SASHIMI2018: Simulation and Synthesis in Medical Imaging

  • 1

Overview

SASHIMI 2018: A MICCAI 2018 Workshop

The MICCAI community needs data with known ground truth to develop, evaluate, and validate image analysis and reconstruction algorithms. Since synthetic data are ideally suited for this purpose, over the years, a full range of models underpinning image simulation and synthesis have been developed: (i) simplified mathematical models to test segmentation and registration algorithms; (ii) detailed mechanistic models (top-down), which incorporate priors on the geometry and physics of image acquisition and formation processes; and (iii) complex spatio-temporal computational models of anatomical variability, organ physiology, or disease progression. Recently, cross-fertilisation between image computing and machine learning gave rise to data-driven, phenomelogical models (bottom-up) that stem from learning directly data associations across modalities, resolutions, etc. With this, not only the application scope has been expanded but also the underlying model assumptions have been refined to increasing levels of realism.

This half-day workshop aims to stimulate the discussion and research in simulation and synthesis approaches, invite new ideas on how to best characterise and evaluate these techniques, and ultimately help bring these two synergistic perspectives closer together.

Topics

Topics of interest include, but are not limited to, the following:

  • Fundamental methods for image-based biophysical modeling and image synthesis
  • Biophysical and data-driven models of disease progression or organ development
  • Biophysical and data-driven models of organ motion and deformation
  • Biophysical and data-driven models of image formation and acquisition
  • Segmentation / registration across or within modalities to aid the learning of model parameters
  • 1
  • Imaging protocol harmonization approaches across imaging systems, sites and time points
  • Image synthesis for normalization and spatio-temporal intensity correction
  • Cross modality (PET/MR, PET/CT, CT/MR, etc.) image synthesis
  • Simulation and synthesis from large-scale image databases
  • Automated techniques for quality assessment of simulations and synthetic images
  • 1
  • Image synthesis in high dimensional spaces (vectors, tensors, spatio-temporal features, etc.)
  • Handling uncertainty and incomplete data via simulation and synthesis techniques
  • Evaluation and benchmarking of state of-the-art approaches in simulation and synthesis
  • Normative and annotated datasets for benchmarking and learning models
  • Novel ideas on evaluation metrics and methods in image-based simulation and image synthesis
  • 1
  • Applications of image synthesis in super resolution imaging and multi/cross-scale regression
  • Applications of image synthesis and simulation in medical image registration and segmentation
  • Applications of image synthesis and simulation in image denoising and information fusion
  • Applications of synthesis and simulation to image reconstruction from sparse data or sparse views
  • Applications of image and data synthesis to real-time simulation of biophysical properties
  • 1

Organization committee

image
Ali Gooya
University of Sheffield, UK
image
Orcun Goksel
ETH Zurich, Switzerland
image
Ipek Oguz
University of Pennsylvania, USA
image
Ninon Burgos
Inria Paris, France

Important dates

Submission Deadline
June 22nd,
2018
Notification of Decisions
July 20th,
2018
Workshop Event
Sept 16th,
2017

Academic objectives

Our objectives are to:
  • Bring together experts of synthesis and simulation.
  • Strengthen the links between machine learning and mechanistic models.
  • Invite new ideas on how to best characterise and evaluate these techniques.
  • Collect curated benchmark datasets and accompanying suitable evaluation criteria.
  • Identify challenges and opportunities for further research.
While the amount of imaging data increases, our uncertainty on the fidelity and variability of data is not decreasing; as such, there is an ongoing interest in evaluating the robustness of the algorithms that our community produces. Setting early on processes and metrics to do so is imperative and timely.