Michelle
Michelle is ARC Medical’s pathogen detection service. It reads one raw spectrum and names the most likely species, with no thresholds to tune.
87.6%combined accuracy on validation spectra
61,599validation spectra tested
295built-in classifiers
2spectrum types: Raman and MALDI-TOF
What it does
- One linear classifier per species; the highest score is returned
- Accepts Raman and MALDI-TOF spectra and can combine both
- All classifiers ship inside the program, with no separate model files
- REST API and built-in web interface for local or LAN use
- New species can be added by retraining
How it works
- Submit a raw spectrum through the API or web interface
- Michelle extracts features for that spectrum type
- Every species classifier scores the spectrum
- The highest-scoring species is returned
Good to know
Stage: Version 0.1. Michelle has no built-in sign-in, so place it behind an authentication proxy before exposing it to a network. Accuracy is lower for some closely related species and for species with few training spectra.
Interested in Michelle?
Write to the Medical team about pilots, partnerships or questions.
Write to us