MicroCT scans of a hybrid poplar leaf dehydrating, with annotated slices for model training

Dataset of a leaf segment of a hybrid poplar (P. maximowiczii x P. nigra ‘Max3’) leaf scanned using microcomputed tomography (microCT) over time as it dehydrates.

Data acquisition methodology

Plants were brought to the TOMCAT tomographic beamline of the Swiss Light Source at the Paul Scherrer Institute (Villigen, Switzerland). Before microCT scanning, a young fully expanded leaf was detached from the plant and a short strip (0.4 x 1.5 cm) was cut between second-order veins. The base of the strip was wrapped in polyimide tape and inserted into a styrofoam block fixed on a sample holder. The strip was immediately scanned by imaging 1801 projections of 100 ms under a beam energy of 21 keV and a magnification of 40x, yielding a final voxel size of 0.1625 µm (field of view: ~416x416x312 µm). The leaf was left to dehydrate in the holder and additional scans were taken 10, 20, 25, and 30 minutes after the initial scan. Scanned projections were reconstructed to a transverse view using both absorption (gridrec; Marone et al. 2012) and phase contrast enhancement (Paganin et al. 2002) reconstruction.

Dataset description

On the reconstructed images a region of interest was identified using a paradermal view (i.e. top to bottom of the leaf) and used to manually align the scans of each time step. Thereafter, all images were cropped to that ROI, ensuring that the same region of the leaf was present in all image stacks.

For all stacks, files start with:
DEHYDRATION_small_Leaf4_time_N_
where N is the time point, with values from 1 to 5 equaling 0, 10, 20, 25, and 30 minutes.

Following this prefix is either GRID (gridrec reconstruction), PAGANIN (phase contrast enhancement reconstruction), or LABELLED (hand labelled slices or ground truth). For GRID and PAGANIN, 8-bit grayscale stacks are provided. The AOI suffix indicates the region of interest.

Stacks have been hand labelled over three orientations (for visual examples of the orientations see Labeled_Sections_order_time1.png and Labeled_Sections_order_time2.png):

CROSS (cross sectional, or transverse, view)

LONGI (longitudinal view: similar to cross sectional view but starting normal to it, i.e. along the depth of the stack starting from the left of the cross-sectional view)

PARADERMAL (top to bottom view: starting at the upper epidermis)

A general idea of the slice range within one LABELLED stack is presented after the orientation, as:
STARTtoENDbyRANGE
The exact position of the labelled slices for each time point can be found in the Labeled_slices_positions.txt file. Note that one-based indexing is used (as in ImageJ), not zero-based indexing (as in e.g. Python).

References

Marone F, Stampanoni M. 2012. Regridding reconstruction algorithm for realtime tomographic imaging. Journal of Synchrotron Radiation 19: 1029–1037.

Paganin D, Mayo SC, Gureyev TE, Miller PR, Wilkins SW. 2002. Simultaneous phase and amplitude extraction from a single defocused image of a homogeneous object. Journal of Microscopy 206: 33–40.