TotalSegmentator : Whole Body Segmentation at your Fingertips

This free tool available online can do full body segmentation, it's called TotalSegmentator.

I have already mentioned this tool in a previous edition of the newsletter, but in this edition I would like to cover few more details about it.

The first thing you need to know about this tool is that it is constantly being updated. More anatomical structures are being added every few months.

The second thing you need to know, is that this tool is available in several channels. Here's how to make the most use of it.

If you're a software developer or ML engineer, then you can check the github repo of TotalSegmentator. In this repo, you can look at the code and run it on your local machine or on a remote server if you like. It's basically a pip install and you're ready to go!

If you're a clinician who is used to working with 3D Slicer, then the best option for you would be to install TotalSegmentator extension directly inside 3D Slicer.

If you're neither a developer nor a clinician but you just want to test out the tool quickly, then I would recommend the online website of TotalSegmentator. In this website, you can simply upload your CT scan and click a button to run the automatic whole body segmentation.

Will TotalSegmentator be useful to your startup or company?

My answer is, mostly likely not so much. Here's why.

If you're just looking for a tool to test things out and to get an idea of what's out there, then TotalSegmentator would be a great place to start.

But if you're planning to build your own AI models and to integrate them as part of your startup's or company's product, then I would NOT recommend TotalSegmentator.

The reason for this is, this tool was designed to be used as it is. Meaning that you don't have lots of options if you want to modify it and tailor it to your specific needs.

Moreover, it's a heavy tool! It takes a lot of time to process your scans, and if you choose the fast option, then you will get degraded quality. We've tested this ourselves.

If your company is trying to build a product in the medical imaging field, then you're better off building your own model and tailoring it to your specific needs.

Most likely you will not need to segment every anatomical structure in the body. Also, you may need to segment parts of the body that TotalSegmentator does not currently support.

When you build your own model, you have so much control over what you can do with it. You can add your custom data to train it. You can use open source data with a combination of your proprietary data. You can deploy your model on any platform you'd like and most importantly, you can update your model anytime you'd like to.

Another important point about TotalSegmentator AI models is that they are mostly based on nnUNet framework. So if you can use this tool directly, then you can build a similar model to TotalSegmentator while maintaining a lot of control over your model. We've actually been using this ourselves at PYCAD to build AI model for our clients.

Feel free to reach out to us if you'd like us to help you with your next medical imaging project!

X Post of the Day

Generate entire 3D MRIs from unseen samples generated from a single

2D slice or an X-Ray!

Abdullah Hamdi @Eng_Hemdi 📢We present X-Diffusion a foundation model for 3D MRI generation. For the first time in medical imaging, entire 3D MRI for unseen samples generated from a single 2D slice or even from an X-Ray !! HUGE! joint work with Emmanuelle & Amir from @Oxford_VGG a thread 🧵 1/n https://twitter.com/i/web/status/1786353918550634513 1:15 PM • May 3, 2024 17 Retweets 96 Likes Read 9 replies

​