Infrared spectra of murine PDX tissues from paper entitled "DREAMER-S: Deep leaRning-Enabled Attention-based Multiple-instance approaches with Explainable Representations for Spectral-histopathology".

Dataset for DREAMER-S: Deep LeaRning-Enabled Attention-based Multiple-instance approaches with Explainable Representations for Spectral-histopathology This repository contains chemical imaging data referenced in the following GitHub repository (https://github.com/AidanDMeade/DREAMER-S). Download the `.mat` files and place them in the `data/` folder to train a model. Data Acquisition Spectroscopic measurements were performed using a Daylight Spero-QT 340 Quantum Cascade Laser (QCL) infrared microscope operating in transmission mode. Hyperspectral chemical images (HCIs) were acquired at low magnification (0.3 NA) across the wavenumber range of 952–1800 cm⁻¹. Each HCI covered a spatial area of 480 × 480 pixels with a spectral depth of 213 wavenumbers. Spectral Pre-processing Rubber-band baseline correction was applied to each spectrum using the Pybaselines library (version 1.2). A threshold value of 0.1 was then applied to the absorbance intensity of the Amide I peak (approximately 1654 cm⁻¹) to distinguish tissue from background. Any spectrum falling below this threshold was classified as non-tissue and replaced with a zero vector. Finally, vector normalisation was performed on the remaining tissue spectra using the Scikit-Learn library (version 1.7.0). The complete dataset comprises 40 hyperspectral images. Each image contains 213 wavenumbers and 230,400 individual spectra, amounting to more than 49 million data points per image.