Carrefour de l'actualité

You are here

Algorithms that diagnose early-stage tumours

Professor Christopher Pal, Department of Computer and Software Engineering

By Catherine Florès
February 15, 2016 - Source : INNOVATIO

Are medical imaging machines more accurate than physicians in detecting a tumour?

Are medical imaging machines more accurate than physicians in detecting a tumour? “When the data consist of medical imaging, programs can actually spot anomalies that would escape human notice,” responds Professor Christopher Pal, an artificial intelligence expert.

The difference is found in learning algorithms, which enable programs to interpret existing, but also new, data on the basis of previous analysis, each time refining the accuracy of results. Professor Pal uses these machine-learning approaches to develop digital medical imaging analysis programs that can identify brain tumours with pixel-level precision.

This project is a collaboration with Professor Yoshua Bengio of the Department of Computer Science and Operations Research at Université de Montréal and will make it possible to diagnose tumours at a very early stage, as well as check on their progression.

The venture uses increasingly powerful graphic cards available on the market that were developed by the burgeoning videogaming industry. Data storage is cloud-computing based. This makes Professor Pal’s team’s technology relatively inexpensive, an asset for marketing prospects.

Suggested Reading

NEWS |  February 22, 2018
Deep neural networks identifies tumours  |  Read
NEWS |  February 19, 2013
NSERC grants three Polytechnique research teams more than $1 million as part of the 2012 Strategic Project Grants and Strategic Network Grants programs  |  Read