The speaker describes an innovative approach to enhancing surgical procedures using artificial intelligence (AI). They highlight the often-overlooked resource of surgical video data, captured during procedures, which remains mostly unexamined. The integration of AI with medical expertise aims to systematically analyze these videos, offering enriching insights and feedback to surgeons and trainees globally. This promising approach could revolutionize surgical training and consistency, leading to improved patient care worldwide.
AI in surgery transcends the traditional use of chatbots to become an essential tool for medical analysis and research. The speaker emphasizes AI's unique capacity to analyze voluminous data rapidly, a feat unattainable by human practitioners alone. Leveraging AI, there's potential to standardize procedures across varied skill levels and reduce the variability currently inherent in surgical education. Through AI's pattern recognition, surgical procedures get analyzed with precision akin to sports analytics, offering tailored learning opportunities to surgeons across differing expertise levels.
Main takeaways from the video:
Please remember to turn on the CC button to view the subtitles.
Key Vocabularies and Common Phrases:
1. endoscopic [ɛnˈdəskopɪk] - (adjective) - Referring to a procedure that uses an instrument to view the interior of a body cavity or organ. - Synonyms: (minimally invasive, scope-inserted, non-surgical)
Perhaps you or someone you know had a surgery this year, and since you're probably under anesthesia or too busy squeezing someone's hand, you might not realize that during surgery, cameras are used to guide surgeons where they need to go, using endoscopic cameras or to zoom on very specific tiny features, like arteries.
2. anesthesia [ˌænəsˈθiːziə] - (noun) - A state of controlled, temporary loss of sensation or awareness, typically used during surgeries. - Synonyms: (sedation, numbness, sleep)
Perhaps you or someone you know had a surgery this year, and since you're probably under anesthesia or too busy squeezing someone's hand, you might not realize that during surgery, cameras are used to guide surgeons where they need to go, using endoscopic cameras or to zoom on very specific tiny features, like arteries.
3. surgical data science [ˈsɜːrdʒɪkəl ˈdeɪtə ˈsaɪəns] - (noun) - A field focused on leveraging AI and data analytics to enhance surgical outcomes and training. - Synonyms: (AI surgery analytics, AI-enhanced surgery, surgical informatics)
This brings us to surgical data science, a field dedicated to developing AI models and analytics in close collaboration with surgeons.
4. predictive analytics [prɪˈdɪktɪv ænəˈlɪtɪks] - (noun) - A branch of data analytics dedicated to making predictions about future outcomes based on historical data and analytical techniques. - Synonyms: (forecasting, trend analysis, predictive modeling)
predictive analytics
5. terabytes [ˈtɛrəˌbaɪts] - (noun) - A unit of digital information storage that is two to the 40th power bytes (one trillion bytes). - Synonyms: (data unit, storage capacity, memory size)
Using microscopic cameras, this generate thousands of terabytes of surgical video data
6. granular [ˈɡrænjələr] - (adjective) - Made up of or characterized by granules or particles; in this context, detailed and precise in analysis. - Synonyms: (detailed, precise, meticulous)
This granularity also allows for targeted research, which can help answer critical clinical questions.
7. intact [ɪnˈtækt] - (adjective) - Undamaged or unharmed; in a complete or whole state. - Synonyms: (whole, unbroken, undisturbed)
These videos represent an intact gold mine of surgical expertise and insights.
8. revolutionize [ˌrɛvəˈluːʃəˌnaɪz] - (verb) - To effectuate a radical and pervasive change. - Synonyms: (transform, change dramatically, innovate)
Videos with the potential to revolutionize medicine.
9. unattainable [ˌʌnəˈteɪnəbl] - (adjective) - Not able to be reached or achieved. - Synonyms: (inaccessible, unachievable, beyond reach)
AI's unique capacity to analyze voluminous data rapidly, a feat unattainable by human practitioners alone.
10. standardization [stændərədɪˈzeɪʃən] - (noun) - The process of making things conform to a standard. - Synonyms: (uniformity, regularization, normalization)
This lack of standardization in surgery must be addressed
Why You Want AI to Watch Your Surgery - Margaux Masson-Forsythe - TEDxLogan Circle
About a month ago, I was heading into the operating room to observe a surgical procedure. The patient had been experiencing strokes because of low blood flow in their brain. So the goal of the surgery was to connect a scalp artery to a brain artery in order to improve blood flow. This was done by choosing the ideal arteries to connect, opening up the skull and sewing the arteries together. What made this surgery even more impressive was the fact that the patient remained awake throughout the whole procedure. This was crucial because at critical moments, the surgeon needed immediate feedback to know that they were not inducing a stroke when they were operating near the vital arteries. So they would ask the patient to squeeze someone's hand to make sure that everything was all right.
