W-Net: Novel Deep Supervision for Deep Learning-based Cardiac Magnetic Resonance Imaging Segmentation

Abstract

Cardiac magnetic resonance imaging (CMRI) segmentation transforms cardiac MR images into semantic regions to define the left ventricle cavity, right ventricle cavity, and myocardium. CMRI segmentation provides ventricles’ volume, mass, and ejection fraction, playing a significant role in cardiac disease diagnosis. This paper aims to propose novel deep supervision in U-Net-based architecture to enhance the segmentation performance. It presents a deeply supervised W-Net, which creates another path in parallel with the decoder path in U-Net-based architecture. The output of every upsampling layer in the decoder path is combined with pixel-wise addition for feature reuse, and loss is computed at each feature dimension on the deep supervision layer, which enables gradients to be implanted at a greater depth into the network and enhances the training of all layers in the network. Proposed W-Net applied on single scanner-based ACDC dataset and Multi-Centre, Multi-Vendor & Multi-Disease dataset, making it more robust in model generalization. W-Net significantly outperforms numerous state-of-the-art methods on the two publicly available CMRI datasets, according to experiments conducted. W-Net obtained better segmentation results ranked in the top three for many metrics. It is evident that the proposed W-Net has considerable potential in CMRI segmentation, cardiac assessment, and disease diagnosis.

Keywords:

• Cardiac magnetic resonance imaging
• Cardiac MRI segmentation
• Deep learning
• Deep supervision
• Generalizability
• U-Net

ACKNOWLEDGEMENT

The authors would like to thank the Head of the Department of Electrical Engineering and Institute Computer Centre of Indian Institute of Technology, Roorkee, for providing the necessary resources for GPU-based computing.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Kamal Raj Singh

Kamal Raj Singh received the BTech degree in electronics and communication engineering from the Uttar Pradesh Technical University, Lucknow, Uttar Pradesh, India, in 2009, and the ME degree in communication engineering from the Birla Institute of Technology and Science Pilani, Pilani, Rajasthan, in 2011. He worked in the industry for nearly one-and-a-half years. He was an assistant professor with the Department of Electronics and Communication Engineering, Inderprastha Engineering College, Ghaziabad, from 2012 to 2018. Currently, he is working as a research scholar with the Department of Electrical Engineering, IIT Roorkee, Roorkee, India. His research interests include medical image analysis, cardiac magnetic resonance imaging segmentation, application of deep learning to biomedical image processing, and artificial intelligence. Corresponding author. Email: [email protected]

Ambalika Sharma

Ambalika Sharma received a BE degree in electronics and communication engineering from the Indian Institute of Technology, Roorkee, India, and an MTech degree in electrical engineering from the Indian Institute of Technology, Roorkee, India and a PhD degree from Annamalai University, India. Currently, she is an assistant professor in the Department of Electrical Engineering at the Indian Institute of Technology, Roorkee, India. Her research interests are in medical image processing, signal processing, deep learning, wavelet and multi-rate signal processing, ECG signal analysis and heart rate variability. Email: [email protected]

Girish Kumar Singh

Girish Kumar Singh received the BTech degree in electrical engineering from the G. B. Pant University of Agriculture and Technology, Pantnagar, India, in 1981, and the PhD degree in electrical engineering from Banaras Hindu University, Varanasi, India, in 1991. He worked in the industry for nearly five-and-a-half years. In 1991, he became a lecturer at the M N R Engineering College, Allahabad, India. In 1996, he moved to the University of Roorkee, Roorkee, India. He was a visiting associate professor with the Department of Electrical Engineering, Pohang University of Science and Technology (POSTECH), Pohang, South Korea, and a visiting professor with the Department of Electrical and Electronics Engineering, Middle East Technical University, Ankara, Turkey. He is currently a professor with the Department of Electrical Engineering, IIT Roorkee, Roorkee. He has been involved in the design and analysis of electrical machines in general and high-phase-order ac machines, as well as power system harmonics and power quality. Recently, a book entitled “Computational Intelligence and Biomedical Signal Processing”, co-authored by him has also been published by M/s Springer Nature. He has coordinated several research projects sponsored by the Council of Scientific and Industrial Research (CSIR) and the University Grants Commission (UGC), Government of India. Prof Singh received the Pt Madan Mohan Malaviya Memorial Medal and the Certificate of Merit Award at the Institution of Engineers, India, from 2001 to 2002. He secured rank 1 in India and 250 world ranking (top 0.15%) in the subject area “Networking and Telecommunications” as per the independent study done and published by Standford University, USA, in 2020. Its' 2021 report on worldwide researcher has placed him at 266 world ranking (top 0.11%) in the subject area “Energy.” Email: [email protected]