Nima L. Wickramasinghe from the Department of Electronic and Telecommunication Engineering (ENTC), together with his mentor Mohamed Athif from the Department of Biomedical Engineering, Boston University (previously an undergraduate at ENTC) has won the 3rd place in 3 categories (All-lead, 3-lead, 2-lead) in the Physionet/Computing in Cardiology Challenge 2021.
PhysioNet supports challenges, which invite participants from all over the world from various institutes to tackle clinically interesting questions that are either unsolved or not well-solved. In cooperation with the Computing in Cardiology conference, PhysioNet has been co-hosting a challenge annually. This year, the conference was held in Brno, Czech Republic from 12th to 15th of September in a hybrid manner. Computing in Cardiology (formerly Computers in Cardiology) is an international scientific conference that has been held annually since 1974. CinC provides a forum for scientists and professionals from the fields of medicine, physics, engineering, and computer science to discuss their current research in topics pertaining to computing in clinical cardiology and cardiovascular physiology.
This year’s challenge was to identify Cardiac abnormalities (26 scored) given the ECG data of the patients. Usually, the ECG data consists of the signals from the 12-leads. But, this year’s challenge focused on whether the same accuracy can be achieved using reduced-lead ECG data. Using a smaller number of leads would enable low-cost, portable, and user-friendly point of care devices.
The team (team NIMA) proposed a novel solution to tackle this Multi-label classification problem by creating a Deep convolutional neural network that used the time domain and frequency domain of the ECG signals to classify the 26 scored cardiac abnormalities. The results showed that reduced-lead ECG data can obtain almost the same accuracy obtained using all leads.
The team was able to obtain 3rd place in the All-lead, 3-lead, and 2-lead categories competing against 39 International teams. The 2nd place was obtained by team DSAIL_SNU, mainly from Seoul National University. And, the 1st place was obtained by team ISIBrno-AIMT, mainly from the Institute of Scientific Instruments of the Czech Academy of Sciences, Brno, Czech Republic.
Remarks: The collaboration between Nima L. Wickramasinghe and Athif Mohamed was facilitated by the ScholarX program from the Sustainable Education Foundation. ScholarX is a 6-month program for Sri-Lankan undergraduates who would like to get free premium mentoring during their study period.
Link to challenge: https://physionetchallenges.org/2021/Read More
A team of undergraduate students (Team NFPUndercover) from the Department of Electronic and Telecommunication Engineering (ENTC) under the supervision of Dr. Chamira Edussooriya from ENTC, University of Moratuwa, became the 2nd runner up in the IEEE Video and Image Processing Cup (VIP Cup) 2021 competition at the 28th International Conference on Image Processing (ICIP) 2021. ICIP is an annual flagship conference of the IEEE Signal Processing Society which is one of the world’s premier associations for signal processing engineers, academics, and industry professionals. This year the conference was held from the 19th to 22nd of September 2021, virtually in Anchorage, Alaska, USA. The IEEE VIP Cup competition is one of the most prestigious competitions in video and image processing for undergraduate students. This year’s task was to develop a robust algorithm to estimate in-bed human poses under heavy occlusion caused by blankets and varying illumination conditions. This algorithm can be utilized for in-bed human behavior monitoring during sleep or rest state, which is crucial for prognostic, diagnostic, and treatment of many healthcare complications. Team NFPUndercover proposed a novel solution for this task leveraging multiple approaches from both computer vision and signal processing such as extreme occlusion-based data augmentation and label smoothening via knowledge distillation. Specifically, their approach is feasible to implement in a resource-constrained setup. Image Processing and Machine Vision is a branch of Artificial Intelligence that models and learns the image representations to autonomously accomplish downstream tasks such as Image Classification, Object Detection, Semantic Segmentation, etc. It utilizes advanced statistical and mathematical algorithms to make a machine see and understand the real world as humans do. This achievement by UoM places Sri Lanka at a higher position in the world of the image processing communityRead More