There is a newer version of this record available.

Other Open Access

Towards the Automatization of Cranial Implant Design in Cranioplasty

Jan Egger; Jianning Li; Xiaojun Chen; Ute Schäfer; Gord of Campe; Marcell Krall; Ulrike Zefferer; Christina Gsaxner; Antonio Pepe; Dieter Schmalstieg

Dublin Core Export

<?xml version='1.0' encoding='utf-8'?>
<oai_dc:dc xmlns:dc="" xmlns:oai_dc="" xmlns:xsi="" xsi:schemaLocation="">
  <dc:creator>Jan Egger</dc:creator>
  <dc:creator>Jianning Li</dc:creator>
  <dc:creator>Xiaojun Chen</dc:creator>
  <dc:creator>Ute Schäfer</dc:creator>
  <dc:creator>Gord of Campe</dc:creator>
  <dc:creator>Marcell Krall</dc:creator>
  <dc:creator>Ulrike Zefferer</dc:creator>
  <dc:creator>Christina Gsaxner</dc:creator>
  <dc:creator>Antonio Pepe</dc:creator>
  <dc:creator>Dieter Schmalstieg</dc:creator>
  <dc:description>This is the challenge design document for the "Towards the Automatization of Cranial Implant Design in Cranioplasty" Challenge, accepted for MICCAI 2020.

Cranioplasty is the surgical process where a skull defect, caused in a brain tumor surgery or by trauma, is repaired using a cranial implant, which must fit precisely against the borders of the skull defect as an alternative to the removed cranial bone. The designing of the cranial implant is a challenging task and involves several steps: (1) obtaining the 3D imaging data of the skull with defect from CT or MRI, (2) converting the 3D imaging data into 3D mesh model and (3) creating the 3D model of the implant for 3D printing. The last step usually requires expensive commercial software, which clinical institutions often have limited access to. Researchers have been working on CAD software as alternative to the commercial software for the designing of cranial implant whereas these approaches still involve human interaction, which is time-consuming and requires expertise of the specific medical domain. Therefore, a fast and automatic design of cranial implants is highly desired, which also enables in Operation Room (in OR) manufacturing of the implants for the patient. Centered around the topic, our challenge provides 200 healthy skulls acquired from CT scans in clinical routine and seeks data-driven approaches for the problem. We inject artificial defects into each healthy skull to create training pairs. The datasets are split into a training set and a testing set, each containing 100 healthy skulls and their corresponding skulls with artificial defects. Participants are expected to design algorithms (such as deep learning) based on these training pairs for an automatic cranial defect restoration and implant generation. In this sense, the problem is being formulated as a 3D volumetric shape completion task where a defected skull volume is automatically completed by the algorithm from the participants. The restored defect, which is in fact the implant we want, can be obtained by the subtraction of the defected skull from the completed skull. The implants reconstructed from the skulls with the artificial defects will be quantitatively evaluated using the Dice Similarity Score (DSC) and the Hausdorff Distance (HD).</dc:description>
  <dc:subject>MICCAI Challenges</dc:subject>
  <dc:subject>Biomedical Challenges</dc:subject>
  <dc:subject>Cranial implant</dc:subject>
  <dc:subject>Deep learning</dc:subject>
  <dc:title>Towards the Automatization of Cranial Implant Design in Cranioplasty</dc:title>
All versions This version
Views 1,055717
Downloads 805544
Data volume 1.5 GB1.5 GB
Unique views 899620
Unique downloads 723501


Cite as