Mammography Detector Ghosting and Image Lag QC

Ghosting and image lag are temporal artifacts: a faint remnant of one mammographic exposure that carries over into the next, either as leftover signal or as a local change in how the detector responds. In digital mammography — where the task is to find small, low-contrast calcifications and masses — a residual image from the previous patient can mimic a lesion or wash one out, which is why temporal-artifact testing belongs in every mammography QC program. ^{1, 3}

Digital mammography detectors are not perfect memoryless devices. Each exposure leaves a small physical trace in the detector that decays over seconds to minutes, and sometimes longer. If the next exposure is read out before that trace has cleared, the two images overlap. The physics behind this is charge trapping, and it behaves differently in the two dominant detector technologies: amorphous selenium (a-Se) direct-conversion detectors and cesium iodide / amorphous silicon (CsI/a-Si) indirect-conversion detectors. ^{2, 3}

This guide explains what image lag and ghosting are, the detector physics that produces them, how the ghost image ratio is measured in quality control, why these artifacts matter for breast cancer detection, and how a medical physicist evaluates and documents them under the ACR Digital Mammography QC Manual and MQSA. DRPS supports mammography facilities with this work through mammography physics and MQSA surveys and accreditation support across Florida, Maryland, Virginia, Washington DC, California, Nevada, and our other service areas.

Introduction

Lag and ghosting are easy to confuse because both appear as a faint version of the previous image, but they are physically different and are measured differently. Image lag is additive: a fraction of the charge generated by one exposure is not read out immediately and instead appears in later frames, adding a ghost of the first image on top of the second. Ghosting is multiplicative: the previous exposure changes the local sensitivity (gain) of the detector, so the same region now converts x-rays to signal at a slightly different rate. ^{2, 3}

In practice, a technologist or physicist sees the symptom — a faint outline of a prior breast, a block, or a phantom that should not be there — and the QC program has to quantify it, compare it to an action limit, and decide whether the unit is fit for clinical use. Because mammography is a screening examination performed on asymptomatic women, the tolerance for artifacts that could generate a recall or hide a cancer is low. ^{1, 6}

This article walks through the detector physics, the QC methodology, a worked ghost-image-ratio calculation, the clinical consequences, practical mitigation, and the regulatory framework that makes temporal-artifact testing a documented part of the medical physicist's annual evaluation.

Topic Explanation

What is image lag?

Image lag is the residual signal remaining in a detector after an exposure has ended and the frame has been read out. When x-rays interact in the detector, they create charge (electron-hole pairs in a-Se, or light in CsI that is then converted to charge in the photodiode). Ideally all of that charge is collected and read out in the frame corresponding to the exposure. In reality, some charge is trapped and released over time, so it appears in the first, second, and later subsequent frames. ³

Lag is usually described as a percentage: the signal in a frame following the exposure, divided by the signal of the exposure itself. Siewerdsen and Jaffray, in their classic characterization of an indirect-detection flat-panel imager, measured first-frame lag of approximately 2–10% depending on incident exposure, falling to roughly 0.7%, 0.4%, and 0.3% in the second, third, and fourth frames at 25% sensor saturation. They also found long-term image persistence — "ghosts" — that could last 30 minutes or longer. ³

What is ghosting?

Ghosting is an x-ray-induced change in detector sensitivity that causes image persistence. After a region of the detector is exposed, trapped charge alters how efficiently that region converts the next exposure into signal. On a subsequent uniform exposure, the previously exposed region reads out at a different level than its neighbors, producing a faint latent image of the earlier exposure even though no residual signal from the first exposure remains. ²

Zhao and colleagues showed that in a-Se detectors, charge trapping in the bulk of the selenium is the dominant source of ghosting: x-ray-generated electrons are trapped, and their subsequent recombination with x-ray-generated holes reduces x-ray sensitivity. The density of trapped electrons increases with accumulated exposure, and the amount of ghosting decreases as the electric field across the selenium is increased. ²

For quick reference, the two artifacts differ as follows:

Direct versus indirect detectors

The two mammography detector families trap charge in different places, so their temporal behavior differs:

• Direct-conversion (a-Se). X-rays are absorbed in a selenium photoconductor and converted directly to charge, which is drawn to the thin-film-transistor (TFT) array by a strong electric field (bias of roughly two to three kilovolts across a sub-millimeter selenium layer). Ghosting here is governed by bulk charge trapping in the selenium, and it is sensitive to the applied field. Charge trapping in the blocking layer at the a-Se/active-matrix interface also contributes to the detector's high-spatial-frequency response, which is why trapping is central to both resolution and temporal behavior. ^{2, 9}
• Indirect-conversion (CsI/a-Si). X-rays are absorbed in a cesium iodide scintillator that emits light, which an amorphous-silicon photodiode array converts to charge. Lag and persistence here arise from trapping in the a-Si photodiodes and in the scintillator (afterglow). ³

Both technologies are capable of excellent image quality — a-Se mammography detectors have demonstrated a presampled modulation transfer function (MTF) reaching 0.1 near 14 line pairs per millimeter and a detective quantum efficiency (DQE) peaking around 54–64% in prototype characterization — but both must manage trapped charge to keep temporal artifacts below the level that affects the clinical image. ^{4, 5}

