I am a Radiation Oncologist (MD, DNB) attempting to use 3D Slicer for my daily clinical cases (1-2 cases/day). I recently moved from an old, basic Windows laptop to a new MacBook Air M5, but the experience has been very difficult and I am unable to use the software.
The primary issues I am facing:
Security/Verification Issues: Upon installation, macOS could not verify the software. I had to manually go into my system settings and select “Open Anyway” just to get the application to start.
Frequent Hanging: Even for simple tasks like loading a CT scan or scrolling through slices, the software “hangs” and becomes unresponsive. I often see the jumping icon in my dock or the spinning wheel, and I have to force quit the app repeatedly.
TotalSegmentator is Unusable: I am completely unable to get the TotalSegmentator extension to work. Every time I try to run a segmentation—even in “Fast” mode—I get red error messages (like ImportError: cannot import name 'GradScaler').
“Tetris-like” Quality: When I did manage to get one test case to run on my old machine, the contours looked like “bricks” or “Tetris blocks” rather than smooth organs. On this new Mac, I can’t even get that far because the AI simply fails to start.
My Background: I am a medical doctor, not a “techie.” I do not know how to use the Terminal or write code. I purchased this MacBook M5 because I thought it would be faster and more stable for my clinical work, but currently, it is much worse than my old, cheap PC.
My Question: Is there a stable, simple version of Slicer meant for the M5 chip that doesn’t require technical knowledge to fix? I just need a tool that “vomits out” smooth, accurate contours for my patients without the software hanging every few minutes.
I am happy to provide logs if you can tell me where to find them in the menus.
Yes, there is download and use the Slicer 5.10 (stable) version. There are dozens of users on the form using Slicer daily on Mac books (I am using a Macbook pro with M3), and Slicer works perfectly fine without any of those issues you report.
Summary - most of these issues are unique to latest macOS systems and are not problems for Windows machines. The Slicer development for macOS has lagged behind a little bit because there are just fewer developers using macOS to help support these platform specific issues. Reviewing the download statistics in the past year, macOS accounts for roughly 23% of downloads compared to 68% for Windows. Unfortunately there is no firm timeline on when macOS on latest Apple hardware will be optimized in an easily downloadable Slicer installer from the Slicer website.
Regarding 1) Yes this is generally a known issue when running the Slicer Preview because it is not signed and notarized each night. Have you tried using the Slicer 5.10.0 stable instead?
Regarding 2) How much memory does your MacBook Air have and how large is the CT scan that you are loading and analyzing?
Regarding 3) There is an issue with a TotalSegmentator dependency requiring a newer torch version however Slicer on macOS is running an intel built version rather than a native Apple silicon ARM64 based version. The Torch dependency however hasn’t been providing latest versions for Intel built apps running on macOS. Slicer is running an intel version on your system through the Rosetta layer which you might have had to install when you got your new computer.
It does ask me to install rosetta2 initially after warning that slicer isnt verified and can’t open. Then I go to privacy settings and choose open anyway. Then it opens slicer. So why does it say it only has the intel version why not a Mac chip version? Can I try an older version?
Precisely the problem. Slicer runs without issues on my old realmebook slim windows laptop with 8 gigs of RAM and no GPU. I can still do fast mode segmentation. MBA M5 isnt even able to run it without hanging. I ll try the NNunet anyways.
This is because the Slicer organization currently only provides a Slicer Intel based version for macOS. There is development work (with no timeline of completion) still required to be able to create a Slicer version that runs natively on Apple Silicon based hardware. All versions leading up to this have been for Intel Macs. So currently the only option is to run Slicer through the Rosetta2 compatibility layer on your M5 MacBook Air.
So keep in mind that Apple turned its back to Nvidia architecture about a decade ago, and now is lagging behind in the tools and ecosystems in AI related software and applications. The robustness of these tools and the performance you can achieve out of them (nnUnet, totalsegmentator) is nowhere close to what you can do on Linux (or even on Windows) with a decent Nvidia GPU.
If your primary goal is to use these tools easily and without any problem and in a performant way right now, you might need to switch hardware.
