Prof. Dr Shaikh Anowarul Fattah, SMIEEE, Fellow IEB
Dept. of Electrical and Electronics Engineering, Bangladesh University of Engineering and Technology (BUET), Bangladesh
Topics: Multi-perspective Deep Learning for Biosignal Analysis
Brief Biography: Dr. Shaikh Fattah received Ph.D. degree in ECE from Concordia University, Canada and later he was a visiting Postdoc at Princeton University, NJ, USA. He received B.Sc. and M.Sc. degrees from BUET, Bangladesh, where he is currently serving as a Professor, Department of EEE. His research interests include signal processing, machine learning, and biomedical engineering. He published more than 215 international journal/conference papers and delivered more than 80 Keynote/invited talks in many countries. Dr. Fattah is the Chair of IEEE SPS Bangladesh Chapter (BDC) and Vice-Chair of IEEE RAS, EMBC, PES, and SSIT BDCs. He was the Vice-Chair of IAS BDC. He served as the IEEE Bangladesh Section Chair during 2015-2016. He is the Chair of IEEE PES-HAC, member of IEEE Public Visibility Committee, IEEE PES Long Range Planning Committee, and IEEE Smart Village Education Committee. He served IEEE HAC (2018-20 Education Chair), various committees of IEEE R10, IEEE EAB, and IEEE SIGHT. He served key positions in many international conferences, such as the General Chair of IEEE R10-HTC2017 and TPC Chair of IEEE TENSYMP2020. Dr. Fattah received several awards, e.g. Concordia University’s Distinguished Doctoral Dissertation Prize (ENS, 2009), 2007 URSI Canadian Young Scientist Award, Dr. Rashid Gold Medal (in MSc), BAS-TWAS Young Scientists Prize (2014), 2016 IEEE MGA Achievement Award, 2017 IEEE R10 HTA Outstanding Volunteer Award and 2018 IEEE R10 Outstanding Volunteer Award. He is the Editorial Board Member of IEEE Access, IEEE potentials and Biomed Research International, Editor of JEE, IEB and Editor-in-Chief of IEEE PES Enews. He is a Senior Member of IEEE and a Fellow of IEB.
Abstract: A major challenge in dealing with a bio-signal is to extract distinguishable features for automatic disease detection. Conventional machine learning-based methods rely on extracting hand-crafted features or developing empirical models to classify the bio-signals. On the contrary, deep learning-based techniques can efficiently solve these problems in numerous multi-dimensional applications. In this talk, some recently reported deep neural network (DNN) architectures used for automatic disease classification and region segmentation will be presented. Here in some cases, features are extracted from varying receptive fields and multiple modalities by incorporating various transformed representations. For the purpose of multi-dimensional data segmentation, multi-scale contextual feature aggregation and fusion-based DNN architecture will be explained. Our recent triple attention-based DNN scheme will be introduced, which combines the channel, spatial, and pixel-level attention to enhance the feature sharing process via joint optimization. Results will be discussed considering 1D (EEG, ECG), 2D (X-ray, Ultrasound) and 3D (CT) bio-signals.