A Special Report from SIIM 2010

At SIIM 2010, Herman Oosterwijk discussed issues that deal specifically with PACS connectivity. He outlined the following  problems:

• Network Issues: A well defined and managed network infrastructure is essential. Proper IP addressing and port number assignment has to be done. Duplicate IP addresses can create issues and are not always easy to troubleshoot. In case this is suspected, a “netscan” utility will show all IP addresses and potential duplicates. Note that DICOM devices rely on fixed IP addresses, as almost none of the PACS vendors make use of the dynamic configuration capabilities defined by the DICOM standard. Dynamic IP addressing is fine as long as the router does not re-assign them to a different address, e.g. when being re-booted or replaced. Note also that DICOM has an “official” assigned port number, i.e. port 11112, which is more reliable than the often used “well-known” port 104.

Not necessarily falling under the network but related is the need to manage AE titles making sure they are also unique. Realize that some devices have multiple AE’s with potential different AE titles. Incorrect net mask definitions and/or VLAN specifications might make certain destinations unreachable. A rather frequent occurrence is the incorrect setting of the switch, e.g. to half duplex or mismatching the device setting, especially when auto-negotiating is configured. Switch issues result in major performance issues and can only be made visible when using a network sniffer.

• DICOM Header Issues: The DICOM image header is generated through mapping RIS data, generation of the modality and manual input by a user. Either one of these sources can potentially generate incorrect and/or invalid data in the image header. Problems are unfortunately not always detected. For example, an incorrectly identified study might be archived in the PACS and get “lost”, only appearing when the data is migrated, which could be years later. Some PACS systems are more conservative than others and check every attribute, while other are more liberal and don’t necessarily complain. A header with an Institution ID exceeding the maximum length of that field might be stored by vendor A while being rejected as an invalid image when being migrated years later.

In this particular instance, the Institution ID could have been mapped from the RIS using a worklist, while not checking for any length violations (note that the source of the data, i.e. the HL7 data elements might not have the same restrictions). Missing and/or incorrect patient demographics can be caused by the RIS being down, or a technologist not using the worklist. This will cause a study to be unverified or “broken” at the PACS. Some PACS applications sort and display images according to image and/or series number instead of according to slice orientation and body part causing the images to be displayed in the incorrect order. When retrieving comparison exams, one can run across some of the older date and time formats in the header, which might cause issues as well.

• Hanging Protocol Issues: Hanging protocols not working is almost always related to incorrect header information or the wrong interpretation of the headers. A common mismatch is related to the way CR and DR systems organize their images into series. Some create a new series for each view (e.g. a Chest PA and LAT), some group them together in a single series. If the viewing software can only be configured to show different series next to each other, there will be some really unsatisfied radiologists. Another frequent issue occurs when some modalities modify automatically series and study descriptions, not taking the values from the worklist and therefore causing these descriptions not matching the hanging protocol configurations at the view station.

• CD import issues: These issues almost always can be traced back to non-compliance with the DICOM standard and/or corresponding IHE profile. Frequent issues are the absence of DICOM image files because the vendor is only providing their proprietary format, a missing directory file, mismatch of the so-called meta-file header with the actual data content, incorrect transfer syntaxes such as compression, and several others. A recent issue has also been splitting up studies over multiple CD’s. In many cases, one can convert the images to an acceptable format that can be imported; however, in some cases it is impossible to read the proprietary information, causing a repeat exam. One also need to make sure that patient identifiers are replaced, including the Accession Number otherwise the integrity of the PACS database could be compromised.

• SOP Class support: Modalities are eager to support new SOP Classes as they contain more information and allow for better viewing and processing. PACS systems traditionally lag with their support for this new functionality. The most common mismatches are due to non-support of the PACS for the enhanced CT and MRI SOP Classes, Structured reports, such as generated by CAD devices and Ultrasound units for measurements, and for new specialties such as ophthalmology, dentistry and endoscopy. In most cases, a modality can be “defeatured” to fall back to an older SOP Class, or alternate encoding (e.g. burn in the CAD marks into a secondary capture), in some cases, one will be stuck with the proprietary information (e.g. MRI spectroscopy).