are just amazing 767 00:24:36,309 --> 00:24:37,910 on the wearable side that are adding whole 768 00:24:37,910 --> 00:24:39,525 new dimensions you we wouldn't have 769 00:24:40,005 --> 00:24:41,625 expected. So that's been really fun. 770 00:24:42,005 --> 00:24:43,365 I'd say on the the negative side is 771 00:24:43,365 --> 00:24:45,365 it is I joined this and kinda went 772 00:24:45,365 --> 00:24:46,964 into this direction because I really love the 773 00:24:46,964 --> 00:24:48,644 physics. I love the coding. I'm the person 774 00:24:48,644 --> 00:24:50,424 who will stay up all night coding, and 775 00:24:50,644 --> 00:24:52,005 I love it. And this is the and, 776 00:24:52,005 --> 00:24:53,509 like, I do not write lines of code 777 00:24:53,509 --> 00:24:55,190 as much anymore. I know some people today 778 00:24:55,190 --> 00:24:56,890 were saying they still code every day. 779 00:24:57,269 --> 00:24:58,470 I do more of you know, I get 780 00:24:58,470 --> 00:25:00,390 to see code in our in our project 781 00:25:00,390 --> 00:25:00,890 meetings. 782 00:25:01,509 --> 00:25:03,269 So I'm I'm still in touch with the 783 00:25:03,269 --> 00:25:04,710 code, but I'm not the person who gets 784 00:25:04,710 --> 00:25:06,230 to run it. And I think that that 785 00:25:06,230 --> 00:25:07,669 that part is a little, like, a little 786 00:25:07,669 --> 00:25:09,005 bit of a change, and and there are 787 00:25:09,005 --> 00:25:11,745 a lot of meetings. So let's say between 788 00:25:11,805 --> 00:25:13,565 those those two things are the kind of 789 00:25:13,565 --> 00:25:14,065 drawbacks. 790 00:25:14,365 --> 00:25:16,125 I think you're the first person I ever 791 00:25:16,125 --> 00:25:18,125 asked that question. You said they actually like 792 00:25:18,125 --> 00:25:20,285 writing Grant post, but I can totally see 793 00:25:20,285 --> 00:25:22,605 what you described as. How what they mean. 794 00:25:22,605 --> 00:25:23,105 Yeah. 795 00:25:23,940 --> 00:25:25,380 Amanda Randalls, thank you very much. It's been 796 00:25:25,380 --> 00:25:26,980 a pleasure to talk with you. Thanks so 797 00:25:26,980 --> 00:25:27,480 much. 798 00:25:36,295 --> 00:25:39,275 That was Amanda Randalls of Duke University 799 00:25:39,894 --> 00:25:40,714 in conversation 800 00:25:41,015 --> 00:25:43,275 with Physics World's Margaret Harris. 801 00:25:43,815 --> 00:25:45,815 Thanks to both of them for coming on 802 00:25:45,815 --> 00:25:46,555 the podcast. 803 00:25:47,494 --> 00:25:49,595 Physics World will be at the APS 804 00:25:50,210 --> 00:25:51,910 Global Physics Summit, 805 00:25:52,289 --> 00:25:56,230 which will take place on March 806 00:25:56,529 --> 00:25:57,990 in Denver, Colorado, 807 00:25:58,690 --> 00:25:59,829 and online. 808 00:26:00,529 --> 00:26:03,190 At the largest physics meeting in the world, 809 00:26:03,410 --> 00:26:05,589 you can join thousands of physicists, 810 00:26:06,335 --> 00:26:09,615 students, and policy leaders for a week of 811 00:26:09,615 --> 00:26:11,154 connection and collaboration. 812 00:26:12,174 --> 00:26:16,195 Explore cutting edge science shaping our shared future 813 00:26:16,575 --> 00:26:19,474 and be part of the global physics community 814 00:26:20,269 --> 00:26:21,409 driving innovation 815 00:26:21,710 --> 00:26:22,210 forward. 816 00:26:22,909 --> 00:26:27,409 Explore the meeting at summit.aps.org. 817 00:26:27,710 --> 00:26:29,089 And when you're in Denver, 818 00:26:29,470 --> 00:26:32,210 look out for us at IOP Publishing's 819 00:26:32,509 --> 00:26:34,929 booth at the conference exhibition. 820 00:26:36,125 --> 00:26:37,964 I'm afraid that's all the time we have 821 00:26:37,964 --> 00:26:39,264 for this week's podcast. 822 00:26:39,644 --> 00:26:43,085 I'm Hamish Johnston, and our producer is Fred 823 00:26:43,085 --> 00:26:43,585 Iles. 824 00:26:44,204 --> 00:26:47,345 Our theme music is called one three seven, 825 00:26:47,565 --> 00:26:49,505 and it was composed and performed 826 00:26:49,869 --> 00:26:50,690 by the physicist 827 00:26:51,390 --> 00:26:52,609 Philip Moriarty. 828 00:26:53,309 --> 00:26:55,169 We'll be back again next week.