Abstract #1147

A Report on the International Workshop on Osteoarthritis Imaging Segmentation Challenge: A Multi-Institute Evaluation on a Standard Dataset

Arjun D. Desai^1,2, Francesco Caliva³, Claudia Iriondo³, Naji Khosravan⁴, Aliasghar Mortazi⁴, Sachin Jambawalikar⁵, Drew Torigian⁶, Jutta Ellerman⁷, Mehmet Akçakaya⁸, Ulas Bagci⁴, Radhika Tibrewala³, Io Flament³, Matt O'Brien³, Sharmila Majumdar³, Mathias Perslev⁹, Akshay Pai⁹, Christian Igel⁹, Erik B. Dam⁹, Sibaji Gaj¹⁰, Mingrui Yang¹⁰, Kunio Nakamura¹⁰, Xiaojuan Li¹⁰, Cem M. Deniz¹¹, Vladimir Juras¹², Ravinder Regatte¹¹, Garry E. Gold², Brian A. Hargreaves², Valentina Pedoia³, and Akshay S. Chaudhari²

¹Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States, ³Radiology, University of California San Francisco, San Francisco, CA, United States, ⁴University of Central Florida, Orlando, FL, United States, ⁵Radiology, Columbia University Medical Center, New York, NY, United States, ⁶Radiology, University of Pennsylvania, Philadelphia, PA, United States, ⁷Radiology, University of Minnesota, Minneapolis, MN, United States, ⁸Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN, United States, ⁹Computer Science, University of Copenhagen, Copenhagen, Sweden, ¹⁰Biomedical Engineering, Cleveland Clinic, Cleveland, OH, United States, ¹¹Radiology, New York University Langone Health, New York, NY, United States, ¹²Biomedical Imaging and Image-Guided Therapy, Medical University of Vienna, Vienna, Austria

Cartilage thickness can be predictive of joint health. However, manual cartilage segmentation is tedious and prone to inter-reader variations. Automated segmentation using deep-learning is promising; yet, heterogeneity in network design and lack of dataset standardization has made it challenging to evaluate the efficacy of different methods. To address this issue, we organized a standardized, multi-institutional challenge for knee cartilage and meniscus segmentation. Results show that CNNs achieve similar performance independent of network architecture and training design and, given the high segmentation accuracy achieved by all models, only a weak correlation between segmentation accuracy metrics and cartilage thickness was observed.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.