Last updated: 2/24/16 12:46PM

Cranioplasties are used to reconstruct the site of craniotomies and other cranial surgeries that remove sections of the skull. Due to risk of infection after such a procedure, creating a well-fitting prosthetic is important for increasing quality of life and risk management. However, creating a prosthetic that perfectly fits the operative hole either requires the patient to be brought back to the operating room at a second date (two-stage surgery) and or needs to be created manually which can take considerable amount of time (10 - 80 minutes) and be inaccurate. New methods that use an overhead projector to aid manual implant creation exist, but are limited by complexity of implant. Recently a system has been developed for using 3D scanner to create a machined single-stage implant, but the effectiveness of the 3D scanners at completely capturing defect shape and bevel is currently unknown. This project will create ground truth models of cranial defects to test and validate accuracy of the 3D scanning system. During this process we will refine and improve the 3D scanning system from implant capture to patient registration.

• Students: Erica Schwarz, Willis Wang
• Mentor(s): Mehran Armand, Chad Gordon, Ryan Murphy

Cranioplasties are used to reconstruct the site of craniotomies and other cranial surgeries that remove sections of the skull. These cranioplasties are also known as secondary cranial reconstructions and are performed for patients who require staged reconstruction after craniotomies. These craniotomy procedures involve the removal of a section of the skull. The resulting skull flap is often not suited for immediate replacement due to issues such as risk of infection or excess removed material. As a result, these skull flaps are often frozen or thrown away altogether and a cranioplasty is performed instead. The cranioplasty is usually performed to alleviate concerns of safety and protection, cosmetic appearance restoration, and treatment of issues associated with leaving a portion of the skull removed, but can carry its own risks. These procedures are generally performed with an implant made of Poly-Methyl Methacrylate (PMMA) or a titanium mesh.

Due to risk of infection after such a procedure, creating a well-fitting prosthetic is important for increasing quality of life and risk management. Recently, an alternative method which involves the implementation of on-site fabrication of the prosthesis has been effective in cutting down on the number of separate surgeries performed. In this system, surgeons use a Customized Cranial Implant, or CCI, made of PMMA. These CCIs are fabricated preoperatively from patient CT scans and modified through omputer Aided Design. These CCIs are made as an oversized section of the operating area based on information from the CT scan. The main advantage of these CCIs is their ability to conform more closely to the unique curvature of the skull. Specifically, the thickness of the skull is taken into account when making these CCIs whereas a prosthetic made of titanium would be unable to achieve the same precision. During the surgery, the surgeon machines the CCI to match the size and shape of the defect. However, this is labor intensive and can take upwards of an hour. Although this single-staged format is already a significant advancement from previously used multi-staged reconstruction, there is still room for improvement. In an effort to further improve procedure times, Murphy et al. [7] have devised a system which includes a Polaris optical tracker and a laser projection system. This system projects the trace onto the oversized CCI for more accurate cutting and shorter operation times. However, the system is not without its drawbacks. Specifically, it struggles with more complicated geometries and has difficulty collecting points describing the bevel angle of the defect. Additionally, the polaris system itself can be difficult to setup and is very expensive.

As 3D handheld scanners become cheaper and more accurate, the viability of replacing the polaris system with newer technology becomes more feasible. In the previous year, there was a group that built upon the existing system by incorporating a relatively inexpensive 3D scanner in the form of the Structure Sensor (an attachment for the iPad) as a cheaper and more effective alternative to the Polaris system. The project was generally a success, but was limited in that it did not incorporate defect bevels and more complicated geometries and also did not evaluate scan-to-patient registration accuracy. This project proposes to further develop this system with updated segmentation algorithms that allow for more complex feature detection and incorporate defect-to-patient registration in order to put the oversized CCI implant and the scanned defect in the same space (a necessary step for later implant fabrication). Will will do this using ground truth test cases that incorporate a variety of realistic defect geometries.

Our specific aims are:

1. Use patient CT scan to create model of skull and ground truth defects of different shapes and complexities.
2. Create a ground truth model.
3. Use a 3D scanner to make a point cloud representation of the defect site. A point cloud will be converted into a mesh of the defect through the patient’s skull that incorporates defect shape, depth, and bevel.
4. Process the defect mesh’s contours and register it to the defect model.
5. Validate and determine accuracy of mesh to defect fit.
6. Improve segmentation and registration algorithms based on findings.

• Minimum: (3/28/16) Completed
  1. Segment and process point cloud of defect to create defect mesh
  2. Register defect mesh to patient
  3. Register mesh to oversized prosthetic

• Expected: (4/24/16) Completed
  1. Create ground truth models
  2. Validate and improve process accuracy
  3. Quantify accuracy of implant creation
  4. Package process as Slicer module

• Maximum: (5/9/16) Finishing during summer
  1. Test process with cadavers
  2. Register oversized prosthetic to UR5 machine
  3. Define UR5 path for cutting fitted prosthetic

1. Use patient CT scan to create a patient-specific model of skull.

• Convert image stack into binary surface using lab program. 2

2. Create artificial defect using 3D modeling program.

• Subtract defined geometries from skull sections using Solidworks. Incorporate a range of bevels and shapes.
• 3D print defect and area surrounding it.

