Simpleware Case Study: AI Automated Segmentation for Custom-Made Implants

Introduction

Founded in 1988, implantcast GmbH specialize in endoprosthesis technology, and provide solutions such as artificial joint replacement and tumor and revision systems. For the last ten years, implantcast have also developed their C-Fit 3D® department for custom-made implants for non-standard anatomies and surgical plans, and deal with around 1,000 cases a year. The team specializes in tumor and revision cases, and exports to over 80 countries. This process starts with CT imaging of patients, followed by creating a virtual bone model and an implant design that can be used as an input to surgery.

One challenge in creating these models is dealing with patient-specific anatomies, particularly in revision cases where bone and metal in a scan are partially indistinguishable. Implantcast uses Simpleware auto segmentation tools to save significant time with AI-based machine learning methods without sacrificing accuracy.

The Simpleware auto-segmentation tools build on the scripting options available in the software to rapidly segment image data. In some cases, the process creates a fully usable model in one click, including options for landmarking. In more complex cases, the AI tool provides a starting point for additional minor revisions. Having worked with the off-the-shelf Simpleware tools for several years to streamline their segmentation tasks, Implantcast is now achieving "1-click"-segmentation in more cases using the most recent updates to the software.

Handling Complex Cases

Sandra Dorn, a Medical Imaging Specialist at implantcast C-Fit 3D®, works with segmentation tasks that would typically require a lot of manual time and effort compared to automating the initial steps of a surgical planning process for custom-made implants. Describing how "auto-segmentation tools are very, very appreciated," Sandra values how the Simpleware tools speed up her workflows.

In complex revision cases, including those with metal artifacts, using thresholding and other manual and semi-automated segmentation tools took approximately 2-3 hours to create a presentable result for designing a custom-made implant. By contrast, running the Simpleware Hip CT tool achieves a segmentation result in a few minutes, with some further work typically needed to account for metal artifacts. Since using the more specialized Hip Revision CT tool, Sandra can achieve “almost perfect segmentation of all important areas” in just 3-4 minutes. Post-processing time can then be approximately 15 minutes, compared to 2-3 hours for carrying out the same tasks manually. In some cases, Sandra has been able to achieve a complete model with just one click of the Hip Revision CT tool, resulting in a “perfect segmentation which doesn’t need any post-processing time. The metal is fully segmented, and we have a perfect virtual bone model. So this is a time saving of 100%.”

Sandra has experienced similar success when working with the Simpleware Spine CT tool for generating virtual models for implant design. Describing how "the workflow couldn’t be easier," Sandra can achieve a segmentation result in 2-3 minutes, with the tool able to work well with extracting small intervertebral joints that would otherwise require extensive manual work prior to using the model in a surgical plan.

Overcoming Bottlenecks with Bone Segmentation

implantcast are also taking advantage of the AI-driven bone filling tool, particularly for the carpal and metacarpal bones in the hand. The tool is designed for general orthopedic segmentation to improve time-consuming workflows. The workflow involves using Simpleware software to apply thresholding to the initial image data before performing a mask flood fill operation. If the mask is not closed, the bone filling tool is used to quickly complete the segmentation, which Sandra describes as "very easy and very, very fast." The entire segmentation process takes 15-20 minutes to generate a virtual bone model of the whole hand and wrist, ready for engineers to continue with surgical planning.

Conclusions

These examples show how implantcast are successfully using Simpleware AI segmentation tools across their workflows to save time, improve accuracy, and maintain consistency in preparing virtual models for surgical planning. Even with challenging revision data, implantcast can achieve reliable results and increase the efficiency of their workflows. For Sandra Dorn, it was "very, very easy to get into the software," with an average time saving for cases of at least 50% up to 100% when "one-click" segmentation is possible.

For implantcast's Mirko Steffen, using Simpleware software means "the whole process of segmentation is completely different", and there is now "no way around AI when it comes to segmentation."

Learn More

• Read more about the Simpleware Auto Segmentation Tools
• Watch implantcast’s presentation and demo in our on-demand webinar Streamline Patient-Specific Implant Design with Simpleware Auto Segmentation Tools