Computer Vision News - April 2021

4 Research Unsupervised Domain Adaptation from Axial to Short-Axis Multi-Slice Cardiac MR Images by Incorporating Pretrained Task Networks Every month, Computer Vision News selects a research paper to review. This month is a special one: spring has finally arrived and the rise in vaccination rate is giving us much needed hope for the coming summer. Filled with all these positive vibes, I decided to celebrate by choosing to review a paper close to my own research field, called Unsupervised Domain Adaptation from Axial to Short-Axis Multi-Slice Cardiac MR Images by Incorporating Pretrained Task Networks, written by several authors located mainly in the German centre for Cardiovascular Research (Heidelberg). by Marica Muffoletto @maricaS8 We are indebted to all of them (Sven Koehler, Tarique Hussain, Zach Blair, Tyler Huffaker, Florian Ritzmann, Animesh Tandon,Thomas Pickardt, Samir Sarikouch, Heiner Latus, Gerald Greil, Ivo Wolf, Sandy Engelhardt) for allowing us to use their images to illustrate this review and especially to Sven and Sandy who provided us with extra beautiful illustrations. I hope this review makes you as enthusiastic about their work as I am after reading their paper, which can be found here. This paper addresses a common problem in medical imaging: the presence of distribution shifts among imaging scans. Of particular relevance, but not very much discussed, is the problem of scans acquired across different axes. In particular, Cardiac Magnetic Resonance, which offers a non-invasive non-radiating option to image the heart, is vastly used to assess measurements of left ventricular and right ventricular volumes for diagnosis and treatment of several cardiomyopathies. This is generally acquired in short axis (SAX) view as