E-Staining DermaRepo

Skin ailments have a huge contribution towards global economy. To reduce the global disease burden there is a need to stop skin disease progression. In-time diagnosis plays a crucial role in reducing the disease impact. Among various skin imaging modalities, whole slide imaging is being utilised worldwide to analyse skin-associated disease biomarkers. Whole slide images when acquired after conducting a biopsy, are passed through a chemical process known as staining. The process helps highlight certain disease-associated structures. Staining is a lengthy procedure involving chemical reactions and also requires skilled pathologists. To reduce this overhead and standardise the procedure, the process can be virtualise using AI models. Our data repository is comprised of 87 un-stained whole slide images, paired with chemically H&E-stained image samples. We have utilised these pairs to train a Contrastive GAN model capable of synthesizing stained images when un-stained input is passed.