Data Structure

MassVision allows users to import modality independent MSI data in a structured CSV or hierarchical HDF5 formats for analysis. These format ensures compatibility with any rectilinear MSI data where all pixels share a common list of ions. The following instructions assume the MSI data has MxN pixels and L ions per pixel.

HDF5 file structure

The HDF5 should have two datasets, one for peak intensities named peaks, and one for ions named mz. The content of the datasets are the following Numpy arrays:

1. peaks: A three-dimensional MSI data as Numpy array with shape (MxNxL) (float or integer)

2. mz: A one-dimensional vector data as Numpy array with shape (L,) (float)

CSV file structure

The CSV should be created based on a spectrum-per-row architecture. The CSV file will have M×N+1 rows (one header row and M×N rows for pixel spectra) and L+2 columns (two location indices and L abundance values):

1. Header Row The first row (header) contains:

   • M: Number of pixels in height (integer)

   • N: Number of pixels in width (integer)

   • L m/z values: The m/z values corresponding to L ions (float)

2. Pixel Data Rows Each subsequent row corresponds to a pixel on the M×N grid. Each row contains:

   • i: The pixel’s y-coordinate (integer, range 0 to M-1, 0 being the top)

   • j: The pixel’s x-coordinate (integer, range 0 to N-1, 0 being the left)

   • Ion intensities: The abundance values for the L ions at this pixel (no specific type or range constraints)

Example For a 3x2 pixel MSI data the spatial indexing of the pixels will look like

(0,0)	(0,1)
(1,0)	(1,1)
(2,0)	(2,1)

Assume the data contains 4 ions with m/z of 281.0375, 494.2507, 600.324, and 831.5288, the CSV structure will look like

3	2	281.0375	494.2507	600.324	831.5288
0	0	26	59	9	133
0	1	45	32	0	24
1	0	0	0	77	0
1	1	112	60	18	72
2	0	0	28	38	22
2	1	0	0	0	18

By following this format, users can import MSI data from diverse modalities into MassVision for analysis and visualization.