UCLA researchers created a deep learning-based approach that may be used to reduce the requirement for human histotechnologists to generate unique stains by computationally converting existing pictures of H&E stained tissue into special stains. This AI-based technique was demonstrated by generating a full panel of special stains used for kidney tissue, namely Periodic acid–Schiff (PAS), Jones silver stain, and Masson’s Trichrome; all of these special stains were computationally transformed from existing images of H&E stained tissue biopsies using specialised deep neural networks.
The researchers used this panel of unique stains in a clinical study to demonstrate the effectiveness of this stain-to-stain transformation approach on a number of clinical samples representing a wide spectrum of kidney disorders. This study, conducted by a multi-institution team of board-certified renal pathologists, discovered a significant improvement in diagnoses made by utilising neural network produced special stains and H&E pictures over using solely H&E images. An further investigation found that the quality of the virtually re-stained pictures is statistically similar to those stained by human specialists histochemically.
A needle core tissue biopsy slice undergoes this stain-to-stain transition in less than one minute. This speed increases the quality of preliminary diagnosis for which specific stains are required, while also saving time and money. These benefits are especially essential for identifying medical diseases such as transplant rejection, when a quick and precise diagnosis allows for prompt treatment, which may result in considerably improved clinical results. Furthermore, because the virtual re-staining approach is applied to existing stains, it is simple to use because it does not necessitate any modifications to the pathology’s present tissue processing workflow.
The research was published in natural communications journal