Teenage leaders and entrepreneurs in science

foreign [Music] [Music] s the smallpox vaccine was discovered I'm saving the lives of 5 million Americans in the 1800s Anastasia was discovered which which allowed us to make make surgery less painful in the 1900s the first kidney transplant was conducted and in 2000s we found a cure for Hepatitis T since the beginning of humanity we've seen advances in health care however what we're seeing now is that the growth of computer science and AI integrated with
Healthcare making technology and Healthcare a really big thing hi my name is Sia goel and today I'll be talking a little bit more about me and my research journey in medical technology I'll also be talking a little bit about the barriers to implementation and Healthcare technology and also current advances in health Tech so to talk a little bit about me I'm originally from West Lafayette Indiana I'm studying computer science specifically artificial intelligence math and biology from Stanford University I'm excited about using technology to improve Society specifically in applications related to health care in high school I published my research on medical technology and presented it in a variety of science competitions as well as conferences most notably I was declared a rise Global winner as mentioned rise is a
program that finds promising teens and gives them support so that they can make more of a social impact without rise I wouldn't have been here today so my research Journey started when I was a freshman in high school specifically I was interested in finding if seeing if lauric acid which is found in coconuts could be a treatment in Alzheimer's to do this I performed a variety of experimental techniques and learned two major things the first thing was that research consists of a lot of failures that may or may not lead to success even after reading a lot of literature papers specifically um looking at the a lot of literature papers about experimental protocols I still had to conduct cell cultures multiple times be before my cells grew correctly in addition because of the lack of resources in my lab I had to come up with creative protocols
thus I learned the role of perseverance secondly I was introduced to medicine and Technology this project gave me the opportunity to attend the international science and engineering Fair specifically I saw projects that combined computer science or Ai and biology I was fascinated by AI as I wondered how such a man-made phenomenon could mimic something as complicated as the brain this started my journey into pancreatic cancer diagnostic research the pancreas is an organ located deep behind the stomach and also inside the intestines as a result it's hard to diagnose specifically the early stage diagnosis rate of pancreatic cancer is at nine percent and it's misdiagnosed 33 of the time with other cancers thus the five-year survival rate of pancreatic cancer has stayed at a constant 10 percent whereas we've seen the survival rate of other cancers increased dramatically
in addition current research often achieves an accuracy of around 70 to 80 percent as it does not take into account this need for early diagnosis and the decreased rate of misdiagnosis thus I created an ensemble algorithm to diagnose pancreatic cancer using micrornas micrornas are a type of genetic material that have been linked to the development of cancer you can see here that my algorithms achieved an accuracy of over 90 percent when distinguishing between pancreatic cancer and no pancreatic cancer and also early and late stage pancreatic cancer what's more is when we look at sensitivity or true positive rates and specificity or true negative rates these scores were also above 90 percent showing that my algorithms were not prone to misdiagnosis in addition we can see that this model achieves an accuracy of over 10 percent compared to previous literature and specificity or true negative rates of over 20 compared to past literature
not just this but I created a user interface so that these algorithms could be easily Deployable into Medical Services specifically all Healthcare Providers or doctors have to do is input information about the patient such as their first name last name and email ID then after conducting a blood test using microarrays they can get the micro RNA expression levels of a patient and input it into the user interface my algorithm runs on this data and results are displayed the results that are shown include whether or not the patient has pancreatic cancer and if they have pancreatic cancer what stage they are in as well as the micrornas that are most differentially expressed what's more is that this research has huge applications in the field of personalized medicine specifically not only these are these micro rnas being used to specifically diagnose a person but the micro rnas can also be used for Target treatment thus instead of using a
one-size-fits-all approach personalized treatment can improve outcomes and reduce the size effects we see of a one-size-fits-all approach now in college what I tried to do was Implement my user interface into more of a clinical setting however as Dr Doris Taylor a previous speaker would say I've received a lot of pushback now me and my naive self I thought this would be a really easy process to do because my algorithms were surprisingly performing really well in terms of accuracy however why were there so many barriers to implementation well the barriers that existed were within AI itself so within AI or an algorithm you know the input and you know the output but you don't necessarily know what's happening inside or why the output is being produced this is called a black box and because an AI is a black box a lot of biological
problems or biological research doesn't favor algorithm research it's very important that you know why you're getting an output in biological research so that you know that the reasoning is due to biological fact a lot of computer science research is looking at the explainability of AI I actually addressed explainable AI in my research and Incorporated it however I still received hesitance from hospitals and insurance companies as they belong because of the preconceived notion that AI is a black box another major barrier was the availability of data a lot of data in biological research is kept private do you do privacy concerns from patients as a result um the data as a result a lot of machine learning studies focusing on genomic data has a have a maximum of around two thousand to three thousand samples which is not enough for generalizability not just this but the data that is
available favors the majority or [ __ ] Asians and men and doesn't often favor the minorities such as people of color or woman as a result people are afraid of the deployment of racist and sexist algorithms thus researchers like myself try to make sure that their data is as diverse as possible in addition as a result computer science has tried to make a call for open science however not a lot of progress has been made because of the fact that a lot of patients would like to keep their data private as well as the fact that research want to keep their data private now the last major barrier that I discovered was in terms of reproducibility specifically in healthcare and AI research this is because of faults and generalizability due to the first two barriers I mentioned also even though computer scientists do a lot of training testing and cross validation a lot of times additional testing is needed to make
sure that the algorithms can be Deployable here we can see the reproducibility crisis in practice specifically there was a there was research published on an algorithm that could detect lung cancer at rates above 90 percent however when Harvard Medical School and the University of Michigan school actually tested this they achieved accuracies around 60 to 70 percent so even though we're seeing these current barriers we're also seeing a lot of current developments in the field of healthcare AI so specifically one of the spaces we're seeing these developments is digital health so digital health is a way to distribute health services and information using technology it's grown rapidly during the pandemic specifically to reduce unnecessary exposure of patients to covid-19 and also reduce the burden on hospitals we've seen a lot of different ways there's been a growth in digital Health
specifically in telemedicine or telepharmacy which basically provides Pharmaceuticals to patients virtually teletherapy or online counseling and remote monitoring screening and Diagnostics which allows for online scans for diseases not just this but we've seen a growth in wearable devices specifically as more people are concerned about their health due to the rise of cardiovascular disease obesity and also concerns regarding kova 19 we've seen more wearable devices being developed some of the some of the different types of wearable devices such as a smart apparel apps biometric sensors and also other accessories can help track pre-biological data such as heart rate blood pressure and also breathing rates virtual reality or VR has also been integrated into a medical technology
spaces specifically VR is used by doctors to see how what a patient is actually experiencing and can easily provide physiological therapy to patients in addition VR is being used in Med schools to train doctors and one of the other spaces we're seeing technology being developed is medical robotics medical robotic support medical procedures and tasks specifically in regards to surgical treatment the reason why we've seen an increase in medical robotics is because of the need for a more accurate and non-invasive surgery shorter visitation times and also the need for more doctor availability a lot of different companies are working on this on a specific type of robot called The DaVinci robot The Da Vinci robot is incredibly special because it has more rotation than the capabilities of the human hand which allows for really specific surgery so all in all we have seen that there
have been developments in healthcare technology however there are barriers to implementation today in the future I hope that doctors researchers Engineers lawmakers and also business people can work together to help solve the issues in healthcare implementation and make Healthcare Innovation more robust thank you [Applause] [Music]

About the Speakers