2021/05/12 | Research | Artificial Intelligence

Space technology in the operating theater

The instruments were developed to search for signs of life on other planets. But now they will also be used to distinguish healthy nerve cells from brain tumor cells. A conversation with astrophysicist Brice-Olivier Demory and Raphael Sznitman, an expert in machine learning and artificial intelligence, about their BrainPol project.

Prof. Dr. Brice-Olivier Demory, Center for Space and Habitability (CSH) and NFS PlanetS. (left; © Ramon Lehmann) and Prof. Dr. Raphael Sznitman, Director CAIM and Director ARTORG Center for Biomedical Engineering Research (right; © Tanja Kurt)

As an astrophysicist, Brice-Olivier Demory is primarily concerned with celestial bodies that are light years away from us. With his team at the Center for Space and Habitability at the University of Bern, he analyzes electromagnetic rays that arrive here on Earth after their long journey through space – and provide information about the nature of exoplanet atmospheres and what they are made of.

In general, he says, astronomy still pays too little attention to the polarization of light, often even trying to filter it out of or eliminate it from signals. "But it transmits information," Demory says. With his team, he is developing special instruments known as spectro-polarimeters that will one day enable researchers to detect signs of life even in space.

Operating in the blind

Surprisingly, polarimetric measurements also reveal differences between healthy and abnormal cells. This is probably due to the fact that the proliferating cells of a tumor crowd closely together – losing the optical properties of the tissue surrounding them. "We know that the polarisation of light is affected by the higher vascularisation and larger cell density of tumours,” Demory says. “What we do not yet know is how exactly the tissue structures shape the polarisation signal.”

At a symposium, Demory spoke with Raphael Sznitman, director of the ARTORG Center for Biomedical Engineering Research. Sznitman is an expert in machine learning and artificial intelligence. The two realized they could start a promising project by combining their expertise.

This is how BrainPol came into being, an interdisciplinary research project largely funded by the National Center of Competence in Research PlanetS. The project aims to master a medical challenge that has so far remained unsolved: the clear detection of early forms of glioma, a malignant and fast advancing form of brain tumor. 

The greatest prospects for successful treatment result when the glioma can be detected early and cut out of the brain before it has spread metastases. "The problem is that the human eye cannot detect the early forms of glioma," Sznitman says. "Neurosurgeons have to operate in the blind, so to speak," Demory adds.

New perspectives

"We want to evaluate the polarimetric signals – and provide surgeons with them live during the operation to enhance their natural visual abilities with technology," says Sznitman. He and his team feed the results of the polarimetric measurements of brain biopsies to their AI algorithms, which search for patterns in the mass of data. On the basis of the diagnoses from pathology the system learns which signals match healthy cells and which indicate glioma cells. The project has only just begun, and it remains to be seen whether it can be clinically implemented.