2018 Robotic Scene Segmentation Challenge

• DOI: 10.48550/arxiv.2001.11190
• Publication Date: 2020-01-01
• Authors: Max Allan; Satoshi Kondo; Sebastian Bodenstedt; Stefan Leger; Rahim Kadkhodamoh...; Imanol Luengo; Felix Fuentes; Evangello Flouty; Ahmed Salih Mohammed; Marius Pedersen; Avinash Kori; Varghese Alex; Ganapathy Krishna...; David Rauber; Robert Mendel; Christoph Palm; Sophia Bano; Güinther Saibro; Chi‐Sheng Shih; Hsun-An Chiang; Juntang Zhuang; Junlin Yang; Vladimir Iglovikov; Anton Dobrenkii; Madhu Reddiboina; A. Aravind Reddy; Xingtong Liu; Cong Gao; Mathias Unberath; Mahdi Azizian; Danail Stoyanov; Lena Maier‐Hein; Stefanie Speidel
• Keywords: Data set

In 2015 we began a sub-challenge at the EndoVis workshop MICCAI in Munich using endoscope images of ex-vivo tissue with automatically generated annotations from robot forward kinematics and instrument CAD models. However, limited background variation simple motion rendered dataset uninformative learning about which techniques would be suitable for segmentation real surgery. 2017, same Quebec introduced robotic 10 teams participating challenge to perform binary, articulating parts type da Vinci instruments. This included realistic more complex porcine as was widely addressed modifications on U-Nets other popular CNN architectures. 2018 added complexity by introducing set anatomical objects medical devices segmented classes. To avoid over-complicating challenge, continued data is dramatically simpler than human due lack fatty occluding many organs.