Multimodal Biomedical Dataset for Evaluating Registration Methods (Full-size TMA Cores)

The dataset consists of 206 aligned Bright-Field (BF) and Second-Harmonic Generation (SHG) images. All images are of the same size of 2048x2048 pixels.

• The training set contains 40 image pairs.
• The validation set 1 contains 25 image pairs for tuning the training hyperparameters.
• The validation set 2 contains 7 image pairs for tuning the registration method.
• The test set contains 134 image pairs for evaluation.

Example of the naming convention in the test set:

• 1B_A1_BF.tif: image in Bright-Field
• 1B_A1_SHG.tif: image in Second-Harmonic Generation

Example of the naming convention in the train/validation sets:

• 1B_A2_WB.tif: image in Bright-Field
• 1B_A2_MI.tif: image in Second-Harmonic Generation

The dataset was originally produced by the authors of Non-disruptive collagen characterization in clinical histopathology using cross-modality image synthesis (https://www.nature.com/articles/s42003-020-01151-5).