On Oct 17, 2023, at 6:58 PM, Barnet, Isabel <ibarnet@hms.harvard.edu> wrote:

 

Hi Alan and Shireen,

 

Thank you for the information!

 

I am the medical student/engineer working with Rebecca and wanted to chime in with a quick question. We understand how to interpret the data corresponding to P1-P5, but could you please clarify the significance of P0?

 

Best,

Isabel

 

From: Alan Morris <alan.morris@utah.edu>
Date: Tuesday, October 17, 2023 at 8:41 PM
To: Beroukhim, Rebecca <Rebecca.Beroukhim@CARDIO.CHBOSTON.ORG>, Shireen Elhabian <shireen@sci.utah.edu>
Cc: shapeworks-users@sci.utah.edu <shapeworks-users@sci.utah.edu>, Barnet, Isabel <ibarnet@hms.harvard.edu>, Alan Morris <amorris@sci.utah.edu>, Schulz, Noah <Noah.Schulz@cardio.chboston.org>
Subject: Re: Shapeworks [EXTERNAL]

Hi Rebecca,

 

The PCA scores in the table describe the position that each subject has along that mode of variation.  The mean shape has scores 0 for all modes.  Moving further from 0 in either direction corresponds to moving along the eigenvector for that mode.  The PCA tab in the analysis module allows you to do this individually for each mode (e.g. +/- 2 standard deviations).

 

If a given PCA mode describes a given feature, such as a lengthening and shortening of the overall shape, then you will find the samples that are long or short will be on opposite sides of the PCA spectrum (e.g. plus and minus).

 

There are lots of resources on the web for PCA and PCA Component Scores that should be applicable to this data.  The values being input to the PCA are the x,y and z of each correspondence point (a high dimensional space).

 

Thanks,

Alan

 

 

From: Beroukhim, Rebecca <Rebecca.Beroukhim@CARDIO.CHBOSTON.ORG>
Date: Tuesday, October 17, 2023 at 4:44 PM
To: Shireen Elhabian <shireen@sci.utah.edu>
Cc: shapeworks-users@sci.utah.edu <shapeworks-users@sci.utah.edu>, Barnet, Isabel <ibarnet@hms.harvard.edu>, Alan Morris <amorris@sci.utah.edu>, Schulz, Noah <Noah.Schulz@cardio.chboston.org>
Subject: RE: Shapeworks [EXTERNAL]

Thanks Shireen,

 

Forgive me if I sound colloquial (I’m copying my engineering team here). We performed a brief analysis of our data and 5 shape modes were generated by the software. However when we exported the PCA data, it sent us 6 columns of data (P0-P5). The table looked something like the one below. We are trying to understand how to interpret this output.

 

Group	P0	P1	P2	P3	P4	P5
1
1
1
2
2
2

 

Thanks

Rebecca

 

From: Shireen Elhabian <shireen@sci.utah.edu>
Sent: Tuesday, October 17, 2023 6:33 PM
To: Beroukhim, Rebecca <Rebecca.Beroukhim@CARDIO.CHBOSTON.ORG>
Cc: Shireen Elhabian <shireen@sci.utah.edu>; shapeworks-users@sci.utah.edu; Barnet, Isabel <ibarnet@hms.harvard.edu>; Alan Morris <amorris@sci.utah.edu>
Subject: Re: Shapeworks [EXTERNAL]

 

* External Email - Caution *

 

Hi Rebecca,

 

The question is a bit ambiguous. Are you referring to the corresponding particles, or the various analytical outcomes? This could include reconstructed shapes, PCA loadings, or DeepSSM results.

 

To best address this, I'd suggest checking out the analysis documentation page.

http://sciinstitute.github.io/ShapeWorks/latest/studio/studio-analyze.html

 

Please let us know if you have further questions.

 

Regards
Shireen


------------------------------------------
Shireen Elhabian, M.Sc., Ph.D.
Associate Professor of Computer Science 
Scientific Computing and Imaging Institute
Kahlert School of Computing
University of Utah
WEB 3608
72 Central Campus Drive, Salt Lake City, UT, 84112
Phone (801) 581-5982

Fax (801) 585-6513
Home Page: http://www.sci.utah.edu/~shireen

On Oct 17, 2023, at 11:11 AM, Beroukhim, Rebecca <Rebecca.Beroukhim@CARDIO.CHBOSTON.ORG> wrote:

 

Hi, we are wondering if you have a user manual for Shapeworks that includes a description of the output?

 

Thank you

Rebecca