Opportunistic CT

Radiation Dose

X-rays are a form of energy – like light and radio waves. X-rays are also called radiation. Unlike light waves, x-rays have enough energy to pass through your body. As the radiation moves through your body, it passes through bones, tissues, and organs differently. This allows a radiologist to create images of them. The radiologist is a specially trained doctor who can examine these images on a computer display. X-rays allow the radiologist to see the structures in your body in very fine detail.

Dose Length Product

Dose length product (DLP) is a measure of CT tube radiation output/exposure (measured in mGy.cm). It is related to volume CT dose index (CTDIvol), but CTDIvol represents the dose through a slice of an appropriate phantom. The dose length product accounts for the length of radiation output along the z-axis (the long axis of the patient). It may be described by the following equation:

DLP = (CTDIvol in mGy) . (length of scan in cm)

The dose length product does not take the size of the patient into account and is not a measure of absorbed dose. If the AP and lateral dimensions of the patient are available, then the size specific dose estimate (SSDE) can be used to estimate the absorbed dose.

Effective Dose

Some estimates of effective dose are summarized in the table below, taken from the cited sources.

Estimates of Effective Dose^1,2,3

Scan	Effective Dose (mSv)
Head	2.2
Full Spine	20.6
Cervical	3.6
Thoracic	12.0
Lumbar	5.0
Chest	6.0
Chest/Abdo/Pelvis	19.0
Abdo/Pelvis	15.6
Knees	0.16
Ankles	0.14

¹Conference of Radiation Control Program Directors (CRCPD) Nationwide Evaluation of X-ray Trends (NEXT) 2005-2006 report.

²CRCPD, Publication E-07-2, Tabulation and Graphical Summary of 2000 Survey of Computed Tomograph, (NEXT), August, 2007

³Biswas et al. Radiation Exposure from Musculoskeletal Computerized Tomographic Scans. The Journal of Bone and Joint Surgery, 2009.

Notes:

• Full Spine: From the base of the skull (C1) to the sacrum (through L5), effective dose of 20.6 mSv (adding up dose of all spine scans)

• Extended Chest/Abdo/Pelvis (or Full Spine + Abdo/Pelvis): From the base of the skull (C1) through the proximal femur. Effective dose of 25.2 mSv (cervical spine + chest + abdo/pelvis). Likely this is an overestimation.

Dose Length Product (DLP) of ARTININ scans is usually between 270 – 400 mGy-cm.

Converting between mGy-cm to effective dose (as described by the American Association of Physicists in Medicine (AAPM) Report No. 96):

E (mSv) = k × DLP

where k is the normalized effective dose per dose-length product for adults.

Estimated effective dose for ARTININ (as according to DLP):

E (mSv) = 0.015 × 400 
E (mSv) = 6 mSv

Note that in the ARTININ consent we list 10 mSv, which is on the high end (which would be for a larger person).

Full spine (estimates of DLP taken from a few participants in the spinal cord injury group): 705 - 958. 55m mGy-cm

Note that DLP is over twice what is observed in ARTININ

E (mSv) = 0.015 × 958.55 
E (mSv) = 14.5 mSv

Return to Contents or Return to Main Page

Calibration

To ensure accurate material properties and comparability across clinical sites and scanners during quantitative analysis derived from clinical CT scans, the native in-scan Hounsfield Units (HUs) must be converted to their corresponding bone equivalent densities.

There are three main types of calibration for CT scans. All of these methods are available in our GitHub repository called Ogo.

Return to Contents or Return to Main Page

Phantom calibration

Phantom calibration is the current imaging gold standard for converting HUs to bone densities. Phantom calibration involves placing a calibration phantom with known density values of either solid CHA or liquid K₂HPO₄ under the patient in the scan field of view, as seen here:

The solid CHA or liquid K₂HPO₄ found in the phantom acts as the bone equivalent reference standard. Phantom calibration requires sampling the in-scan HUs of the bone equivalent reference rods (from the calibration phantom) and comparing those values to the known reference values. A simple linear regression between the known values within the calibration phantom rods and the in-scan HUs can be created to convert the HUs to bone equivalent density (BMD).

The calibration curve takes the form of

BMD = (HU⋅a) + b

where a is the slope of the resulting calibration curve and b is the intercept. This calibration curve is applied to specific bone compartments in the image or to the entire image, creating a bone-equivalent density image. The values of a and b  depend on many factors, such as the composition of the X-ray spectrum emitted by the X-ray tube, tube current, reconstruction kernel or table height, and as a result are scanner and image protocol dependent.

To run phantom calibration in Ogo, you will need two things:

1. The raw CT scan (.nii or .nii.gz)
2. A segmentation mask of the phantom rods

Typically, this is done using ITKSnap or some other automated method [TBD], and you segment the inside of the rods, making sure to stop before reaching the edge of the rod to avoid partial volume effects.

An example of the phantom rods segmented:

Return to Contents or Return to Main Page

Asynchronous calibration

This also uses a phantom, but not at the same time as the scan of the patient. Sometimes it can be months or even years away. (This is an ongoing edited section...)

Return to Contents or Return to Main Page

Internal calibration

This doesn't use a phantom, but instead uses the internal tissues as referents.

Return to Contents or Return to Main Page

DICOM Meta Data

DICOM Dump

If you want a comprehensive list of tags then I suggest "dicomdump" to get output like the following, and then apply some filters:

(0008,0070) LO "Manufacturer" : [GE MEDICAL SYSTEMS] (18 bytes)
(0008,0080) LO "InstitutionName" : [FOOTHILLS Hospital] (18 bytes)
(0008,0090) PN "ReferringPhysicianName" : [] (0 bytes)
(0008,1010) SH "StationName" : [MRGESEA1] (8 bytes)
(0008,1030) LO "StudyDescription" : [QUANT MR-(RF_FMC_QUANMR_01_13360)] (34 bytes)
(0008,103E) LO "SeriesDescription" : [OAx T2-Mapping REG NEW] (22 bytes)
(0008,1090) LO "ManufacturerModelName" : [DISCOVERY MR750] (16 bytes)
(0008,1111) SQ "ReferencedPerformedProcedureStepSequence" : (1 item)
  ---- SQ Item 0001 at offset 844 ----
  (0008,1150) UI "ReferencedSOPClassUID" : [1.2.840.10008.3.1.2.3.3] {Modality Performed Procedure Step SOP Class} (24 bytes)
  (0008,1155) UI "ReferencedSOPInstanceUID" : [1.2.840.2948978595] (60 bytes)

Filters to apply to get "just the good stuff":

1. Remove anything in group 0002
2. Remove anything in odd-numbered groups (private vendor attributes)
3. Remove anything that is indented (these are the nested attributes)
4. Remove anything that is "UN", "SQ", "OB", "OW", "OF", "OD", "OL", "OV"

After you've removed all that stuff, you can extract the attributes (e.g. SeriesDescription, etc.).

I recommend putting them all into a file (I usually call this file "query.txt") instead of having them all on the command line:

  dicomtocsv  ./path_to_data -q query.txt -o output.csv

The query.txt will have contents like this, one attribute per line:

   Manufacturer
   StudyDescription
   SeriesDescription

I don't recommend using "--default" because future versions of dicomtocsv might change what counts as "default". It's best to list all of the wanted tags explicitly.