Media, Awards, general
2 min read
E-Mail Facebook Twitter

New Evidence on our award winning MS lesion segmentation algorithm

Peer-reviewed paper about the MICCAI challenge presents an award winning method to detect new lesions in MRI sequences for MS diagnostics.

New Evidence on our award winning MS lesion segmentation algorithm

The MICCAI Challenge offers innovative companies and leading academic groups to test the quality and precision of their deep learning algorithms under objective and standardized conditions. Mediaire applied for the challenge focusing on automatic segmentation and detection of lesions on MRI sequences, comparing our algorithm with international research institutes and leading competitors. 

From 24 with 30 different algorithms mediaire took the first and third place with two different algorithms. The award winning algorithm of mediaire reached the highest F1-score, expressing the detection accuracy by combining recall and precision of the algorithm detecting new lesions in longitudinal MRI sequences. The method even outperformed 3 out of 4 experienced neuroradiologists in detection accuracy (F1 score) and being on par in segmentation accuracy (Dice score).


This is clinically important for diagnosis, prognosis, and monitoring of MS treatment. In addition, this method is highly beneficial to improve diagnostic quality by providing a “second pair of eyes”. Taking into account that the time to segment and detect MS lesions with the algorithm is less than two minutes, the MICCAI challenge proves how mediaire’s deep learning method reliably supports radiologists in their day-to-day practice. This award winning method is implemented in the commercial product mdbrain 4.6 – book your personal product demo here.

If you are interested in learning more about the method and the MICCAI challenge, you can read the published paper in ‘Frontiers in Neuroscience’ for free.