Computer Vision News - December 2018

10 Computer Vision News Project Management Tip Management Availability of relevant data It is a well-known issue for the project manager in computer vision. This problem is even more serious in medical projects, for many different reasons. Sparseness and limited number of existing cases in the real world; privacy constraints; scarce expert resources for annotation; reticence of those who own data to share it with others. These and other reasons explain project managers’ struggle in finding medical data. It is true that open sources do exist and medical challenges help making more data available for all. Of course, besides checking ownerships and legitimacy of any data, it must be verified if it was already labeled (when needed) and (if yes) whether annotation was done correctly. In many cases there are no labels and we shall explain in a follow-up article how to add them correctly and rapidly. At this point, it is key that the project managers have a clear idea of the data they need and what their software is going to do. A good hint for a direction is to think at what a doctor in the real world would care to look for in a patient. When an expert sees a medical scan and declares that it contains important information, it is generally possible to train a neural network or other system to do the same and reach the same conclusion. If the content is significant, the algorithm may see additional important information. Indeed, we have already seen computerized systems outperform physicians. Of course, it is certainly better when this observer is a medical expert, like a doctor or a radiologist. However, expertise is rare and most times we have to make do with other experts, whose field of expertise overlaps certain aspects of the radiologist’s work. They are biotechnologists, biologists or other professionals who have already done this kind of work in the past. They can be expected to know enough of the terminology and the process. In this way we are able to validate what we are looking for and where it can be found in the scan. When we can't get a doctor or a biologist Collecting and Selecting Data for Medical Projects RSIP Vision’s CEO Ron Soferman has launched a series of lectures to provide a robust yet simple overview of how to ensure that computer vision projects respect goals, budget and deadlines. This month Arik Rond tells us about Collecting and Selecting Data for Medical Projects . It’s another tip by RSIP Vision for Project Management in Computer Vision . If we understand what experts look for in the data, we can use this knowledge to generate input.