Computer Vision News - February 2020

2 Summary Medic l Imaging 10 As time progressed and demand increased dramatical ly, the entry point lowered and created unique opportunities. The balance of power has changed now. Software can give an edge without being a master of hardware, and thanks to deep learning technologies, there’s no need to spend years ref ining your detection and segmentation algorithm anymore. However, Leo is quick to point out that “there is no free lunch” and the price you pay for such a fast turnaround is a large amount of annotated data. “Companies that are in a position to obtain large amounts of annotated data can turn around an appl ication relatively quickly and put it in the cl inical workf low,” he explains. “In that sense, the game has changed considerably from the industry point of view because smal l players can now enter the f ield and make a contribution. In the last Radiological Society of North America (RSNA) meeting, there were about 170 companies out of 700 or so claiming to have some AI in their product. Most of them were start-ups or smal l companies. You wouldn’t have seen that 10 years ago.” He also thinks a new generation of radiologists has made a di fference. “I grew up driving with paper maps. When I saw the f irst GPS, I was l ike, I don’t real ly need that. I know the streets of Jerusalem. Of course, the continuation of this story is obvious, right? Even experienced taxi drivers can see the value of using Google or Waze now. Things have changed and I ’m enthusiastic. These are great times to be in the f ield of medical image processing and computer aided interventions.” In the last 12 months, things have been moving at a pace, and Leo says we have seen the basic model -based methodology widely replaced. But he warns people “not to throw the baby out with the bathwater ”. In other words, some of the old techniques are sti l l useful and in fact, he is excited that model -based techniques are now being used in combination with deep learning. “It is wel l known that deep learning is very data hungry,” he tel ls us. “there is no free lunch” "These are great times to be in the field of medical image processing and computer aided interventions.”