Mapping scents opens the possibility of early disease detection and more
Spotted: With their advanced sense of smell, dogs are able to detect and alert humans to a number of diseases, including cancer, diabetes, and epilepsy. Researchers working to digitise smell may be able to scale that early detection ability and offer new disease screening methods. Right now, human senses such as sight and touch have been digitised and adapted for robotic and other uses, and the team at Osmo is seeking to do the same for smells.
Beginning by building a digital fragrance profile for perfumes, the company’s larger goal is to improve human health and wellbeing through applications of a digitised sense of smell. Led by Alex Wiltschko, entrepreneur in residence at Google Ventures (GV), Osmo is working to read, map, and write the elements that make up a scent. To do this, the company uses machine learning to build Principal Odor Maps (POM) of a variety of smells and their molecules, much like colour is mapped out according to its Red, Green, and Blue (RGB) values.
Smell is difficult to map in part because of the number of different types of receptors used by the human body to detect a scent. With the advent of machine learning, however, mapping something in many dimensions has become possible and is something that the Osmo team has built its work on. The researchers use a kind of machine learning called Graph Neural Networks.
What the team has been finding is that scents that are chemically related are also biologically related, even if they are put together very differently. Now, the researchers are working to identify the components of some of the more common scents in order to replicate them digitally. Work published in Science found that the POM could outperform human panellists in identifying odours, a finding that could help accelerate the use of this technology in healthcare.
Bringing the human sensory experience to robotics is an exciting area of innovation, with innovations in Springwise’s database featuring a number of different ‘skins’ for robots, including one that can sense toxic chemicals through touch and another that is self-healing.
Written By: Keely Khoury