Towards the Automatization of Cranial Implant Design in Cranioplasty: 2nd MICCAI Challenge on Automatic Cranial Implant Design

Cranioplasty is the surgical process where a skull defect resulting from previous surgery or injury is repaired using an implant that restores the original protective and aesthetic function of the skull. Implications range from decompressive craniectomies to performing brain surgery. Although the patient’s autologous bone is routinely used as the implant material, it may not always be possible due to infection, fracture or bone tumour. Using autologous bone also poses significant risk of requiring a secondary surgery due to bone resorption [1]. Artificial patient-specific implants (PSI) created using computed-assisted design reduce overall patient risks as well as operating time [2]. Developing automatic skull reconstruction methods will increase the availability of PSIs as well as enable their design and subsequent manufacturing directly inside the operating room. Prior to the first AutoImplant Challenge (AutoImplant 2020, held in conjunction with MICCAI 2020 https://autoimplant.grand-challenge.org/), automatic design of cranial implant has been a under-researched area, due to a lack of proper formulation of the problem from a technical perspective. In AutoImplant 2020, we formulated cranial implant design as a volumetric shape completion and 3D shape learning task. Based on the
formulation, various data-driven approaches, such as deep learning and statistical shape model can be employed in solving the problem. AutoImplant 2021 is a substantial extension to the AutoImplant 2020 challenge, where only synthetic defects were used for training and evaluation. In AutoImplant 2021, real clinical defective skulls from craniotomy and skulls with traumatic defects will be used in the evaluation phase, each serving as a separate track (task). The original AutoImplant 2020 task will serve as a third track of AutoImplant 2021. Using real cases for evaluation will guarantee the clinical usability of the winning algorithms.

[1] Göttsche, J., Mende, K.C., Schram, A. et al. Cranial bone flap resorption—pathological features and their implications for clinical treatment. Neurosurg Rev (2020). https://doi.org/10.1007/s10143-020-01417-w

[2] Gilardino, M. S., Karunanayake, M., Al-Humsi, T., Izadpanah, A., Al-Ajmi, H., Marcoux, J., Atkinson, J., & Farmer, J.-P. (2015). A Comparison and Cost Analysis of Cranioplasty Techniques. The Journal of Craniofacial Surgery, 26(1), 113–117. https://doi.org/10.1097/scs.0000000000001305