WEBVTT Kind: captions Language: en 00:00:12.920 --> 00:00:20.280 I'm a computer scientist I specialize in medical image analysis and large-scale analysis of biomedical data. 00:00:20.280 --> 00:00:26.580 So we take images usually from CT or MRI scans that you get as a patient in the hospital 00:00:26.600 --> 00:00:33.320 and we try to mine those scans for additional information and use that to derive predictive treatment algorithms 00:00:33.640 --> 00:00:36.560 for patients that are undergoing cancer care. 00:00:44.680 --> 00:00:49.160 I graduated from the School of Computing almost a decade ago and after that I 00:00:49.170 --> 00:00:53.340 went down to the United States worked at Vanderbilt University as a postdoc and 00:00:53.340 --> 00:00:57.180 then I transferred to a position at Memorial sloan-kettering Cancer Center 00:00:57.180 --> 00:01:02.100 in New York which is the oldest and largest and probably best cancer center 00:01:02.100 --> 00:01:07.650 in the world. And I was the only PhD that specialized in image analysis at that 00:01:07.650 --> 00:01:12.750 institution and Memorial sloan-kettering produces a tremendous amount of imaging 00:01:12.750 --> 00:01:17.940 data and so I worked through that data for a few years and really worked with 00:01:17.940 --> 00:01:23.040 clinicians and demonstrated that there was value in that data. 00:01:23.320 --> 00:01:30.540 And when I found out that Queen's is the repository for the cancer clinical trial data for all 00:01:30.540 --> 00:01:35.580 of Canada, and much of the biomedical data for Canada. I realized that that was 00:01:35.600 --> 00:01:39.600 a unique opportunity that we simply don't have in the United States. 00:01:39.660 --> 00:01:49.020 The United States does not centralize any kind of access to patient data. 00:01:49.020 --> 00:01:52.920 So I mean it's a country that can't agree that they should have access to universal healthcare, right? 00:01:52.920 --> 00:01:58.380 Never mind have inter-operable patient data for you know transfer to different institutions. 00:01:58.380 --> 00:02:02.380 So I came back to the School of Computing, I was recruited into the School of Computing 00:02:02.440 --> 00:02:06.360 and into the department of biomedical molecular sciences to take advantage of the data 00:02:06.360 --> 00:02:11.420 that were collecting here and to use that and exploit that for potential cancer clinical trials. 00:02:19.380 --> 00:02:24.550 I'm interested in biomedical data so some of that data is from imaging data 00:02:24.550 --> 00:02:28.810 so those CT or MRI scans that that we take of patients when they come in for 00:02:28.810 --> 00:02:33.250 their care sometimes that data is genomic proteomic metabolomic but what 00:02:33.250 --> 00:02:36.310 I'm really interested in is how we put that data together. So how we put that 00:02:36.310 --> 00:02:42.070 data together interrogate that data to make new biomarkers. A biomarker is a 00:02:42.070 --> 00:02:47.530 thing that a clinician uses to make a decision. So do we send this patient for 00:02:47.530 --> 00:02:54.120 chemotherapy, do we send this patient for a surgery, do we send them for radiation therapy. 00:02:54.380 --> 00:02:58.440 And right now we don't really have a good way to do any of that. 00:02:58.440 --> 00:03:02.380 We treat everyone to treat a few, so if you're a cancer patient for example, you 00:03:02.380 --> 00:03:07.000 come to the hospital, we'll probably give you one line of chemotherapy and then 00:03:07.000 --> 00:03:09.790 when that fails we'll give you another one and then when that fails we'll give 00:03:09.790 --> 00:03:13.300 you another one. And then we might do a genomic test to figure out if there's a 00:03:13.300 --> 00:03:18.250 targeted therapy that we should do for you. And so so I take all of these 00:03:18.250 --> 00:03:21.850 different types of data and use machine learning to put them together into 00:03:21.850 --> 00:03:25.540 predictive and prognostic markers. 00:03:33.100 --> 00:03:37.560 Currently there are no AI and machine learning systems in use clinically. 