Saturday, October 1, 2011

Eliot Siegel Q&A

This is a transcript of the Q and A session of the September vDHIMS eposium presentation by Eliot Siegel about advanced visualization workstations. If you are interested, you can listen to the full one-hour presentation, simply register at and enjoy. 

Q: There is a difference in the quantitative output of several advanced visualization workstations among the different vendors, do you see a potential standardization and/or certification by a company such as ECRI?

A: That is a great question and it is of tremendous concern to several people including myself with regard to quality control as we seem to look for the esthetics of how the images look but when we are making quantitative measurements either manually or by using the software, the measurements vary considerably. 

I propose to do a couple of things, and we have been talking with some vendors about them. The first thing would be to have standard scans of phantom data, for example creating a phantom for lung nodules or carotid stenosis. Another option is to work with NIST, which has created standard objects that have been measured very precisely that we can scan. Yet another option is to create a mathematical model, so we would not have to use the scanner to create a data set, and there are interesting data sets that are well known and which can be submitted to the vendors. 

The problem is that it is hard to reproduce the human anatomy with phantoms, therefore one might use a de-identified data set, with patient approval, and share those and use them to create a semi-standard. It would be great if one could go to RSNA or another meeting and go to a vendor and look at a standard data set for carotid imaging or cardiac etc. So I think it is a great idea and, as a customer and a person who is interested in quality improvement, I would very much like to pursue that. 

Q: Do you keep the thin axial CT slices and what would you recommend for a typical hospital? 

A: I work in multiple clinical settings and at the University of Maryland, we keep them for only three to six months unless it is designated as a research study or need to be kept for other purposes. At the VA, we keep all of our thin slice data indefinitely. My recommendation would be for everyone to keep the thin slices indefinitely. However, I think that if you look across the country, only a minority of institutions keeps the thin slices. 

When we talk with the legal folks about what data to retain, the answer that they give us is that you should retain data that you used for making your original clinical diagnosis. I and other people are doing image interpretation from the thin slices and therefore logically the conclusion would seem to be that if we use the data for making the day-to-day diagnosis, we ought to be keeping that information because my decision was partly predicated upon what I would see in an oblique image or a reconstructed image that was synthesized from the original data. I don't really have a record of what I saw unless I am able to save the thin slices. Therefore my philosophy is to save it, especially with the cost of storage declining. One compromise for institutions who are having cost issues would be to compress thin sections in a number of different ways. You could store the thick sections uncompressed and then use, for example, a JPEG compression for the thin sections. Therefore my philosophy is that in the near future everybody will start saving the thin sections.