Hamid Tizhoosh Contouring medical images -- deciding what tissue is normal and what is cancerous in an MRI, ultrasound, CT or other scan -- can be a tricky business. First, it’s time consuming. Depending on the body part and type of cancer involved, it can take a doctor anywhere between several minutes and three hours to accurately identify the cancerous tissue in all images from a patient. Second, doctors generally don’t agree with each other. If you give the same scan of a tumor to 10 different doctors, you’ll get 10 different answers about the boundaries of the tumor.

Hamid Tizhoosh, an Associate Professor at the University of Waterloo, is aiming to solve both these problems. In 2008, Tizhoosh founded Segasist Technologies, a company which produces revolutionary contouring software that learns from the doctor using it and provides consensus contours -- suggested contours based on contours by numerous doctors. But, Tizhoosh’s journey to owning a successful business was not a simple one and along the way he learned many valuable lessons about research, entrepreneurship and nitty-gritty of innovation.

Tizhoosh was born in Iran, but when the political turmoil there in the 1980s prompted him to leave, he moved to Germany. He studied electrical and computer engineering and then enjoyed what he calls “a rather regular engineering job.” Then, in the mid-1990s, his grandfather was diagnosed with and died from lung cancer. For Tizhoosh and his family, one of the most frustrating aspects of this experience was waiting to get a treatment plan from his grandfather’s doctors. Not understanding why it was taking so long, Tizhoosh investigated and learned about the problems with medical imaging and contouring that affect so many cancer patients. He changed his academic focus to medical imaging and uprooted his life to Canada to study it.

The challenges his family encountered in the 90s still exist, and are getting worse. According to Tizhoosh, the number of cancer patients will increase by 22 percent between 2010-2020, because of the aging population and other factors. In the same 10 years, the number of oncologists will only increase by 2 percent. Contouring is going to have to get faster and more collaborative to handle the influx of patients.

Tizhoosh realized this and, in his research, worked on creating software that would improve the speed and accuracy of contouring. He quickly realized that while there are many methods to improve speed, improving accuracy is a bigger challenge because there is no way to measure it. How do you know what accuracy is, when doctors don’t agree? Doctors told Tizhoosh that if “they could get multiple doctors in the same room, then they could build consensus contours, a kind of smart average of many contours.” They would be able to develop a measure of accuracy and identify when someone is significantly deviating from that, leading to establishing a baseline for accuracy and quality control. But, as Tizhoosh says “we can hardly afford to have one doctor per patient, so multiple doctors is too expensive. We simply don’t have the resources.”

Medical contouring image The answer was to build software that can learn. Tizhoosh and his colleagues developed a program that tracks a doctor’s contours as he or she works, and hence, is able to predict how that doctor would contour a new image. With this ability, a doctor can load the image data of multiple patients before they leave the office, let them run overnight and inspect the software’s contours in the morning. The software can also store the contours of other doctors on the network and learn from them, providing consensus contours.

Tizhoosh initially developed the software at the University of Waterloo. After receiving grants to support the project, he finished the prototype in 2007. Realizing the commercial potential of the software, he gave the commercialization rights to the University of Waterloo and they assisted him with intellectual property protection and finding his first investor. The newly-formed company moved to office space in the Medical and Related Sciences (MaRS) Discovery District in downtown Toronto. MaRS is a non-profit innovation center, whose mission is to stimulate innovation and accelerate the growth of new Canadian health care businesses. In their current location, “there is a hospital two minutes away in every direction,” says Tizhoosh -- a great place for any medical technology start-up.

Last year, Segasist received US Food and Drug Administration (FDA) approval for their first commercially available software, which is tailored to prostate cancer. Their next product, an improved software package that can be used for any type of cancer, is being released this month and they hope to get FDA approval later this year.

image of prostate cancer medical image contours For his next innovation, Tizhoosh has his head in the cloud. With internet-based storage, Segasist can build software that will learn from doctors all over the planet and make that information available to any single doctor. “Just imagine,” he says, “that you have access to the contouring knowledge of 5,000 oncologists in North America.” He predicts that 10 years from now, there could be well-established quality control for contouring in cancer treatment via radiotherapy. Any doctor will be able to request and receive consensus at anytime based on knowledge of hundreds of experts for a given type of cancer. This knowledge could be used to develop clinical guidelines and standards for contouring.

When asked to reflect on the path from academia to entrepreneurship, Tizhoosh admitted there are many challenges in getting these two groups to work together. “There are many obstacles that consume resources and kill innovations,” he said, “There are many tech people -- engineers, computer scientists -- who are sitting in their offices developing brilliant ideas and nice solutions, but none of them are used in a clinical setting, and there are many doctors who have nice ideas, because they know the problems first hand, but they might not have the necessary technical skills for developing solutions.”

Improving this situation is a challenge, but creating support structures for truly multidisciplinary research would be a good first step, according to Tizhoosh.

Tizhoosh encourages academic researchers to work towards practical implications for their work, but to do it intelligently. “If you have the slightest ambition to change something, the first recommendation is do not do anything before you know who will be the user of the technology. Go and find that user, and find out what he really needs...Billions of dollars of research money gets wasted in spite of the fact that we have many brilliant people working at research institutions and universities. But, they may develop things that the world doesn’t need. The fact that it’s new doesn’t mean anything. Invention is something. Innovation is something else.”