In this series, we’re going behind the technology to take a look at the people that make Flywheel tick. In this session, we caught up with Mehul Sampat, Flywheel’s Machine Learning Technical Lead. 

Quick facts 

• Medical imaging scientist with 20 years of experience in image analysis and machine learning. 
• Expert in the application of deep learning algorithms for Computer Vision tasks such as image segmentation, image classification and object detection. 
• Expert in traditional image processing methods such image registration, image feature extraction, texture analysis and image segmentation. 
• Published 15 journal papers in top-tier journals and 50+ abstracts at various conferences (2800+ citations).  

From academia to industry 

Mehul’s journey in the world of medical imaging and technology began long before he joined Flywheel in 2022. In graduate school, he studied medical imaging and stayed in the competitive world of academia for a decade, at prestigious universities such as Harvard Medical School, UCSF and Yale.  

“I was mostly interested in medical imaging, so all my schooling and graduate work has been in this area.” Mehul says. 

He then transitioned from academia to the medical imaging industry where he worked at Bioclinica (which is now part of Clario). At Bioclinica, he trained multiple deep learning models for the classification of images from multiple medical imaging modalities. In addition, he designed and implemented various traditional medical image processing algorithms and integrated them in production.  Lastly, he played a crucial role in the timely completion of multiple phase-3 imaging studies for several clients. 

Eventually, he longed for a new challenge, so in 2022, he joined Flywheel. It was a natural fit, allowing Mehul to combine his technical skills with his passion for medical imaging. He was interested in Flywheel’s ability to incorporate AI algorithms and its ability to work with nearly any medical image format. Also having the opportunity to work with and help researchers both in industry and academia was very appealing. 

"The factoid I remember from my first interview was that “Flywheel is data-agnostic” Mehul says. “We don't care if it’s DICOM or X-ray or MRI. Any medical imaging image, any medical imaging format, even JPEG or PNG, we should be able to capture it and do whatever processing you want.” 

Championing machine learning at Flywheel 

At Flywheel, Mehul plays a pivotal role in helping clients take advantage of the platform’s machine learning capabilities. Mehul has been interested in machine learning and its application to imaging since 2005, “before it was sexy,” he says, when he wrote his first paper on the subject.  

One of the things that excites Mehul about Flywheel is the ability to take any algorithm (ML or non-ML), package it into a gear and run it on the appropriate imaging datasets.  

With Flywheel users can run a segmentation algorithm on CT scans and compare it against another or against results generated by humans to see what works best for their use case. While some of the steps are possible with some other platforms, the Flywheel platform is truly exceptional since it captures and logs all aspects of the data journey. “In over 20 years, I have not seen a platform as comprehensive as Flywheel and that is phenomenally exciting to a medical imaging scientist like me”, Mehul says.  

“I always joke that machine learning models may generate some crap, and you wonder, ‘is it any good?’” Mehul says. “With Flywheel, you can look at your data, compare it with your machine learning output and see how it does, and all of that is within the Flywheel ecosystem. The combination of running inference with any Machine Learning algorithm and reviewing the results with Reader Tasks, is an incredible Flywheel superpower.” 

Balancing innovation with personal passions 

Beyond his professional achievements, Sampat enjoys an active lifestyle, with a particular love for sports like soccer, chess and basketball. A resident of Santa Clara, California, Sampat loves to travel to nearby national parks and dreams of visiting each one in the United States, with many adventures already logged in the Rockies and Glacier National Park — although those plans may be on hold, for the time being. 

“My 3-year-old keeps me busy these days,” Sampat laughs. “But when I can, I like to be outdoors and to travel.” 

Flywheel’s medical imaging and AI experts are ready to help with your individual imaging data needs. Whether it’s developing algorithms, testing algorithms on your data or helping you gain visibility into all of your imaging files, we can help you get more from your data. Schedule a call to get started.