3. After ground truth defect is made, use a 3D scanner to make a point cloud representation of the defect site.

• Perform segmentation of defect site.
• The point cloud will be converted into a mesh of the defect through the patient’s skull.

4. Process the mesh’s contours and register it to the oversized implant.

• Use a smoothing algorithm to eliminate noise and harsh angles on the interior wall of the defect.
• Find best fit of defect mesh’s surface to surface of patient CT scan model and evaluate accuracy

5. Improve registration and segmentation algorithms based on findings

• Evaluate what cases produces worst and best fit
• Adjust algorithms to be more sensitive to unusual geometries

6. Using a fabrication device, cut the oversized implant into the form of the mesh.

• Mesh representation must be converted into a machining path
• Oversized implant must be registered to machine space

Structure Sensor (Status: Completed)

• Sensor to be used for scanning incision site. Provided by Dr. Armand.

iPad (Status: Completed)

• iPad to use with structure sensor. Provided by Dr. Armand.

Software Repository (Status: Completed)

• Provided by Ryan Murphy. Contains existing lab code, system, and test data. This will also be where we store and document our software modules.

Patient CT Scans (Status: Completed)

• Will be used to create ground truth models. Provided by Ryan.

3D Printer (Status: Completed)

• Needed to fabricate ground truth models. Ryan Murphy will order prints we send him from Wyman.

Operation Observation (Status: On Hold)

• Currently scheduling operation viewing to better motivate understanding of the problem.

UR5 Machine (Status: On Hold)

• Machine for fabricating prosthetic. Not current priority.

1. Milestone name: Create Ground Truth Models
   • Planned Date: 3/14/16
   • Expected Date: 3/27/16
   • Status: Completed
2. Milestone name: Complete Initial Accuracy Evaluation
   • Planned Date: 4/4/16
   • Expected Date: 4/4/16
   • Status: Completed
3. Milestone name: Complete Validation and Improvement
   • Planned Date: 5/2/16
   • Expected Date: 5/2/16
   • Status: Completed

• Project Plan
  • Project plan presentation
  • Project plan proposal
• Project Background Reading
  • See Bibliography below for links.
• Project Checkpoint
  • Checkpoint Presentation
• Paper Seminar Presentations
  • Seminar presentation - Erica Schwarz
  • Critical review - Erica Schwarz
  • Seminar paper 1 - Erica Schwarz
  • Seminar paper 2 - Erica Schwarz
  • Seminar presentation - Willis Wang
  • Critical review - Willis Wang
  • Seminar paper - Willis Wang

• Project Final Presentation
  • PDF of Poster
• Project Final Report
  • Final Report
  • links to any appendices or other material

[1] Aspert, Nicolas, Diego Santa Cruz, and Touradj Ebrahimi. “MESH: measuring errors between surfaces using the Hausdorff distance.” ICME (1). 2002.

[2] Cates JE, Lefohn AE, Whitaker RT. GIST: an interactive, GPU­ based level set segmentation tool for 3D medical images. Med Image Anal. 2004 Sep 8 (3):217­31.

[3] Cignoni, Paolo, Claudio Montani, and Roberto Scopigno. “A comparison of mesh simplification algorithms.” Computers & Graphics 22.1 (1998): 37-54.

[4] Gordon CR, Fisher M, Liauw J, Lina I, Puvanesarajah V, Susarla S, Coon A, Lim M, Quinones­ Hinojosa A, Weingart J, Colby G, Olivi A, Huang J. Multidisciplinary Approach for Improved Outcomes in Secondary Cranial Reconstruction: Introducing the Pericranial­ Onlay Cranioplasty Technique. Neurosurgery. 2014 Jun 10 Suppl 2:179­89.

[5] Herbert M, Pantofaru C. A Comparison of Image Segmentation Algorithms. ​Carnegie Mellon University 2005. The Robotics Institute

[6] Huang GJ, Zhong S, Susarla SM, Swanson EW, Huang J, Gordon CR. Craniofacial Reconstruction with Poly (Methylmethacrylate) Customized Cranial Implants. The Journal of Craniofacial Surgery. 2015 Jan;26(1):64­70.

[7] Murphy RJ, Wolfe KC, Gordon CR, Liacouras PC, Armand M, Grant GT. Computer-assisted Single-stage Cranioplasty. IEEE. Jan 2015.

PreBuilt Environments: \\jhnasrd1.win.ad.jhu.edu\lcsr$\BIGSS\Slicer\Installs-Prebuilt\VS2013_x64

Project Documentation and Test Data: \\jhnasrd1.win.ad.jhu.edu\lcsr$\BIGSS\Cranioplasty

Software Repository: https://svn.lcsr.jhu.edu/bigss

BIGSS Wiki Page: https://intranet.lcsr.jhu.edu/BIGSS