Computer Vision News - August 2019

What was it about it that you thought would be technologically interesting? Computer vision has of course been very prevalent in machine learning and has been growing in popularity, as well as the approaches available to make progress in using machine learning techniques to identify any sort of diseases or anomalies. “..for humankind to be able to benefit from this because radiologists can more effectively identify and diagnose this disease.” We thought it would be interesting because we understand that pneumothorax specifically as a disease is very difficult for radiologists to detect. Pneumonia by contrast might be seen as a bit more advanced in that there are methods available to easily and effectively diagnose it, but pneumothorax we understand is a more challenging lung disease for radiologists. Anything we can do to try to advance the state of the art for medical professionals, and to improve their ability to detect these potentially life-threatening diseases, is of course of great interest to us. What are your best expectations from this competition? For our community to be exposed to something that is a very important topic to improve their learning in the area of deep learning for computer vision. I think that the community can advance the state of the art for this particular problem, so I hope that for our host, they can discover the most trending and effective techniques for identifying pneumothorax. Then just broadly for humankind to be able to benefit from this because radiologists can more effectively identify and diagnose this disease. Do you have any advice to the participants that are taking on this challenge? First, I would say there are radiologists and people who have exposure to this field of study who have been pretty vocal on discussion forums and are interested in teaming up. I would say, as much as possible – especially if you’re new to either computer vision or this topic – it can be a benefit to reach out and work with experts in the field to really understand what those diagnostic features are of the disease. “radiologists can more effectively identify and diagnose this disease.” Secondly, because this is a computer vision problem which may require Challenge 16