Image Database for Article "How to gain valuable insight from scarce data with ML: a post-hoc explanation tool to identify biases in biological images classification", Bolut et al., 2026.

This folder contains all the images used to train the models described in the article "How to gain valuable insight from scarce data with Machine Learning: a post-hoc explanation tool to identify biases in biological images classification", C. Bolut et al., 2026.

The images are in the tif format. They are organized into sub-folders: j03/nalox, j03/ptlip, j10/nalox and j10/ptlip. j03 and j10 correspond to the times elapsed after injury  (3 or 10 days). Nalox (for "naloxone") gathers the images corresponding to mice that underwent the tissue regeneration procedure, ptlip (for "petite lipectomie", small size resection) corresponds to the tissue scarring procedure.

The name of each image provides several information:

• the reparation class: nalox or ptlip.
• the time elapsed injury: j03 (D3) or j10 (D10).
• the number of z stacks imaged: z30 (30 distinct stacks acquired) or ztt (all the tissue-depth).
• the name of the experiment, eg "manip17" for the 17th experiment.
• an identification of the corresponding mouse, signaling a mark made on the mice's ears: either no mark or letters corresponding to which ear was marked (for instance "OG" means "oreille gauche" (left ear)).
• then for the end of the name three nomenclatures were used:
  
  • one of the type "nalox_j03_ztt_manip17_afm1.lsm - Ch1 - C1 Z154 T1.tif" where the "afm" part corresponds to where in the tissue the images were acquired, "Ch1 - C1" (resp "Ch2 - C2") indicates the corresponding channel (1 for fibers and 2 for nuclei), "Z154" is the position of the image in the z-stack column and "T1" is the timepoint (always "T1" because there was only one timepoint).
  • another of the type "nalox_j03_z30_manip19_OD_afm3bis_ch1-0014.tif" where the "afm" part corresponds to where in the tissue the images were acquired, "ch1/2" indicates the corresponding channel and the last number is the position of the image in the z-stack column.
  • another of the type "nalox_j03_ztt_manip17_OD_mos160_tile1_ch1-0028.tif" where "mos160" indicates that 160 tiles were acquired on this tissue, "tile1" is the number of the considered tile (here the first one among 160), "ch1/2" indicates the corresponding channel and the last number is the position of the image in the z-stack column.