Good news! A paper written by members of our group has been accepted in CVPR2020. The work is called “Learning nanoscale motion patterns of vesicles in living cells” and presents a new application for Machine Learning and super-resolution in biology. The main authors are Arif Ahmed and Ida Opstad, who received the support from several departments at UiT:
Physics and Technology: Balpreet Singh Ahluwalia and Krishna Agarwal
Faculty of Health Sciences: Åsa Birna Birgisdottir
Clinical Medicine: Truls Myrmel
Computer Science: Dilip K. Prasad
As a context, when ranked by h5-index CVPR (240) occupies the Top 10 in a list led by Nature (368), The New England Journal of Medicine (352) and Science (338).
By using Machine Learning, the authors were able of classifying the motion of vesicles at nanoscales. Due to its size, the resolution at which we can observe vesicles in motion is limited by the diffraction of light, which makes the study of them quite difficult for the human eye. This is where super-resolution techniques can help. In particular, researchers applied MUSICAL to the resolution-limited images of the moving vesicles. This technique created by Krishna Agarwal allows to reconstruct super-resolved videos from a stack of fluorescence images. The authors applied then MUSICAL in order to increase the lateral resolution, followed by a classifier based on a Convolutional Neural Network. As a result, they managed to classify the nanomotion of single vesicles into 5 categories with 82% of accuracy!
Congratulations to Arif, Ida, and the whole team!