Believe it or not, this is an essential part of my work as an AI engineer. Globally, hundreds of millions of people undergo a surgical procedure every year. Perhaps you or someone you know had a surgery this year, and since you're probably under anesthesia or too busy squeezing someone's hand, you might not realize that during surgery, cameras are used to guide surgeons where they need to go, using endoscopic cameras or to zoom on very specific tiny features, like archeries. Using microscopic cameras, this generate thousands of terabytes of surgical video data. And if you're wondering who's watching these videos after a procedure is over, it's basically no one. Video data packed with information about diverse rare procedures, including successful and thought cases. Videos with the potential to revolutionize medicine. They mostly sit on a shelf in a drive, accumulating dust. But I watch these videos me and AI developed in collaboration with surgeons.
So when I said AI here, this is probably what you're thinking about. In the past couple of years, AI has mostly been represented as chatbots built on top of giant, environmentally costly models. Personally, though, I don't see AI as chatbots. I've been working in AI for almost 10 years now. And at first I started to work in AI because I wanted to use algorithms to analyze scientific data. I wanted to bridge the gaps between computer science and other complex sciences. And mostly what I loved the most was to analyze visual content, like videos and images. So I specialize in computer vision, which is the subfield of AI used to find patterns in visual data, like on this image. And my main focus over the years has been to figure out where my skills could be the most impactful. Meaning where can AI help humankind and actually make a difference?
So I've tried to work on several projects trying to use AI to solve problems. My research took me to places like the tropical forest in Panama, where I was testing computer vision models that measured the diameter of trees for carbon sequestration estimation. So, as you can imagine, when I heard about the thousands of terabytes of unused surgical video data, I was pretty intrigued. These videos represent an intact gold mine of surgical expertise and insights. However, the large volume of these recordings, meaning hours and hours of videos, makes it impossible for humans, even expert surgeons, to watch and analyze them efficiently or effectively. But AI has something that surgeons don't time and the ability to watch hours of videos in minutes. It can also absorb what engineers don't know medical expertise. AI can sort through these vast amounts of data, identify patterns, and can provide feedback and analytics to surgeons and trained residents.
Because surgery is so common, you might imagine a rigid set of written rules for every single procedure. But in many ways, surgery still remains more of an art than a science. Every surgeon is trained by different masters in the field. So the methods and tools a surgeon uses are heavily dependent on the specific teacher in the specific part of the world, in their specific hospital, in the specific field. To make this even more complicated, some surgeries or unique medical circumstances don't happen very often. So an early career surgeon might only have the opportunity to watch or assist in a rare procedure only a couple of times before they have to do it themselves. In some parts of the world, these opportunities shrink even further. This viability can lead to inconsistent and inequitable patient outcomes. I mean, imagine if every pilot had their own unique way of flying a plane. This lack of standardization in surgery must be addressed.
This brings us to surgical data science, a field dedicated to developing AI models and analytics in close collaboration with surgeons. By talking to them, watching them perform their jobs, and understanding their challenges, we can gather valuable data and insights. We're still in the very early days of this innovation. However, the possibilities are already becoming clear. We're moving towards a future where surgical procedures are analyzed by championship sport games.
We are already developing AI models capable of tracking surgical tools with precision. It measures speed, movement and proximity to critical areas, just like tracking asset performance. Let's do an example. These are the analytics from the tennis match Federer vs Murray during the 2012 London Olympics. You can see all the different moves performed by the two players. Likewise, these are the moves performed by a surgeon with each specific tool during a surgery. But here is where it gets really exciting. AI can be used to break down surgeries into distinct phases and actions, creating a playbook for excellence. Envision. A medical resident wants to perfect a specific technique with AI powered analysis. They could search for and study how Leading experts use a particular tool or perform a crucial action. Just like a tennis player reviewing Serena Williams legendary serve. Frame by frame.
This granular access to expertise allows for targeted learning, which has the potential to accelerate skill development globally. This granularity also allows for targeted research, which can help answer critical clinical questions, such as knowing how much time the surgeon spent on removing the tumor, or how many times did they have to stop the bleeding, and how these correlate to patient outcomes. So the ultimate goal? Predicting patient outcomes. Picture an AI system that not only provides feedback on a surgery, but forecasts the potential outcome of a procedure. An AI system that says, based on this technique, at this moment of the surgery, here's what will likely happen. In some cases, this could mean knowing in advance if you will need a second surgery.
Think about that. predictive analytics. Working with surgeons to save lives for patients. This translates to more successful surgeries through improved post surgery care, through better access to surgical experts all around the world, and potentially fewer surgeries. All of this leading to reduced cost. That's why you want AI to watch your surgeries. However, we must be vigilant about bias. AI is only as good as the data we give it, and we need to understand its limitation from the beginning. We are, in the early days, laying the foundation for the future of surgeries. It requires visionaries, technologists, policymakers, medical professionals, working together with ethics and transparency in mind. By integrating AI into surgical practice, we have the potential to unlock the vast reservoir of knowledge stored in those terabytes of unused videos. We can standardize procedures, reduce errors, train more surgeons, encourage collaboration, and ultimately improve patient outcomes.
From my experience, this is one of the most exciting and promising innovation in AI. This is why I go to the operating room. I'm there not just as an observer, but as a partner in innovation focused on using AI to solve impactful problems. Thank you.
Artificial Intelligence, Innovation, Technology, Surgical Data Science, Medical Analytics, Computer Vision, Tedx Talks