Multimodal colorectal cancer dataset (FT-IR spectroscopy and Whole Slide Imaging)

This dataset contains paired mid-infrared (MIR) spectroscopic data, collected using Fourier-transform infrared (FT-IR) spectroscopy, and Whole Slide Images (WSIs) from human colorectal cancer tissue microarrays (TMAs). The dataset was generated to support research in computational pathology, multimodal learning, tissue classification, and molecular imaging. Data was collected from the measurements of 9 TMAs each containing 30-48 tissue cores. Each tissue core in a TMA is represented by the following MIR data:- A full MIR hyperspectral image (.txt format) (7744 spectra with X and Y coordinates in a table)- A visible light RGB reference image (.bmp format) taken by the microscope from the MIR measurement area- A helper file to edit the .bmp images with the Perkin Elmer software (.stg format)- Metadata associated with every single MIR measurements (.json format) WSIs are stored for every TMA as:- A folder containing the body of the WSI- A thumbnail allowing for data analysis on the WSI (.mrxs format) Each TMA has its own distribution of normal (NC), primer colorectal carcinoma (CRC) and colorectal liver metastasis (CRLM) tissue cores with the cancer labels, core IDs, respective patient IDs, and data availability (has MIR data, is on a WSI):- Metadata files containing all of the above (.csv format) The tissue core ID layout that is consistent with every TMA was stored as a .pdf image that depicts the thumbnail of a WSI side-by-side with the IDs belonging to each tissue core. To avoid confusing the core IDs when rotating the TMA, four marker tissue cores were added on the two sides of each TMA, making core identification trivial.