Structured BMD Data Permits Easy Query and Data Analysis

The two prior posts, Bone Density Reporting and PACS and The Evolution of Bone Density Reporting, prompted feedback from readers and BoneStation users.  The articles mentioned that quantitative bone mineral density data (BMD, t-score, z-score, etc.) is available in a structured form in the DICOM format.  Apparently this is quite appealing to physicians and researchers who would like to analyze and mine bone density data.

In this posting we will provide more information the bone density data in DICOM files.  We will describe where the data is stored, how it may be accessed, and the types of things that can be done with it.

Bone density data is available in the DICOM transmissions of bone density scans.  Specifically, BMD data is available in two forms – a raw image and a structured form.  The raw image is of little use in terms of analysis because the numerical information (area, BMC, BMD, t-score, and z-score) can not be extracted out of the image.  However, the structured form may be of considerable value because it can be parsed.

The structured BMD data is not visible when looking at a DICOM image.  The data is stored in private DICOM elements.  GE/Lunar and Hologic use their own proprietary formats.  Fortunately, each DXA manufacturer documents their format.  We have seen very few systems that utilize the structured BMD data stored in DICOM and have yet to encounter a PACS that makes use of the private data.

The DICOM standard supports many modalities – CT, Ultrasound, etc.  Unfortunately, DXA is not one of them.  This is the reason the DXA manufacturers have created their own private DICOM fields for storing BMD data.

BoneStation depends heavily on the structured BMD data.  It parses the data and stores it in its database.  From there, BoneStation can display the data in ways that are useful to physicians.  It can:

  • Perform calculations, such as change in BMD between arbitrary scans
  • Highlight scans performed on different DXA machines or with different scan modes
  • Highlight questionable scan values
  • Assist the physician in assessing an exam – for example, an interpretation may be provided based on t-score

Of course, more than just scan data is available.  BoneStation captures additional information, some of which is customized per user.  Some of this information is entered during the review process and some via an online patient history questionnaire.  A sample of data that may be available:

  • Treatments, current and past
  • FRAX risk factors
  • ICD9 codes
  • Vertebral Fracture Assessment (VFA) fractures, including severity and type
  • Etc…

All of this information is stored in a standard relational database and may be queried using Structured Query Language (SQL).  Tools such as Microsoft Excel and Crystal Reports may be used to access the database. A wide variety of queries may be performed.  Here is a very small sample of the types of queries that may be of interest.

  • How many bone density scans were performed by month for the past year.
  • Find all patients with a t-score within a range – say t-score <= -2.5.
  • Find all male patients under 65 with a t-score below -2.5.
  • Find all patients being treated for osteoporosis who are osteopenic.
  • How many patients are being treated with a specific ICD9 code for each of the past 3 years.
  • Find patients with a moderate or worse VFA fracture.
  • How many scans is each physician reviewing.
  • How many scans have poor quality.
  • How much time does each physician take to interpret scans.

Some astute readers picked up on the value of structured BMD data in DXA DICOM transmissions.  Structured data can be stored in an organized fashion and easily queried and mined for clinical, quality, research, and financial purposes.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: