3D Slicer Alpha Training Summary

By Louise van der Werff

Last week I ran my newly developed 3D Slicer training workshop over three content-packed afternoons. Five willing participants were able to come along and get their first peek at the training material, provide very helpful feedback related to the structure of the workshop, and brainstorm ways in which this software may be harnessed in their own work.

For those who may not be familiar with 3D Slicer, it is an open source software package for image visualisation and analysis. More specifically, 3D Slicer is tailored towards the analysis of medical scan data such as that generated via MRI and CT scans. Although 3D Slicer has a wide range of functionalities, this workshop was primarily focussed on generating 3 dimensional volume renderings and surface models of anatomical features from medical scan data.

After giving an introduction on the principles of image processing, I conducted a tour of the 3D Slicer graphical interface, then gave the participants a series of challenges to generate 3D surface models of different anatomical features.

The first step towards generating a 3D model involves image segmentation, which is the process of separating an image into distinct components to make it more meaningful for software to analyse. This is done by assigning each pixel belonging to a particular object a label.

Segmentation of a photo into three distinct components. 

Anatomical structures are segmented from medical scan datasets in 3D Slicer by generating a labelmap over the feature of interest. Anatomical structures we segmented during the workshop included bone, lungs, airways, lateral ventricles, and a trachea and larynx. We explored both manual and automated segmentation methods, their appropriateness being predominantly dependent on the level of contrast between the feature of interest and the surrounding volume.

A particularly challenging case was manually segmenting the trachea and larynx from an MRI scan. Below is a picture of the original scan data, the segmented labelmap, the generated 3D model, and a 3D print of the model to-scale.

An MRI scan of a child’s trachea and larynx. These were manually segmented before a 3D surface model was generated and then 3D printed to scale. 

In addition to segmentation, we also touched upon basic image registration, adding annotations such as fiducials and rulers to a dataset, using statistical tools to calculate volumes of segmented regions, and creating scene views.

The generation of 3D models of anatomical features may be beneficial to many researchers and clinicians, for teaching and training purposes, surgical planning, the creation of custom fit implants and prosthetics, and simple visualisation.

As well as using local installs of 3D Slicer on laptops, we are also currently exploring the effectiveness of running 3D Slicer from the NeCTAR Research Cloud via resbaz.cloud.edu.au (which is powered by the DIT4C engine). A couple of participants tried this approach out with promising results. One major benefit of running graphically intensive software from the cloud is that performance is not limited to the specifications of  local devices, and resources can be scaled as required.

Major points of feedback from this alpha 3D Slicer workshop was that participants preferred more practical content to theory, and wanted to get hands-on with the software as quickly as possible. It was great to see ideas flowing near the end of the workshop from participants about how they might apply 3D Slicers functionalities to their own research projects. Some participants were also interested in exploring whether 3D Slicer could be applied to non-medical applications, such as in the Materials Engineering field.

The alpha training material can be viewed and commented on here.

Please keep an eye out for more training sessions, soon to be announced! If you have any questions, please don’t helistate to contact me at louisevanderwerff@gmail.com or tweet me @LouWerff.

Top