TernaryNet: faster deep model inference without GPUs for medical 3D segmentation using sparse and binary convolutions

International Journal of Computer Assisted Radiology and Surgery
Mattias P HeinrichOzan Oktay

Abstract

Deep convolutional neural networks (DCNN) are currently ubiquitous in medical imaging. While their versatility and high-quality results for common image analysis tasks including segmentation, localisation and prediction is astonishing, the large representational power comes at the cost of highly demanding computational effort. This limits their practical applications for image-guided interventions and diagnostic (point-of-care) support using mobile devices without graphics processing units (GPU). We propose a new scheme that approximates both trainable weights and neural activations in deep networks by ternary values and tackles the open question of backpropagation when dealing with non-differentiable functions. Our solution enables the removal of the expensive floating-point matrix multiplications throughout any convolutional neural network and replaces them by energy- and time-preserving binary operators and population counts. We evaluate our approach for the segmentation of the pancreas in CT. Here, our ternary approximation within a fully convolutional network leads to more than 90% memory reductions and high accuracy (without any post-processing) with a Dice overlap of 71.0% that comes close to the one obtained when using ...Continue Reading

References

Oct 1, 2010·Medical Image Computing and Computer-assisted Intervention : MICCAI·Amal FaragRobert Falk
Jun 21, 2011·Current Opinion in Neurobiology·Maurice J ChacronLeonard Maler
Nov 16, 2011·IEEE Transactions on Pattern Analysis and Machine Intelligence·Michael CalonderPascal Fua
Jan 1, 2013·Information Processing in Medical Imaging : Proceedings of the ... Conference·Mattias P HeinrichJulia A Schnabel
Oct 15, 2014·IEEE Transactions on Medical Imaging·Xiaofan ZhangShaoting Zhang
May 24, 2016·IEEE Transactions on Bio-medical Engineering·Yuhang XuKerry R Mills
Jan 5, 2017·Scientific Reports·M L JinC Q Jin
Feb 23, 2018·Neural Networks : the Official Journal of the International Neural Network Society·Lei DengGuoqi Li

❮ Previous
Next ❯

Citations

Nov 16, 2018·International Journal of Computer Assisted Radiology and Surgery·Max Blendowski, Mattias P Heinrich
Jul 27, 2021·Journal of Healthcare Engineering·Meixiang HuangDexing Kong
Aug 1, 2021·BMC Medical Informatics and Decision Making·Kinshuk Sengupta, Praveen Ranjan Srivastava
Aug 21, 2021·Medical Image Analysis·Rongzhao Zhang, Albert C S Chung
Aug 31, 2021·Journal of Imaging·Boris ShirokikhMikhail Belyaev

❮ Previous
Next ❯

Related Concepts

Related Feeds

Brain Imaging of Neural Circuits

Neural circuits are groups of interconnected neurons which carry out specific functions when activated. Imaging these neural circuits allows researches to further elucidate their mechanisms and functions. Follow this feed to stay up to date on brain imaging of neural circuits.

Brain Imaging of Neural Circuits (MDS)

Neural circuits are groups of interconnected neurons which carry out specific functions when activated. Imaging these neural circuits allows researches to further elucidate their mechanisms and functions. Follow this feed to stay up to date on brain imaging of neural circuits.

Related Papers

International Journal of Neural Systems
Noel Lopes, Bernardete Ribeiro
Quantitative Imaging in Medicine and Surgery
Lin ShiDefeng Wang
Proceedings of the International Society for Magnetic Resonance in Medicine ... Scientific Meeting and Exhibition
Zhenghan FangD Shen
© 2022 Meta ULC. All rights reserved