AutoImplant II Challenge (MICCAI 2021)

Towards the Automatization of Cranial Implant Design in Cranioplasty II

MICCAI 2021 Challenge

---

📚 Proceedings

The official proceedings can be found on SpringerLink.

#	Title & Link	Code / Repository
01	Personalized Calvarial Reconstruction in Neurosurgery	-
02	Qualitative Criteria for Feasible Cranial Implant Designs	-
03	Segmentation of Defective Skulls from CT Data for Tissue Modelling	-
04	Improving the Automatic Cranial Implant Design in Cranioplasty by Linking Different Datasets	code
05	Learning to Rearrange Voxels in Binary Segmentation Masks for Smooth Manifold Triangulation	code
06	A U-Net Based System for Cranial Implant Design with Pre-processing and Learned Implant Filtering	-
07	Sparse Convolutional Neural Network for Skull Reconstruction	code
08	Cranial Implant Prediction by Learning an Ensemble of Slice-Based Skull Completion Networks	code
09	PCA-Skull: 3D Skull Shape Modelling Using Principal Component Analysis	code
10	Cranial Implant Design Using V-Net Based Region of Interest Reconstruction	-

---

🛠️ Code Snippets for Implant Post-Processing

(Thinning & Border Adjustment)

import nrrd
import pyvista as pv
import numpy as np
import vtk
import trimesh

# Load original mesh
mesh = pv.PolyData('original.stl')

# Extract the interior surface of an implant
mesh.compute_normals(cell_normals=True, point_normals=False, inplace=True)
ids = np.arange(mesh.n_cells)[mesh['Normals'][:,2] < 0.0 ]
top = mesh.extract_cells(ids)

ind = []
newList1 = []
newList2 = []

# Convert PyVista datatype to list for fast processing of points
for i in range(len(top.points)):
    newList1.append([top.points[i][0], top.points[i][1], top.points[i][2]])

for i in range(len(mesh.points)):
    newList2.append([mesh.points[i][0], mesh.points[i][1], mesh.points[i][2]])

for i in range(len(top.points)):
    idx = newList2.index(newList1[i])
    ind.append(idx)

# Rescale x, y, z
for i in range(len(ind)):
    mesh.points[ind[i]][2] = mesh.points[ind[i]][2] * 1.05
    mesh.points[ind[i]][1] = mesh.points[ind[i]][1] * 0.95
    mesh.points[ind[i]][0] = mesh.points[ind[i]][0] * 1.00

mesh.plot()
mesh = mesh.extract_surface().triangulate()
mesh.save('result.stl')

# Note: After generating the processed mesh, intersection can be done using MeshLab:
# MeshLab -> Loading result.stl and original.stl -> Filters -> Remeshing, Simplification, Reconstruction 
# -> Mesh Boolean: Intersection