01 05 2017 Kick off meeting

Goals: Understand the goals Identify who’s doing what Deliverables, deadlines

Bea: Our main deliverable: Software People Data will come from Craniomaxillofacial R01s Neuroimaging R01s Cancer R01 Timelines NIH cut us by 1 year, but we should get a no-cost-extension We should start dissmination efforts Subcontract administration

Administrative stuff: Weekly software dev meetings in the beginning Which talks about architecture of SALT itself. Biweekly software dev t-cons Whole meeting t-con before progress report (mid April) Similar personal meeting in Sept 2017 Data sharing: Common Google Drive (CloudHQ) Folder organization? Wiki options on github? (Paid) Individual meetings for each methodology to discuss their needs and architecture. We should have frequent releases for software as long as we have hooks for more advanced stuff to come. Hackfest for all the developers in the next two months

Major work in the coming months would be to figure out the architecture: slicer and additions Software: Workflow: Inputs (segmented) Data importer from fsl, freesurfer, autoseg Will give visual feedback Data preprocessing Quality control for spherical and non-spherical People are responsible for correcting Hooks would be needed from the output to the data processing as well. Shape Correspondence Visualizing correspondence computations, etc. Ability to adjust correspondences. We start with a simplified versions. Correspondence postprocesing and quality control Shape analysis outputs (4d regression methods, etc) There should be a way skipping some steps Shape statistics packages available from Stephen’s book group on PDM Statistics. We should know what they have. There is an opportunity some of our methods available to them and vice versa.

Learning from SPHARM: Visual feedback on problems with Input daata Speed Going away from Batchmake because they create challenges for users and evelopers

Immediate software goals: SPHARM-PDM Remove batchmake: use iPythong, Slicer’s pipeline architecture (CWL), girder Python gui for shape analysis module Study improvements to time-bottlenecks Need to figure out a way to share optimizer We’ll have github repository for SlicerSPT with all the functionalities to vtk and itk available. Bea and Hina will be reviewing the code and ctests High quality software

Methodology:

Martin: Study specific shape analysis Improve SPHARM-PDM correspondences over all datasets in a study Provide a version of the method in Slicer and make it as user friendly as possible. Need GUI, tests, Need binding to spharm-pdm. And landmark inclusion needs to be available Public validation data Simulation data needs to be added with known differences Improve visualization for quality control Landmark manipulation Bea : Topology checker Non-spherical topology tool available and needs to be packaged and given a gui.

Steve: S-reps: is being rebuilt from scratch S-reps fitting would need a lot of material from the shape-stats community. Have tools for generating a library or a base structure for creating a reference skeleton model Pablo visualizations should be implemented here JP’s method for changing a 3d distance function into spline and hence derivatives is up for grabs to be added into the software.

CPNS is better over PCA for PDMs but slow. Would it be possible to give an option to choose between CPNS and PCA

Guido Gerig: Geodesic regression and acceleration controlled methods. Inputs: Series of shapes Output: 4d shape model, shape features can be measured continuously Current state of software: geodesic regression: part of deformatcia, but is ready to be shared in SALT. Acceleration controlled: Shape4d Need a GUI. Use both methods for common interface Visualization tools, graphs and shape feature visualizations are being looked forward to.

Hongtu: Matlab or python implementations of methods available. Multiple tools with some parallel processing and large data analysis challenges. We can pick up one of the tools and add it to SALTSlicer.

Data:

Neuroimaging: Need data for evaluation and validation of the shape analysis algorithms, but keeping in sync with the research objectives of the data providers. IBIS: There are multiple structures available segmented from the brain datasets (per Martin). Processing, statistical analysis will be done by Mahmoud.

John’s study: Lateroventrical.. Finding markers for schizophrenia Relating post-natal volumes with pre-natal measures. Relating the shape features for ventricles as well. (atrial width prenatal with cortical thickness) Maybe using Guido’s longitudinal studies?

Craniomaxiloofacial:

2 example studies are being used… Need correspondence manipulation improvements Orthognatic surgery: Mandibular regression. All whole head segmentation.

Colon Cancer: Colonoscopy: Polyps have been missed? WHat are the things that were missed? Visualizing the surfaces, and finding holes. Modeling colons in high resolution that could let us find the holes: Holes being things that the colonoscope failed, or polyps. Cross patient shape analysis will help in finding the holes. Inter patient analysis. This would be basically about coming up with coordinate system mappings. A: Can we get model types to include cylindrical and spherical models B: Can we come up with a way to deal with this coordinate change which is basically correspondence problem

Organize a meeting of Stephen with Julian and Steve to demonstrate depthmap generation for endoscope images.

Dissemination:

We’ll have one paper regarding software at the end of year 2. At the end of year two will also start dissemination in terms of tutorial and workshops Grad students should develop tutorials Skeleton for tutorial will be provided

Workshops For Cancer data: American associan of physicists (AAPM) GI conferences

Moving forward: SPHARM into SALT Dashboards and testing infrastructure involved Comparison framework. Dissemination infrastructure website and forums Start methodological integration Architecture needs to be decided early on. Publication related research All the students should be familiar with ITK