<aside> 📌 Diagnosing diseases through speaking in your smartphone and analyzing vocal cord fluctuations .
</aside>
I’m a 16-year-old Activate student in Toronto, Canada; and for those familiar with the area, I’m actually from Brampton but its always easier to say Toronto! Majority of who I am is a compound of years and years of varied experiences which all connect together.
On the surface, I’m an athletic, organized and deep thinker. On the sports side, I was a pre competitive gymnast, played basketball, soccer, flag football, cricket and even badminton. This athletic side has also gotten me out of my house on spontaneous nature walks! You’ll often catch me on my laptop, tidying up my space or breaking down a complex topic.
However, a large portion of who I am has to do with my work because of my passion for it. For the past year and a half being in TKS, I’ve been driven by one key problem: healthcare accessibility around the world. This multi-faceted problem always struck me with confusion, motivation and drive. Being someone who gets sick 3-4 times a year and had quite a few hospital visits as a kid, if I didn’t get the right medicine, at the right time with the right facilities, the outcome of my health could and probably would have been lightyears away from what it currently is. Hence, the motivated and high-intensity work ethic side of me. In fact, these experiences and seeing broken healthcare systems functioning around the world particularly got me interested in using AI for non-invasive diagnosing.
Although, I’ve been driven by the problem of healthcare accessibility for years now and even dove deep into how nanotechnology can be used for the problem for almost a year, I realized to tackle one aspect of the issue I had to use a technology at scale; hence why I indulged on this journey of using AI for diagnosis through voice based systems.
What initially got me interested in the field was hearing a Vinod Khosla talk about health moonshots and hearing this:
Someday you will have better cardiac care in a village in India — which relies on an AI system — than at Stanford — which still relies on the cardiologists they just hired. The cardiologist will be an expert, but the AI will have 25 years of learning and improving every single day. — Vinod Khosla
We require a highly complex AI system to tackle every aspect of high quality care and this might even mean incorporating AGI, however, that’s exactly when this intersection of digital care becomes 10x more interesting. Currently, we’re seeing the rise of telemedicine and remote monitoring. During COVID-19 specifically, every patient would talk to their doctor on the phone rather than going to the clinic. That’s telemedicine.
My vision for the future of this intersection is similar to Vinod’s but extending this further. The problem of healthcare accessibility is different for each country. The hindering factor could be money, supplies, lack of diagnosing, lack of human capacity, lack of education, etc. Out of the five listed, AI can help combat 4/5. Currently, people lack healthcare coverage and majority of this coverage is in doctors visits or repeat check ups, AI can do this. Diagnosing is a booming field for AI to take over. Human capacity and lack of education can directly be solved with the use of AI in clinical settings.
However, if there’s anything I’ve learned in the past three months, its that this process is lengthy and not as simple as we might think. Theoretically speaking, this makes perfect sense. However, privacy concerns, AI safety, access to data, implementation and need are all questions which come to mind. The main insight I’ve gained is to ensure that this is what people need and want. Although cheap diagnosing is a major milestone, imagine getting told you have a disease by your phone? What are the next steps? What if theres a lack of resources to treat the issue? After constantly thinking this, I’ve begun to get in contact with doctors and understand the patients point of view from their experience.
Three months ago, I had worked minimally with AI and had the voice based biomarker (VBB) idea in my head, but worked backwards to try and solve it. So, I hopped on a call with my directors and explained the idea and we formulated my plan to work backgrounds and build this system out in just three months.
As mentioned previously, the output was actually creating the VBB system for Parkinson’s and Huntington’s. However prior to this, I wrote an article on how the baseline for these systems work and then created a contact tracing clustering project for COVID-19. All content below.
During this bumpy journey, one of my main insights was the power of simplicity. During mentorship and feedback calls, I would always go into them with ambitious ideas which were far too complex and would not analyze the risk associated. During these three months, I also went spoke at WebSummit (world’s largest tech conference) in Lisbon, Portugal for one week and lost lots of time to prepping and unwinding from the conference. However, I still wanted to finish every aspect of the project to a high degree with great complexity.