The tiff-loader is a NiiVue plugin that converts TIFF bitmap images into NIfTI voxel-based images. It uses the geotiff library to parse TIFF files.
The Tagged Image File Format (TIFF) became popular in miscroscopy due to features including support of high precision (16-bit depth), the ability to store multiple 2D slices in a single file, and the ability to define custom tags that report scanning important parameters. Various tools extend the TIFF format with their own metadata:
- Zeiss LSM (Laser Scanning Microscope) images are based on TIFF but incorporate custom with a proprietary tag with image details. These files also contain thumbnails. This library is able to interpret the custom tag and extract thumbnails.
- OME-TIFF (Open Microscopy Environment) introduces standardized tags to improve compatibility across imaging platforms. Since different software tools define their own TIFF metadata conventions, compatibility can vary, making specialized loaders necessary for correct interpretation.
- ImageJ, a popular image analysis tool, embeds proprietary metadata (e.g., using frames and slices to define 4D datasets). ImageJ provides multiple TIFF loaders to handle different tag variations (ImageJ, OME, LSM).
Since different software tools define their own TIFF metadata conventions, compatibility can vary, making specialized loaders necessary for correct interpretation. The goal of this NiiVue loader is to automatically detect and handle these variations.
Another challenge is that a single TIFF file can contain 2D images of different size and bit-depth. In contrast, NIfTI requires that all slices in a file have identical dimensions. Using ImageJ terminology we refer to all the 2D slices that share dimensions as a stack, and a TIFF file that has multiple stacks as a hyperstack. For example, Zeiss Laser Scanning Microscopes often create TIFF images (with the .LSM extension) that include images in both full resolution as well as reduced resolution thumbnails (illustrated in the ImageJ sample datasets). To handle this, the included loader.js includes two functions: tiff2nii() always returns the first stack in a TIFF image. In contrast, tiff2niiStack() returns one stack (by default the first) and a listing of all the stacks in an image. Through successive calls to tiff2niiStack(), one can sequentially convert all the stacks in a hyperstack to separate NIfTI images. The tiff2nii demo program illustrates this.
To illustrate this library, tiff2nii is a node.js converter that can be run from the command line. The wrapper batch_convert.js allows you to convert all the tiff/tif/lsm files in a folder:
git clone [email protected]:rordenlab/tiff-loader.git
cd tiff-loader
npm install
node ./src/tiff2nii.js ./tests/testData/shapes_deflate.tif
node ./src/batch_convert.js /path/to/tiffs
Note that Python equivalents (tiff2nii.py uses imio; tiff2nii2.py uses tifffile and nibabel). However, the Python converters are unaware of the tags used by ImageJ, LSM and OME. Therefore, these fail to correctly detect and order images based on slice, timing and channels, nor do they provide information about the physical size.
You can also embed this loader into a hot-reloadable NiiVue web page to evaluate integration:
git clone [email protected]:rordenlab/tiff-loader.git
cd tiff-loader
npm install
npm run dev
While TIFF is a popular 2D image format for bitmaps, it is also used by scientific instruments for multi-frame datasets with high precision (e.g. 16-bit scalars).
- ImageJ samples including ImageJ TIFF and LSM (Leica variation of TIFF).
- OME-TIFF sample data.
- Example TIFF images illustrating edge cases such as rare compression schemes.
For scientific applications, we need to preserve the precision of the source data (retaining 8, 16 or 32 bits per channel) and read 4D datasets (with 3D slices and different timepoints or contrasts). This limits the number of suitable libraries. This repository uses geotiff for speed and compatibility.