Ultrasound Nerve Segmentation Using Deep Probabilistic Programming

Deep probabilistic programming concatenates the strengths of deep learning to the context of probabilistic modeling for efficient and flexible computation in practice. Being an evolving field, there exist only a few expressive programming languages for uncertainty management. This paper discusses an application for analysis of ultrasound nerve segmentation-based biomedical images. Our method uses the probabilistic programming language Edward with the U-Net model and generative adversarial networks under different optimizers. The segmentation process showed the least Dice loss (‑0.54) and the highest accuracy (0.99) with the Adam optimizer in the U-Net model with the least time consumption compared to other optimizers. The smallest amount of generative network loss in the generative adversarial network model gained was 0.69 for the Adam optimizer. The Dice loss, accuracy, time consumption and output image quality in the results show the applicability of deep probabilistic programming in the long run. Thus, we further propose a neuroscience decision support system based on the proposed approach.

[1]  Manuel Jorge Cardoso,et al.  Variational inference for medical image segmentation , 2016, Comput. Vis. Image Underst..

[2]  Reinhard Klette,et al.  Concise Computer Vision , 2014, Undergraduate Topics in Computer Science.

[3]  Zoubin Ghahramani,et al.  Probabilistic machine learning and artificial intelligence , 2015, Nature.

[4]  Sampath Jayarathna,et al.  A Survey of Attention Deficit Hyperactivity Disorder Identification Using Psychophysiological Data , 2019, Int. J. Online Biomed. Eng..

[5]  Jiqiang Guo,et al.  Stan: A Probabilistic Programming Language. , 2017, Journal of statistical software.

[6]  Léon Bottou,et al.  Large-Scale Machine Learning with Stochastic Gradient Descent , 2010, COMPSTAT.

[7]  Noah D. Goodman,et al.  Pyro: Deep Universal Probabilistic Programming , 2018, J. Mach. Learn. Res..

[8]  Martin Hirzel,et al.  Extending Stan for Deep Probabilistic Programming , 2018, ArXiv.

[9]  Sampath Jayarathna,et al.  EEG-based Processing and Classification Methodologies for Autism Spectrum Disorder: A Review , 2019, Journal of Computer Science.

[10]  Li Wang,et al.  STRAINet: Spatially Varying sTochastic Residual AdversarIal Networks for MRI Pelvic Organ Segmentation , 2019, IEEE Transactions on Neural Networks and Learning Systems.

[11]  Thomas B. Schön,et al.  Automated learning with a probabilistic programming language: Birch , 2018, Annu. Rev. Control..

[12]  Jean-Michel Marin,et al.  Approximate Bayesian computational methods , 2011, Statistics and Computing.

[13]  Anna Goldenberg,et al.  TensorFlow: Biology's Gateway to Deep Learning? , 2016, Cell systems.

[14]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[15]  Dinggang Shen,et al.  Medical Image Synthesis with Deep Convolutional Adversarial Networks , 2018, IEEE Transactions on Biomedical Engineering.

[16]  Mubarak Shah,et al.  ThoughtViz: Visualizing Human Thoughts Using Generative Adversarial Network , 2018, ACM Multimedia.

[17]  Neil D. Lawrence,et al.  MXFusion: A Modular Deep Probabilistic Programming Library , 2018 .

[18]  Sébastien Ourselin,et al.  Aleatoric uncertainty estimation with test-time augmentation for medical image segmentation with convolutional neural networks , 2018, Neurocomputing.

[19]  Mubarak Shah,et al.  Brain2Image: Converting Brain Signals into Images , 2017, ACM Multimedia.

[20]  David Broman,et al.  Delayed Sampling and Automatic Rao-Blackwellization of Probabilistic Programs , 2017, AISTATS.

[21]  Nikhil Ketkar,et al.  Introduction to PyTorch , 2021, Deep Learning with Python.

[22]  Pushmeet Kohli,et al.  TerpreT: A Probabilistic Programming Language for Program Induction , 2016, ArXiv.