Jianning Li - Towards the Automatization of Cranial Implant Design in Cranioplasty II: Second Challenge, AutoImplant 2021, Held in Conjunction with MICCAI 2021, ...

More information about this subseries at https://link.springer.com/bookseries/7412

This Springer imprint is published by the registered company Springer Nature Switzerland AG

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Different countries can have difference clinical practices of cranioplasty - a surgical procedure to repair cranial defects. For some countries and clinical institutes, a cranial implant is the primary choice for the repairment of cranial defects. However, the design and manufacturing of cranial implants, especially patient-specific implants (PSIs), remains time-consuming and expensive. Hence, current workflows of cranioplasty demand improvements.

The second AutoImplant cranial implant design challenge (AutoImplant 2021, https://autoimplant2021.grand-challenge.org/ ) was organized as a satellite event of the Medical Image Computing and Computer Assisted Interventions (MICCAI 2021) conference, focusing specifically on the clinical usability of the automatic cranial implant design algorithms. Thus, three task tracks were created for AutoImplant 2021; Task 1 and Task 3 focused on the generalization ability of the algorithms on varied synthetic defect patterns, as we observed from the prior edition of the challenge that improving the generalization ability is non-trivial, yet essential for the problem of automatic cranial implant design. Task 2 provided 11 pre-cranioplasty skulls with real defects from the clinical routine for a clinical evaluation of the algorithms. Clinical experts were invited to manually assess the clinical usability of the predictions for Task 2 based on a scoring system ranging from 1 point (not usable) to 5 points (flawless). The conference was held virtually on October 1, 2021, and featured 11 talks, including two invited talks from German and American neurosurgeons who specialize in cranioplasty and cranial implant design.

The challenge proceedings are comprised of 10 papers (818 pages long), including one invited paper from clinical experts about cranioplasty management. A descriptor for the creation of the Task 1 dataset was also provided by the organizing team members. Clinical experts evaluations for the Task 2 submissions were compiled as an independent paper, as we believe the qualitative evaluation criteria used by clinical experts are valuable to the (automatic) cranial implant design community since general quantitative metrics alone are not closely correlative of the actual practical usability of the cranial implants. The challenge papers were reviewed in a single-blind manner and each challenge paper received three to four reviews. It was required that, for the camera-ready version of the accepted papers, the reviewers' comments must be addressed and incorporated. There is one paper that got accepted after a major revision.

We are grateful to the organizing team members, the authors, and the speakers for making our conference a success, and for contributing to the advancement of automatic cranial implant design.