Imaging QA: Image Eval Errors

Every Image MUST include (may be in DICOM header)

Required by law…

Patient Name 

Patient DOB 

Date of Exam 

Location of Exam 

Rt or Lt Markers

NOT required by law on images...

MR#

Tech’s ID (initials, etc..)

EI Number

Referring Physician

Reading Radiologist

Common Errors …

Wrong Patient - may have x-rayed correct pt, but send under wrong ID

Wrong Body Part - sent under wrong body part, wrong accession # = Hard for physician to locate 

Wrong side - Wrong marker, but correct side OR wrong accession # for bilateral examinations 

Wrong Annotation - Upright vs. Supine; PA vs. AP

Receptor Exposure

The amount of radiation striking the receptor 

Mottle

Saturation - soft tissues eliminated 

patient and technical factors need consideration 

S#s and EI#s are brand specific and will not be assessed on the ARRT Exam 

(Image shows saturation - Maybe Big Belly? ) 

Processing Errors 

Histogram Analysis 

* Histogram analysis error- need to collimate

• Values of Interest

• Rescaling - Modifying image to what the computer things it should look like

• LUTs - processed with the right amount of brightness and contrast 

• Correct Collimation is crucial - lowest exposure regions of a histogram can represent collimated borders

  • Asymmetrical collimation causes problems in CR b/c CR is processed looking for 2 or 4 borders 

Artifacts 

Any part of an image that does not accurately represent the anatomic structures present within the subject being evaluated

Damages:

Receptor physical damage to the receptor

Hyperdense artifacts 

DR dexel malfunctions (Dead dels) 

Damages:

Pixel Malfunctions - malfunctions with the screen

Dead or damaged pixels

Located in the exact same location every image 

Detector Calibration Errors

Grid Line Suppression Artifact 

Dexel Drop - Out Effects

In DR systems, dead detector elements (dexels) can fail = not produce a gray square = broken 

Entire rows or columns of dexels can also drop out due to electronic failure 

• Mild - moderate issues are compensated for 

• Many rows (severe), the detector plate will need to be replaced 

* the most common way to correct these is using a software Kernel in which the values of the 8 pixels surrounding a dead pixel are averaged, then this value is inserted into the dead pixel

• Steps

  • SUM eight surrounding pixels

  • AVERAGE these values

  • INSERT result into centered

Software can compensate for moderate cases of Dexel Drop-out, but not for severe cases like this (image), where whole sections of rows have dropped out

Data Clipping

If the dynamic range or bit depth of a digital processing system is too limited, it is possible for data clipping to occur when either brightness or contrast is adjusted. 

The dynamic range of the software, supported by the bit depth of the computer hardware, must extend sufficiently above and below typical input values to allow for all probable adjustments to the image.

Would restrict the amount of data shared to the Radiologist and therefore limit their windowing ability. 

Mid - Gray Clipping - its usually bright whites and dark blacks and that would be ok

Aliasing (Moire Artifact) 

An interference pattern 

A false presentation of artefactual lines 

One way is to overlap two similar patterns of alternating, periodic structures, especially high contrast structures, such as grid lines from two overlapping grids.

In CR Processing, an electronic version called “aliasing” is much more likely to occur if…

A stationary grid is used, and the grid lines run parallel to the scanning lines of the CR Reader

If the frequency of the grid lines is close to the scanning frequency of the CR reader and they run parallel to each other, what may occur?

Artifact

*Older grids often have a frequency of 40-50 lines/cm

TO Prevent aliasing artifacts…..

Use high - frequency grids at >50 lines per cm (expensive)

Use new multi-hole grids

Use short-dimension (SD) grids positioned such that the grid lines run perpendicular to the CR scanning lines

Long Dimension vs. Short Dimension

Lead strips run parallel vs. perpendicular

17 inch vs. 14 inches

Alignment Issues

For CR, exposure indicator errors are likely unless at least 30 percent of imaging plate is exposed 

• Therefore, for tightly-colllimated views of digits, it is recommended that two or three views be taken on one imaging plate.

• Use at least 1/3 of the CR plate for an exposure

  • CR IRs come in different sizes

NOTE-

*Because they scan the receptor plate in sections, DR systems are not subject this 30% rule for plate coverage

Bilateral Projections 

When “manual” technique is used, CR sytems generally have no difficulty processing bilat views without processing errors

However, when AEC is used, special care must be taken to active the two side detector cells and ensure that the anatomy is positioned directly over the energized cells, to avoid early shut-off and unacceptable mottle.

*Partition pattern recognition errors/segmentation errors

Artifacts; CR

Dust, aging phosphors, ghosting, scratches

Laser jitters = wavy appearance 

Laser obstruction 

Radiation Fog 

Ghost Images 

• CR only issue 

• Incomplete erasure of the CR receptor 

Backscatter (artifacts)

Are more likely to appear in situations with more scatter (ie, large patients or wide-open collimation) or if the x-ray beam is not fully intercepted by the detector

If projections of detector electronics are visible on a patient image, options for a repeated image include using tighter collimation (decreasing the field of view), verifying good imaging geometry, or placing additional shielding behind the detector, such as a lead apron or plate to block the backscatter from striking the detector.

1. Brightness

All pixel brightness levels within the anatomy of interest should be neither completely white nor pitch black, but should possess along a broad range from very light to very dark gray *see through that area

2. Contrast and Gray Scale

Should be such that the number of details present in the image is maximized, and there is sufficient visual differentiation between adjacent details

3. Signal-to-Noise Ratio

High signal, low noise = a high ratio

Must be achieved in every image by; 

a. increasing the signal reaching the image receptor 

b. decreasing the electronic noise, scatter radiation, and all other forms of noise (artifacts) reaching the image receptor 

To ensure adequate penetration of the signal, sufficiently HIGH kVp must be used!

High SNRs are ideal for quality images 

4. Spatial resolution (sharpness of detail) 

should be apparent in electronic iamges:

• at least 8 LP/mm for static images

• at least 6 LP/mm for digital fluoroscopy

• Depends upon:

  • geometrical factors in the original projection

  • digital processing

  • Vertical and horizontal resolution of the display monitor

  • Zoom (magnification) level

5. Artifacts 

of all kinds must be absent

6. Shape Distortion 

Must be minimized, such that accurate representation of the anatomy of interest is achieved 

• Depends entirely upon geometry of original projection

7. Geometric Magnification (Size Distortion)

of the original image should generally be minimized

• strictly controlled by the SID/SOD ratio and related positioning

• Long SID; Short OID

• Measuring SID, minimizing OID

8. Display Magnification

of the digital image shound not be so extreme that the image becomes pixely where individual pixels become apparent and sharpness of details is lost

• Controlled by display screen size, matrix size, and field of view