1 of 22

SlicerAstro: an interactive 3D visual analytics tool for HI data

Davide Punzo*,

Thijs van der Hulst,

Jean-Christophe Fillion-Robin.

*D.Punzo@astro.rug.nl, punzodavide@hotmail.it

Image credit: T. Oosterloo

2 of 22

1

2

Each data-cube contains ~100 HI sources

An Apertif

data-cube:

~ 1011 voxels

most (99%) will be dominated by noise

 

Data: Ramatsoku et al., 2016.

Apertif (Verheijen et al., 2009, in Panoramic Radio Astronomy, PoS )

will detect HI in hundreds of thousands of galaxies in the northern sky.

Data rate 10 cubes/week. We will enter in the Big data domain for two reasons:

WALLABY (Southern sky; Johnston at al., 2008)

will have similar characteristics

8deg2

Subcubes around sources will typically contain~ 105 voxels

3 of 22

Galaxies in the Ursa Major cluster

Data: Busekool and Verheijen

4 of 22

extra-planar gas tidal filaments ram-pressure tails

3D interactive

visualization techniques

can help finding such features

Image credit: M. Verheijen

 

Our aim

5 of 22

3D interactive

visualization

of HI datasets

Filtering and 3D

3D selection

Analysis, modeling

in full 3D

SlicerAstro

6 of 22

Downloading SlicerAstro

7 of 22

Well documented:

High level of modularity

Open Source

CPU/GPU Volume Rendering

Multi-platform (Qt5 and CMake)

Long-term maintainability

  • SlicerWiki
  • github.com/Slicer/Slicer
  • Mantis bug report
  • https://discourse.slicer.org/

Synergies between astronomical and medical visualization

www.slicer.org

See also Punzo et al., 2015,

Astronomy and Computing, 12, 86.

3DSlicer

8 of 22

Data:

M. Ramatsoku

Inspecting HI in a galaxy in 3D

XY

YZ

XZ

See also Punzo et al., 2015, Astronomy and Computing, 12, 86.

Overview of coherent structures in space and velocity

3D

pros:

9 of 22

HI Filament

CPU (OpenMP) / GPU(OpenGL) filtering techniques:

Box

A

B

C

Gaussian

Intensity-driven gradient

Discovering faint structure

3D + filtering

pros:

See also Punzo et al., 2016, Astronomy and Computing, 17, 163.

Inspecting a faint HI filament in 3D

10 of 22

Modify interactively

the selection

Selecting ROI for

further analysis

3D

pros:

See also Punzo et al., 2017, Astronomy and Computing.

3D selection: CloudLasso

11 of 22

Quantitative visualization: Histogram

12 of 22

Quantitative visualization: Profiles

13 of 22

Quantitative visualization: PV Slice

14 of 22

Quantitative visualization: PV Diagram

15 of 22

Quantitative visualization

  • … and :

  • Load and visualize masks;
  • Interactive masking;
  • Statistics in a selection;
  • Generate moment maps;
  • Reprojection of data;
  • Contours;
  • Load catalogs and overlay markups on data.

See https://github.com/Punzo/SlicerAstro/wiki

16 of 22

Quantitative visualization

  • … and :

  • Load and visualize masks;
  • Interactive masking;
  • Statistics in a selection;
  • Generate moment maps;
  • Reprojection of data;
  • Contours;
  • Load catalogs and overlay markups on data.

See https://github.com/Punzo/SlicerAstro/wiki

17 of 22

Quantitative visualization

  • … and :

  • Load and visualize masks;
  • Interactive masking;
  • Statistics in a selection;
  • Generate moment maps;
  • Reprojection of data;
  • Contours;
  • Load catalogs and overlay markups on data.

See https://github.com/Punzo/SlicerAstro/wiki

18 of 22

Quantitative visualization

  • … and :

  • Load and visualize masks;
  • Interactive masking;
  • Statistics in a selection;
  • Generate moment maps;
  • Reprojection of data;
  • Contours;
  • Load catalogs and overlay markups on data.

See https://github.com/Punzo/SlicerAstro/wiki

19 of 22

 

Automated fitting:

See also Di Teodoro et al., 2015, MNRAS; https://github.com/editeodoro/Bbarolo

NGC 5055

Image credit: T. Oosterloo

Image credit: T. van der Hulst

Image credit: Rogstad et al. 1974

Tilted-ring model fitting

20 of 22

Discovering subtle structures

3D + modeling

pros:

See also Punzo et al., 2017, Astronomy and Computing.

NGC 2403

Kinematic models as masks

21 of 22

Visual Analytics example: manual model refinement

22 of 22

The main aim of SlicerAstro is to aid source inspection

and interactive analysis of spectral line data;

The 3D visualization gives an immediate overview of coherent structures in space and velocity;

The 3D visualization (coupled with filtering ad modeling) greatly helps the discovery of faint and subtle structures;

SlicerAstro source code available at:

https://github.com/Punzo/SlicerAstro

Binaries (Linux and Mac) are available in the

3DSlicer Extension Manager

http://download.slicer.org/

A

B

C

D

We thank S. Pieper (Isomics), A. Lasso (Queen's University)

and K. Martin (Kitware) for their feedback and help.

Final Remarks