MONTREAL, Feb. 27, 2024 – A team led by Professor Mathieu Dehaes and Dr. Sébastien Perreault, researchers at CHU Sainte-Justine, has just developed a semi-automatic segmentation method for the quick and accurate measurement of neurofibromas, tumours that infiltrate nerves. The technique uses an algorithm that analyzes pixel intensity in magnetic resonance imaging (MRI) scans to detect tumours in various areas of the body identified by the doctor, generating an accurate measurement of tumour volume in just a few minutes. This method, developed by biomedical engineering graduate student Dorsa Sadat Kiaei and published in the journal Heliyon (Cell Press), is a breakthrough in the way we monitor children with neurofibromas. 

Complex tumours that are difficult to measure

The use of algorithms to segment tumours is not new, and many are commonly used in the clinic. However, until now, no automatic algorithmic method has been shown to be effective for neurofibromas. “These are very peculiar tumours of varying shapes and sizes,” explained Mathieu Dehaes, who is also professor of radiology and biomedical engineering at Université de Montréal. “Since they infiltrate the nerves and are distributed in small masses, they can be found throughout the body and in various tissues.” Due to these characteristics, the segmentation and therefore measurement of neurofibromas is highly complex. That’s why radiologists usually proceed manually, measuring only the length of the neurofibromas. This is a quick, but not very accurate way of monitoring the development of complex tumours.

Switching to a much more representative volume assessment is therefore a breakthrough in the monitoring and treatment of neurofibromas. “By using 3D segmentation, we can detect very subtle changes even in a complex tumour,” said a delighted Dr. Sébastien Perreault, pediatric neurologist at CHU Sainte-Justine and assistant clinical professor at Université de Montréal. Moreover, the semi-automatic nature of the analysis, which only requires you to identify the areas to be examined, makes the process very quick. If implemented in therapeutic trials and clinics, it would provide a more accurate assessment of treatment efficacy.

Towards a fully automatic method?

This study is the first step in the development of a fully automatic MRI analysis method based on machine learning. In collaboration with Polytechnique Montréal and with funding from Fondation du Grand défi Pierre Lavoie and Association de Neurofibromatose du Québec, the team is already working on applying to neurofibromas algorithms that have proved effective for segmenting gliomas, another type of tumour that’s simpler in terms of shape and distribution. “Eventually, we’d like to integrate such algorithms into our medical imaging services,” explained Professor Mathieu Dehaes. “That way when a child has an MRI scan for neurofibromas or other tumours, we would have a volumetric assessment right away. This would allow us to adjust treatment without having to wait for the next appointment, thereby optimizing medical care for each child.”