15.12.2012 Views

Digital Imaging and Communications in Medicine (DICOM)

Digital Imaging and Communications in Medicine (DICOM)

Digital Imaging and Communications in Medicine (DICOM)

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

7.3 Storage 135<br />

C-Store bottlenecks<br />

If you implement <strong>DICOM</strong> software that uses C-Store to send image series,<br />

never <strong>in</strong>terpret a s<strong>in</strong>gle image store failure as a failure for the entire series.<br />

A s<strong>in</strong>gle image <strong>in</strong> a series can fail C-Store for many totally benign reasons<br />

(for example, the image already exists at the dest<strong>in</strong>ation) <strong>and</strong> it should not<br />

stop the transmission of the rema<strong>in</strong><strong>in</strong>g images.<br />

More importantly, never <strong>in</strong>terpret a C-Store failure for some images as an<br />

excuse for halt<strong>in</strong>g any subsequent communications. We had this experience<br />

with one CR device that was send<strong>in</strong>g images to a digital archive. It<br />

got stuck on one particular study with an <strong>in</strong>correctly entered patient ID<br />

(which was illegally blank). The dest<strong>in</strong>ation archive recognized the ID<br />

problem <strong>and</strong> refused to accept the study. Surpris<strong>in</strong>gly enough, the CR unit<br />

<strong>in</strong>terpreted this s<strong>in</strong>gle-study failure as a major disaster <strong>and</strong> stopped send<strong>in</strong>g<br />

all subsequent studies after the one that failed, even though the new<br />

ones had absolutely correct patient IDs! In a short time, all unsent studies<br />

queued on the CR, filled up its local hard drive, <strong>and</strong> the unit refused to<br />

process any new patients.<br />

This provides an excellent example of how a s<strong>in</strong>gle m<strong>in</strong>or problem, magnified<br />

by very poor software design, bottlenecked the entire PACS workflow.<br />

7.3.2<br />

C-Store DIMSE<br />

C-Store takes care of <strong>in</strong>struct<strong>in</strong>g the image-accept<strong>in</strong>g AE on what needs to be<br />

done. Just like C-Echo, C-Store has a request part (sent from the C-Store SCU)<br />

<strong>and</strong> a response part (replied from the C-Store SCP). If we compare Table 26<br />

to the C-Echo-Rq we studied earlier (see 7.2.2) we f<strong>in</strong>d a few new parameters,<br />

which we describe below:<br />

1. Priority: as the table suggests, priority can be low, medium (normal), or high.<br />

Many other DIMSE messages <strong>in</strong>clude a priority field <strong>in</strong> their specifications,<br />

but its support is optional <strong>and</strong> defaults to medium (0000). In reality, different<br />

message priorities are almost never implemented by <strong>DICOM</strong> manufacturers,<br />

<strong>and</strong> this is noted <strong>in</strong> their conformance statements. This makes<br />

practical sense because ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g a functional PACS network is much<br />

more important than prioritiz<strong>in</strong>g messages, <strong>and</strong> any priority-assign<strong>in</strong>g<br />

mechanism only adds unnecessary complexity <strong>and</strong> unpredictability. Delays<br />

<strong>in</strong> data acquisition, transmission, <strong>and</strong> retrieval, which are <strong>in</strong>evitable <strong>in</strong> any<br />

real cl<strong>in</strong>ical environment, have a much stronger impact on any messag<strong>in</strong>g<br />

scheme than predef<strong>in</strong>ed message priorities. In short, SOP message priority<br />

is rarely needed.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!