Current Position :homepage  News  News

News

The 146th Youth Academic Salon of CBEIS: Deep learning based Cardiac MRI Segmentation Using Deep learning incorporating CNN Uncertainty and Shape Prior

【Publisher】:生物医学工程与仪器科学学院【Time】:2021-12-07 【Frequency】:54

On the morning of October 9, 2021, the 146th Youth Academic Salon and Postgraduate Research Ability Enhancement Series of CBEIS was held both online and offline as scheduled. Dr. Fumin Guo from the Sunnybrook Institute of the University of Toronto made an academic report to the students.

  

Dr. Guo works as a postdoctoral fellow at the Sunnybrook Institute and the Department of Medical Biophysics at the University of Toronto. He focuses on deep learning methods for lung and cardiac image segmentation, registration, motion correction, and biomarker quantification for image-guided therapeutic interventions. Up to now, Dr. Guo has published more than 30 papers on medical journals including Medical Image Analysis, IEEE TMI, Radiology, etcHe is also the first Chinese researcher who won the John Charles Polanyi Medical Award since 1987.


When applying the deep learning method on the segmentation of cardiac magnetic resonance images (MRI), performance degradation often occurs when using small datasets with sparse annotations, and testing in unknown areas. In this speech, Dr. Guo introduced a new method to solve these key problems, which greatly promote the wider application of deep learning based on U-net++ integration algorithm in heart segmentation task. This kernel cut segmentation model incorporating the shape prior of the left ventricle as well as the CNN uncertainty. Dr.Guo’s team evaluated the performance of their approach and the experimental results showed improvements than previous methods, even with sparse annotations and unknown datasets. These results suggested that their approach provides a way to enhance the applicability of deep learning in research and clinical patient care, confirming it is a well-established global optimization method.

  

Students benefited a lot from Dr. Guo’s report. At the end of the salon, Dr. Guo and students had a lively discussion on related academic issues, the lecture ended in warm applause.