3D Printing of human body parts

Deep learning algorithms for more accurate implants

life sciences and ehealth

3D Printing of human body parts

3D Printing of human body parts

Deep learning algorithms for more accurate implants

This project focuses on challenges faced in medical 3D printing of human body parts. Medical 3D printing requires translating ‘human data’ into ‘virtual data’. This is error-prone and subsequently results in voids in the 3D surface model of the patient. Critical steps in in this proces, even more than the printing process itself, are:

  • Image acquisition
  • Image processing
  • 3D printing

Each of these steps can introduce errors into the fabrication process. This can lead to misfitting 3D printed implants and life-threatening complications during and after surgery. The application of deep learning algorithms may lead to more accurate implants.

Patient image acquisition currently results in voxel-based data (~1GB) representing different tissue types hence grey scales. These grey scales (voxels) have to subsequently be translated into 3D surface models using segmentation and surface rendering algorithms. This process is very compute-intensive and to date requires a wide range of different algorithms and software packages:

  • Pattern recognition, edge detection
  • Segmentation and the Marching Cubes algorithm
  • Computer-aided design (CAD) and reverse engineering
  • Matlab
  • Starting with artificial intelligence and machine learning

The application of deep learning algorithms may lead to more accurate implants. The result of this research could open new avenues for individualized treatments all over the world and will allow surgeries to be performed more accurately, shorten the intervention period, minimize complications and reduce costs.

eScience Coordinator Dr. Elena Ranguelova

Elena is specialized in image processing, and works as an eScience Engineer on medical imaging and other multidisciplinary projects.

Profile page
eScience Research Engineer Wouter Kouw, MSc

Profile page

Sign up for our newsletter