Deep Learning for Cranioplasty in Clinical Practice: Going from Synthetic to Real Patient Data

Deep Learning for Cranioplasty in Clinical Practice: Going from Synthetic to Real Patient Data

WoS Article

Correct virtual reconstruction of a de-
fective skull is a prerequisite for successful cranioplasty
and its automatization has the potential for accelerat-
ing and standardizing the clinical workflow. This work
provides a deep learning-based method for the recon-
struction of a skull shape and cranial implant design
on clinical data of patients indicated for cranioplasty.
The method is based on a cascade of multi-branch vol-
umetric CNNs that enables simultaneous training on
two different types of cranioplasty ground-truth data:
the skull patch, which represents the exact shape of the
missing part of the original skull, and which can be eas-
ily created artificially from healthy skulls, and expert-
designed cranial implant shapes that are much harder
to acquire. The proposed method reaches an average
surface distance of the reconstructed skull patches of
0.67 mm on a clinical test set of 75 defective skulls. It
also achieves a 12% reduction of a newly proposed de-
fect border Gaussian curvature error metric, compared
to a baseline model trained on synthetic data only. Ad-
ditionally, it produces directly 3D printable cranial im-
plant shapes with a Dice coefficient 0.88 and a surface
error of 0.65 mm. The outputs of the proposed skull
reconstruction method reach good quality and can be
considered for use in semi- or fully automatic clinical
cranial implant design workflows.

Correct virtual reconstruction of a de-
fective skull is a prerequisite for successful cranioplasty
and its automatization has the potential for accelerat-
ing and standardizing the clinical workflow. This work
provides a deep learning-based method for the recon-
struction of a skull shape and cranial implant design
on clinical data of patients indicated for cranioplasty.
The method is based on a cascade of multi-branch vol-
umetric CNNs that enables simultaneous training on
two different types of cranioplasty ground-truth data:
the skull patch, which represents the exact shape of the
missing part of the original skull, and which can be eas-
ily created artificially from healthy skulls, and expert-
designed cranial implant shapes that are much harder
to acquire. The proposed method reaches an average
surface distance of the reconstructed skull patches of
0.67 mm on a clinical test set of 75 defective skulls. It
also achieves a 12% reduction of a newly proposed de-
fect border Gaussian curvature error metric, compared
to a baseline model trained on synthetic data only. Ad-
ditionally, it produces directly 3D printable cranial im-
plant shapes with a Dice coefficient 0.88 and a surface
error of 0.65 mm. The outputs of the proposed skull
reconstruction method reach good quality and can be
considered for use in semi- or fully automatic clinical
cranial implant design workflows.

Cranioplasty; Skull Reconstruction; Cranial Implant Design; 3D Convolutional Neural
Networks

Cranioplasty; Skull Reconstruction; Cranial Implant Design; 3D Convolutional Neural
Networks

KODYM, O.; ŠPANĚL, M.; HEROUT, A.

2022

01.10.2021

0010-4825

Computers in Biology and Medicine

137

104766

United States of America

1

10

10

https://www.sciencedirect.com/science/article/abs/pii/S0010482521005606?via%3Dihub

cranio_cbm_2021