Digital Health Interviews: Amelia Okulewicz, “BioCam” — Capsule Endoscopy and Machine Learning to Find Pathologies

It’s our new episode of Digital Health Interviews. Today we have a new guest — Amelia Okulewicz, Machine Learning Specialist at “BioCam” (Wroclaw, Poland). Together we found out what a capsule endoscopy is and what conditions can be diagnosed using capsule endoscopy.

Introduction:

Amelia Okulewicz: currently, she is working in “BioCam” as a Machine Learning Specialist. Her background is in Biomedical Engineering, but she switched and focused on AI Medicine for two years. She graduated in Poland and Australia to specialize in Machine Learning. She liked computer science in healthcare, so she joined a group for Image Recognition. She is a knowledge collector and AI in the healthcare sphere enthusiast. Her article for developers on “Medium” became very popular; it summarized her adventure of standardized data in medical healthcare.

Lots of startup founders in digital health are not technical. They mostly have either a business management background or a medical one. So, before we talked about the startup Amelia is currently working at, we explained to our audience what machine learning and artificial intelligence are in a simple, not very technical way.

Amelia Okulewicz: When we came up with computers and the computation power as developers, we said: “Okay, maybe it can help us to make decisions.” Then we designed systems that weren’t mended but looked like they were necessary. Patients wouldn’t feel safe because it’s a machine; you need some kind of human interface to use it. Today's AI in healthcare is a small box that will help you make better decisions. It wouldn’t be a decision for you, it’s not the replacement of your help, but you decide how to use it. It will speed up the processes you don’t like. It will show you the most concerning parts, but still, you are at the very end of the decision process. We are just a small pipeline.

ML and AI are buzzwords for every investor. Still, the funny thing is that there aren’t too many successful startups and companies using machine learning very effectively for many reasons. “IBM Watson Health” is a perfect example of it: after so many years of great marketing of the system, doctors and clinicians considered the results unsuccessful. Are ML and AI still not ready to become adequate to solve real problems and help medical workers?

Amelia Okulewicz: I worked for some time with IBM as an NLP system provider, so it was a huge problem to integrate ML and AI correctly. From a professional perspective, it’s incredible how it works! The main problem is communication differences; there are still mistakes: we still don’t have a good connection. We work closely with gastroenterology, and it’s not a simple solution to connect that work right. And procedures sometimes translate our needs, like we need specific labels for different pictures. The specialists in the medical sphere would say it could be this or that diagnosis, but it’s not our case. The machine does not see a diagnosis. It considers a pathology or some changes in tissues. We need to put some more focus on how we communicate with staff. I like two areas of AI that are both beneficial. The first direction is data-driven AI. It’s less about the algorithm black box; it’s more about what we put into the box because machine learning is more general than AI. You can say: “Garbage in — garbage out.” If you put the wrong data, even with a godly designed polished algorithm, it still doesn’t work; it will give you the wrong prediction. The second is ethical AI. It concerns medical providers and doctors all around the world. It’s like their job. They also have specific regulations about it. We also need to be aware of AI: when only writing an algorithm and putting it to paper, it would go in another direction. You still have space for ethical AI between it and business, data and healthcare.

The company we spoke about — “BioCam” — designs and manufactures an endoscopic capsule for detecting potential threats. Capsule endoscopy is an examination of the gastrointestinal tract. It is a minimally invasive procedure that can give insight into digestive system health. The analysis uses a disposable, wireless capsule with a digital camera and LED lamp. The patient swallows the pill. Then, it passively passes through the gastrointestinal tract and takes pictures of the small intestine. Meanwhile, the images are transmitted to the patient's recording device on a belt around his waist. Capsule endoscopy usually takes around eight hours to complete. However, the device can be active for up to twelve hours, and the capsule passes through the digestive system within forty-eight hours. During the examination, the patient can carry out daily activities as usual.

Amelia Okulewicz: First, we designed and manufactured the little thing you can swallow and test in the capsule endoscopy. It’s an alternative to classical endoscopy: not invasive, not unpleasant for patients! Many don’t want to go for screening because of it. The second part is machine learning. It’s a more AI-staffed feature for doctors to see through data. We can catch every hot spot with our algorithms, saving a lot of time for doctors. We recommend the areas that can be seen first, and maybe that will be enough. We can see the exact frame where the capsule is now. We are in the pre-phase of clinical trials and ready to be certified as a product. We consulted a lot with the specialists in the field. So we work as doctors intended, not like we want to. In the healthcare field, you need to remember that the probes you get are not of such importance as a thing you cannot miss.

What makes "BioCam" unique, and what advantages are compared to similar solutions?

Amelia Okulewicz: It’s not a technology we invented. We popularize patient-friendly examination of the whole gastrointestinal tract using our capsule endoscopy platform based on AI technology. We believe capsule endoscopy will be our solution's go-to examination of the gastrointestinal tract.

Traditionally, in the final part of every episode, there are some invaluable pieces of advice for startup founders in digital health. But now we have some recommendations for startup founders in digital health who decided to hire an ML specialist: how to build good communication and set the goals and expectations correctly?

Amelia Okulewicz: I’d start from teambuilding. Try to find a person that would fit your team. Some years ago, companies were searching just for ML specialists. They are now exploring a specialist who will work mainly with algorithms regarding specific things they want to design. They also need to think about cloud systems and continuous delivery… I tell you, as such a specialist: If you don’t understand us, ask! We don’t know when we are speaking jargon.

Our previous episode was with Dhruv Agrawal: The Future of Prosthetic Devices is Already Here