00:03:37.570 --> 00:03:41.350 So there are no AI are machine learning systems that are used to treat an 00:03:41.350 --> 00:03:45.010 individual patient. That all of the work that's done in AI and machine learning 00:03:45.010 --> 00:03:50.710 has failed to be clinically translated and so in the next 10 years I would like 00:03:50.710 --> 00:03:57.700 to see clinically relevant biologically relevant biomarkers that marry the idea 00:03:57.700 --> 00:04:03.430 of genomics, of radiomics or radiology images, and putting these data together 00:04:03.430 --> 00:04:11.200 to to identify a unified cogent story of cancer that can be mechanistically 00:04:11.200 --> 00:04:15.420 targeted through some through some treatment that we derive. 00:04:15.420 --> 00:04:21.600 What if we can actually, using AI and machine learning tools, come up with a better strategy and 00:04:21.600 --> 00:04:24.080 a better ordering of those treatments? 00:04:24.080 --> 00:04:26.020 And that's something that I would like to see. 00:04:34.460 --> 00:04:38.360 My research has changed over time in a few ways, one is that as a graduate 00:04:38.370 --> 00:04:42.270 student at Queen's, it was much easier to do clinical work that could potentially 00:04:42.270 --> 00:04:47.610 influence the patient care and I thought that was something that you could do 00:04:47.610 --> 00:04:50.520 anywhere because you could do it at Queen's and so then I went down to the 00:04:50.520 --> 00:04:54.720 US and I realized that there are so many barriers to that work and to really 00:04:54.720 --> 00:04:59.280 doing clinical work and you know passing things through the FDA and making 00:04:59.280 --> 00:05:03.800 clinically usable biomarkers that it brought me back to Canada. 00:05:03.800 --> 00:05:09.860 And so now that the equation has changed so coming back to Canada you can really rethink 00:05:09.860 --> 00:05:14.960 how you create these markers and the data with which you have to do that. 00:05:15.200 --> 00:05:18.960 But the other way that it's really changed so I've had an independent lab for four 00:05:18.960 --> 00:05:24.570 years now, so when I started the lab we spent a lot of time annotating data and 00:05:24.570 --> 00:05:29.310 collecting data because data isn't just available through the clinical system 00:05:29.310 --> 00:05:33.540 you have to extract it from the system you have to label it. And as a result of 00:05:33.540 --> 00:05:38.220 that labeling we now sit on a mountain of data that will be really 00:05:38.220 --> 00:05:43.720 interesting to use AI and machine learning tools to interrogate. 00:05:43.720 --> 00:05:47.640 And so while we spend a lot of time doing that initially, I think now is a really 00:05:47.640 --> 00:05:53.220 exciting time to be in the lab because we can now mine that data for clinically 00:05:53.220 --> 00:06:01.860 interesting biomarkers and so it just represents a really unique time to join 00:06:01.860 --> 00:06:07.919 the lab and a really exciting time in AI research. Now we can really benefit 00:06:07.920 --> 00:06:13.500 from all this work that's been happening for the last few years and do some cool stuff. 00:06:21.640 --> 00:06:24.580 That's a great question. I haven't had grad students for a while 00:06:24.590 --> 00:06:29.539 because I've been at a Cancer Center so we didn't have a University to draw 00:06:29.539 --> 00:06:34.849 graduate students from but what I've done since I've come to Queen's to pick 00:06:34.849 --> 00:06:38.320 graduate students is that you know you always look for diversity. 00:06:38.320 --> 00:06:43.780 Diversity of thought, diversity of background to bring a group of people together that think 00:06:43.789 --> 00:06:47.810 differently about problems. You know we're an interdisciplinary group, we work 00:06:47.810 --> 00:06:54.120 with radiologists, pathologists, surgeons oncologists, computer scientists, biostatisticians 00:06:54.140 --> 00:06:59.780 So any background related to anything can really be usable in our group. 