AI mHolmes Estimates Time Since Death Using Cadaver Microbes

Researchers at Huazhong University of Science and Technology built mHolmes, an AI system that estimates postmortem interval from skin microbiome samples. Trained on daily data from 34 human cadavers over 21 days, it achieves an average error of under two days, versus roughly ±3 days for previous methods.

• mHolmes was trained on microbiome samples from 34 cadavers over 21 days
• Average prediction error is under two days, versus ±3 days for older methods
• The system stays stable even when over half the input data is removed
• Seven bacterial groups were linked to known decomposition stages