Artificial intelligence for the interpretation of muscle biopsy images

Quantifying the number and size of muscle fibres remains a complex, repetitive and time-consuming task. It is hampered, in particular, by the low resolution of biopsy images. French engineers and biologists have developed a new artificial intelligence (AI) tool to try to remedy this.

  • MyoSOTHES is a tool derived from the fusion of two algorithms that have appeared on the market, Cellpose and QuPath.
  • It is configured to analyse images with haematoxylin-eosin (HE) staining.
  • The gain over conventional methods of quantification on slices is 31%.

The authors were also able to test it in the context of preclinical therapeutic protocols (sarcoglycanopathy).

 

Artificial intelligence workflow quantifying muscle features on Hematoxylin-Eosin stained sections reveals dystrophic phenotype amelioration upon treatment. Reinbigler M, Cosette J, Guesmia Z et al. Sci Rep. 2022 Nov 19;12(1):19913.