00:06:59.780 --> 00:07:03.680 We're interested in enthusiasm you know we want people that 00:07:03.680 --> 00:07:08.000 are interested in curing cancer. We're not here necessarily to write another 00:07:08.000 --> 00:07:13.340 paper although we're gonna do that too we're here to really move the bar on 00:07:13.340 --> 00:07:19.009 cancer care. We're here we have a mission and we are fulfilling that mission and 00:07:19.009 --> 00:07:24.229 we fulfill that by doing these projects but these projects and these thesis are 00:07:24.229 --> 00:07:30.380 a microcosm of the whole system. You know we're here to generate new ideas and to 00:07:30.380 --> 00:07:34.190 really look at cancer in a different way. Students can have a variety of 00:07:34.190 --> 00:07:37.340 background so you know some of the people that we work with are medical 00:07:37.340 --> 00:07:41.419 students or medical residents or fellows but we also work with people that are 00:07:41.419 --> 00:07:45.050 computational students we work with people that are from a basic science 00:07:45.050 --> 00:07:48.880 background that might have some basic science question that they want to ask 00:07:48.880 --> 00:07:53.320 of data that they might use computing to interrogate that solution. 00:07:53.320 --> 00:07:57.940 And we want all of those different types of people and there's no way really, no one way to 00:07:57.949 --> 00:08:02.599 really pick those people. You know I was someone that was a good student but I 00:08:02.599 --> 00:08:08.029 wasn't a stellar student you know and but because of my ability to talk to all 00:08:08.029 --> 00:08:12.349 kinds of different people that's my skill set and so I recognized versions 00:08:12.349 --> 00:08:16.699 of that and other people some people you know want to work by themselves and 00:08:16.699 --> 00:08:20.990 work through a problem and we love those people too and then there's other people 00:08:20.990 --> 00:08:23.659 that don't work that way and it really is just really about 00:08:23.660 --> 00:08:26.120 building a team a common mission. 00:08:34.520 --> 00:08:40.160 Computing is flexibility. So you can get any job once you're a computer scientist you can go 00:08:40.160 --> 00:08:44.300 into any field. You can go into business, you can go into medicine, you can go into 00:08:44.300 --> 00:08:49.760 computing, you can go anywhere. There's a place for a computer scientist in every 00:08:49.760 --> 00:08:54.770 discipline. And computing means that you can make your own path and that you're 00:08:54.770 --> 00:09:00.050 always employable so even when the going gets tough in your job and every job is 00:09:00.050 --> 00:09:04.700 a grind at some point even the best jobs are a grind at some point and when 00:09:04.700 --> 00:09:08.720 you're in the grind and you're questioning your life choices, computing 00:09:08.720 --> 00:09:12.770 gives you the freedom to say "I don't want to be in this grind anymore". I can 00:09:12.770 --> 00:09:17.690 always go and take a job as a programmer in xfield 00:09:17.690 --> 00:09:22.220 or reinvent myself as something else. And I think about that in my career I think 00:09:22.220 --> 00:09:25.430 about this is the thing that I'm doing right now that I want to do for the next 00:09:25.430 --> 00:09:29.060 while but I don't have to do it because I'm a computer scientist so I can go do 00:09:29.060 --> 00:09:32.240 whatever I want to do and that's why you go into Computer science. 00:09:40.860 --> 00:09:44.500 What I like about computing is that you can go into a room where you don't know 00:09:44.500 --> 00:09:48.430 anything about what's going on in that room. So I can go into a room that's full 00:09:48.430 --> 00:09:53.440 of clinical people that understand deeply their discipline and they speak a 00:09:53.440 --> 00:09:58.150 completely different language and I can listen to what they're doing and I can 00:09:58.150 --> 00:10:02.080 think about ways that computing can help them solve a problem that they either 00:10:02.080 --> 00:10:05.260 didn't even know that they could solve or a problem that they didn't even know 00:10:05.260 --> 00:10:09.190 that they had. And that's what's so exciting about computing is that you can 00:10:09.190 --> 00:10:12.780 bring things together that maybe no one has ever thought about before. 00:10:21.460 --> 00:10:25.920 Computing is great and I spend a lot of time in my head in computing. So whenever I can I 00:10:25.930 --> 00:10:29.650 try to get out of my head I try to get outside I try to do a lot of physical 00:10:29.650 --> 00:10:35.710 activity, go to the gym, throw sandbags at the ground to let out some steam because 00:10:35.710 --> 00:10:39.790 you know things it's important to handle stress and to deal with your life and 00:10:39.790 --> 00:10:44.860 it's important for you to be a productive member to take care of yourself. 00:10:44.860 --> 00:10:49.360 And we sit in desks all day so it's important to undo sitting in a desk all day.