Title: Slow and unstable performance


(Lucas Formighieri) #1

Operating system: Win10-Pro (Portuguese), Intel 5820k, 32gig ram, nvidia 970gtx, nvme ssd.
Slicer version: 4.10 and nightly builds
Expected behavior: fluid workflow
Actual behavior: interruptions and delays in the workflow

Hello,

I am a radiologist working with angiotomography exams (up to 2k slices each). I have experience with dicom programs such as Carestream Vue, Vitrea, GE-ADW, Osirix and Radiant. My focus now is segmentation and 3D printing.

During tasks such as loading exams and segmentation there are such sings as the “not responding” on the title bars on the “whitening” of the program´s window. Loading exams my take up to one minute. Sometimes the program “hangs” during a task and then just quits. Watching the resources usage CPU and GPU are always below 20%. Memory may reach peaks of 50 or 60% but usually hovers around 30%.

Even in systems more powerful than my own (Xeon 5118 Gold, 32 gig ram, nvdia Quadro) it is slow and unstable, although a little bit less so.

Is there anything I can do to improve this?

Thanks.


(Andras Lasso) #2

Typically on a regular desktop computer (few ten GB of RAM, few-year-old i7 CPU) can handle typical images - size of up to 512^3, 16 bits per pixel. If you have a 2000^3 image (about 60x larger) then you’ll need 60x more memory space (about 150GB, preferably physical RAM) and 60x faster CPU to get similar experience as with typical images.

DICOM image review workstations that load and display only a few slices of the image at a time, so for them it does not matter how large the entire image is. 3D workstations either refuse to load extremely large images or downscale them when loaded and may load full-resolution representations for smaller regions when needed.

In Slicer, you can crop&downscale the images using “Crop volume” module. Increase “Spacing scale” parameter or decrease the cropping region size until every component of the cropped volume dimensions (shown in Volume information section) gets below 512. After that, click “Apply” to generate the image and delete the original image from the scene. With this new image, all Slicer functions should work reliably. Before you start, configure at least 150GB virtual memory in your Windows system settings to make sure you don’t run out of memory.