FAQ

Biomedisa bases on a semi-automatic segmentation algorithm. That means, before you perform the segmentation process, you must pre-segment some reference slices in a segmentation editor of your choice. The information given by these slices is then used to automatically segment the remaining volume. To start the segmentation process, upload your image data and your pre-segmented slices as Multipage-TIFF or as an Amira file, click the start button, and download or visualize your final segmentation.

To segment the image data, several weighted random walks start in the pre-segmented reference slices and walk through the volume. Over time, the voxels are hit by these random walks. Based on the respective number of hits, it is possible to determine the probability that a voxel belongs to one particular segment. The segmentation can thus be performed by assigning each voxel to the label from which most of the hits originate. You can imagine Biomedisa as an interpolation taking into account the underlying image data.

If you use Biomedisa online, you do not need any special hardware. If you want to run Biomedisa locally, you must have at least one state-of-the-art Nvidia GPU (minimum Kepler architecture). The most important points are the GPU memory and the memory of your system (RAM). How much memory you exactly need depends on the size of the image to be processed. To give you an idea, we tested two of our showcases, which you can download in our gallery. First, we used a workstation with a GeForce 1080 Ti (11 GB GPU memory). The system has 60 GB of RAM. Second, we used our development server with 4x V100 NVLink (each of it has 32 GB of GPU memory) and 800 GB of RAM.

1. The biological screw is a rather small image (419 x 492 x 462 voxels). About 7 GB of RAM and 3 GB of GPU memory were used. Two objects were pre-segmented every 20th slice.

1x GeForce 1080 Ti (11 GB) ⇨ 6 min 36 sec

4x V100 NVLink (each 32 GB) ⇨ 1 min 2 sec

2. Female parasitoid (1077 x 992 x 2553 voxels). Images of this size are widely used on Biomedisa. About 40 GB of RAM and 10 GB of GPU memory were used. Originally, we pre-segmented 56 individual body parts every 10th slice. But for this test, we only used pre-segmentations of every 20th slice to speed up the computation.

1x GeForce 1080 Ti (11 GB) ⇨ 1 h 25 min 34 sec

4x V100 NVLink (each 32 GB) ⇨ 19 min 48 sec

The following three-dimensional data file formats are supported: Multipage TIFF, Amira mesh (AM), MHD, MHA, NRRD and NIfTI (.nii & .nii.gz). In addition, a zipped folder containing two-dimensional slices as DICOM, PNG, or TIFF that represent the volume can also be uploaded. Popular analytic tools like Amira or Fiji/ImageJ can save your data as a TIFF file. The data type can be 8-bit, 16-bit or 32-bit integer or float. You can even upload 64-bit images but they are automatically converted to 32-bit during the calculation.

The label file contains pre-segmented reference slices. It must have exactly the same xyz-dimensions like the uploaded image file. Each gray value or color in this image corresponds to one segment. If you use an Amira label file, all your information, like names etc., will be preserved so that post-processing is easy after importing to Amira again. Attention! Once you label a slice, all segments occuring in this slice must be labeled. Otherwise, these areas are considered as background, and your segmentation will not be correct. Download our showcases and see how the label files were created.

Yes, it is possible to label in different planes. To enable this, please click on the settings icon next to your label file and activate this feature. You can then label in all three planes simultaneously or only in one or two planes of your choice. Attention! There must be at least one empty slice between your pre-segmented slices. The algorithm needs this to detect in which planes you have labeled slices.

Unfortunately, it is not possible to answer this question in general. It depends on several factors. But Biomedisa checks whether your data can be processed or not. If your data is too large, it will automatically complain and stop the process. To give you an idea of how big your data can be. Depending on your labels, Biomedisa will automatically crop your image to a proper region of interest. This region of interest can be roughly 8 GB large (e.g. 2000 x 2000 x 2000 for 8bit images). So even if your volume is much larger, it can be processed as long as your region of interest fits into the bounding box. It also depends on the number of reference slices you have pre-segmented. The more slices were labeled, the smaller the region of interest can be.

Sometimes it is useful to ignore a specific area. Therefore, you can specify a label that should be ignored. No random walks are started from this label. Note that an ignored area is different from an unlabeled area because an unlabled area is considerd as background, and random walks are started for the background.

In addition to the regular diffusion result, you can choose between cleaned (removed outliers), filled holes, post-processed with active contours, uncertainty (a probability map on how uncertain your segmentation result is), and a smoothed version. You can fine-tune these features in the settings .

  • All axis enables pre-segmentation in all orientations (not only xy orientation)
  • Compression results are compressed
  • Human heart enables the adaption for human heart segmentation
  • Normalize (AI) training data is scaled to the same mean and variance
  • Position (AI) input of the network is extended by the coordinates of the voxels
  • Epochs (AI) the number of epochs (how often the network "sees" a patch)
  • Stride size (AI) the stride made to create overlapping patches of size 64x64x64
  • Scale size images are sclaed to this size before training
  • Smooth number of smoothing steps
  • Active contour alpha balance factor between expansion and shrinking
  • Active contour smooth number of smoothing steps after each iteration step
  • Active contour steps number of iteration steps
  • Delete outliers remove islands smaller than this threshold (default 90% of the largest object)
  • Fill holes fill holes smaller than 90% of the label size
  • Ignore label ignore a specific label (no random walks start in this label)
  • Compute only label single label or list of labels (e.g. 1,2,3) calculated exclusively
  • Yes, you can share your data. You can either share it with one or multiple users by entering their user names or you create a password protected download link .

    You can use the move icon to move your data between your projects or from and to your storage.

    Yes, you can have a quick look at your data with the slice viewer or visualize it with our 3D rendering software by clicking on the file icon .

    You can train a neural network by selecting the training image files and the corresponding fully segmented label files. Then press the AI button . When training is complete, select the trained network and one or more images for which you want to predict the segmentation. Finally, press the predict button . Optionally, you can upload your data as a tarball. Here, the names of the label files must exactly match the names of the image files. See Gallery for an example. You can also refine your network (this is especially usefull for large images). After the first network has been trained, you can again select the image data, the label data and now the already trained network and press the AI button. A prediction with refinement is performed by selecting both the first and the second network.

  • Download the latest version Downloads.
  • Run install.sh for an automatic installation on Ubuntu 18.04 and follow the instructions.
  • Alternatively, follow the instructions for manual installation in README.
  • Access the application by entering the IP of the machine in a web browser (e.g. 192.168.176.33).
  • If you do not use an apache server you have to specify a port (e.g. 192.168.176.33:8000).
  • Biomedisa wants to speed up and improve your daily work. Instead of manually labeling each slice of your volume, Biomedisa uses only a few pre-segmented slices to segment the rest of your data. But it is not a model-based segmentation method. That means, if you do not see any structures in your data or you can not see the structures are separated, then Biomedisa is probably not able to see it either.

    Biomedisa will soon be an open source project. Until then you can use it for free online by creating an account. If you use Biomedisa results in your publication please cite the following paper:
    Lösel, P. and Heuveline, V. (2016): Enhancing a diffusion algorithm for 4D image segmentation using local information, Proc. SPIE 9784, Medical Imaging 2016: Image Processing, 97842L, doi: 10.1117/12.2216202.

    Lösel, P. and Heuveline, V. (2016): Enhancing a diffusion algorithm for 4D image segmentation using local information, Proc. SPIE 9784, Medical Imaging 2016: Image Processing, 97842L, doi: 10.1117/12.2216202.