Computer Vision News - March 2020

2 Summary Challenge 32 or ground glass, which are clear and di ffuse nodules. The LNDb cha l lenge i s made up of four par ts : the main F lei schner c lass i f i cat ion chal lenge and three sub-chal lenges . The idea i s that par t i c ipants have to detect , segment and character i ze nodules in order to ca l cul ate a F lei schner score. Par t i c ipants may choose to take par t in as many of the cha l lenges as they wi sh. This is the f irst time that the Fleischner classi f ication has been done automatical ly at this scale, so it truly is a grand chal lenge, and any result wi l l be a breakthrough. Breaking down some of the numbers involved, Gui lherme tel ls us there have been 229 CT scans, 768 nodules with segmentation and characterization, 58 test scans, 874 participation requests, 45 submissions, and 20 submissions for the test set. In terms of the geographic spl it of participants, The Lung Nodule Database (LNDb) Chal lenge is the chal lenge of this month. We have asked Gui lherme Moreira Aresta to tel l us about it. Gui lherme is a PhD student in the Faculty of Engineering at the University of Porto; he has co- organised this chal lenge alongside João Pedrosa and Carlos Ferreira . Lung cancer is the deadl iest type of cancer worldwide for both men and women. We know that using low-dose CT can help to reduce lung cancer mortal ity , and that computer-aided diagnosis can help physicians and radiologists to further improve early detection. Early diagnosis can make a big di fference to survival rates. Gui lherme tel ls us that Europe fol lows the Fleischner guidel ines for patient fol low-up . Patients are classi f ied according to the number and type of nodules that are found, both in terms of volume and texture e.g. i f they are sol id; partial sol id; From left to right, the LNDb challenge co-organizers: João Pedrosa, Carlos Ferreira and Guilherme Aresta LNDb dataset