Key Technical Principles

Quantifying lag

If $S_0$ is the signal generated by an exposure and $S_n$ is the residual signal measured in the $n$-th frame after readout, the frame lag is:

Because trapped charge is released over a spectrum of trap lifetimes, the residual signal decays as a sum of exponentials rather than a single rate:

where $a_i$ and ${\tau }_i$ are the fractional amplitude and release-time constant of each trap population. The slowest components are what produce long-lived persistence — the ghosts that survive many seconds after a normal inter-exposure interval. ³

Quantifying ghosting: the ghost image ratio

Ghosting is evaluated by deliberately creating a non-uniform exposure history and then reading the detector response to a uniform exposure. A common approach exposes the detector with a partial attenuator — for example a block or a half-field attenuator covering one part of the field — reads it out, then delivers a uniform ("flood") exposure. The ghost image ratio (GIR) compares the residual signal difference between the previously blocked and previously exposed regions to the signal of the uniform exposure:

A perfectly memoryless detector would give $\mathrm{GIR}=0$ because the two regions would respond identically to the uniform exposure. A nonzero value means the exposure history left a sensitivity (or residual-signal) imprint. The specific phantom geometry, timing, region-of-interest placement, and the pass/fail action limit are defined in the ACR Digital Mammography QC Manual and the manufacturer's QC procedures for the specific unit; the physicist follows the procedure the facility has adopted rather than a generic rule of thumb. ¹

Worked example

Suppose a physicist performs a ghost-image-ratio test. After a partial-field exposure and a uniform exposure, the mean pixel values in three regions of interest are:

• Previously exposed region: $S_{\text{prevexposed}}=1512$
• Previously blocked region: $S_{\text{prevblocked}}=1488$
• Dark (offset) level: $S_{\text{dark}}=50$

The uniform-exposure signal above offset in the previously blocked region is $1488-50=1438$. The ghost image ratio is:

A value of about 0.017 (1.7%) indicates only a slight residual sensitivity difference. The physicist compares this to the action limit in the adopted QC manual; a result far above the limit would trigger flat-field recalibration and, if it persisted, a service call. The point of the calculation is not the specific number but the method: a reproducible, documented measurement against a defined limit, repeated over time so drift is caught early. ¹

Why field strength and timing matter

Because ghosting in a-Se is reduced at higher applied electric field, and because lag decays with time, two levers reduce temporal artifacts without touching the hardware: an adequate inter-exposure interval, and keeping the detector's bias and calibration within specification. Zhao and colleagues' finding that ghosting falls with increasing field is the physical basis for manufacturers' bias and refresh strategies, and it is why a detector running out of specification can suddenly start showing artifacts that were previously absent. ²

Clinical Impact

In mammography, a temporal artifact is not a cosmetic nuisance — it can change a diagnosis. The clinical task is the detection of small, low-contrast objects: clustered microcalcifications a few hundred micrometers across, and subtle masses and architectural distortions. A faint ghost from a prior exposure competes directly with these findings.

The two failure modes are symmetric and both serious:

• False positive. A residual outline or a sensitivity artifact can be read as a density or a cluster, prompting additional views, ultrasound, or a biopsy for a finding that is not real. In a screening program, unnecessary recalls carry cost, anxiety, and procedural risk.
• False negative. A ghost that overlaps a real lesion can reduce its apparent contrast, and a sensitivity change can distort the local signal-to-noise ratio enough to push a subtle finding below the detection threshold.

Ghosting also propagates into other parts of the imaging chain. Flat-field (gain) calibration assumes the detector responds uniformly; a ghost imprint corrupts that assumption and can bake a faint artifact into every subsequent image until the next recalibration. Automatic exposure control that samples detector signal can also be perturbed, nudging technique and dose. Because mammography detectors are pushed to high spatial resolution and low noise — a-Se direct-conversion detectors reach an MTF around 0.55–0.64 at 5 line pairs per millimeter, higher than CsI/a-Si and far higher than computed-radiography phosphors, with correspondingly high DQE — the system is sensitive enough that even small temporal artifacts are visible against the intended image. ^{4, 5, 10}

Practical Optimization Tips

A practical approach to temporal artifacts in mammography QC follows a consistent sequence.

1. Separate lag from ghosting before you troubleshoot

Decide which artifact you are seeing. A residual signal in a dark frame taken immediately after an exposure is lag. A latent image that appears only when you deliver a new uniform exposure is ghosting (a sensitivity effect). The fix is different, so the diagnosis matters.

2. Respect the inter-exposure interval

Many apparent ghosting complaints are really a workflow problem: exposures taken faster than the detector clears. Confirm that routine clinical timing allows the detector to recover, especially after a high-exposure view.

3. Keep flat-field calibration current

Periodic gain (flat-field) recalibration is the single most effective routine control for ghost imprints. If artifacts appear, recalibrating often clears them; if they return quickly, the detector is trapping charge faster than expected and needs service.

4. Test after service and after detector swaps

Any time the detector or its bias supply is serviced or replaced, repeat the temporal-artifact test before returning the unit to clinical use. A new or repaired detector can behave differently.

5. Trend the numbers

Record the ghost image ratio (and any lag metric) over time. A value that is within limits but steadily rising is an early warning of detector degradation — far more useful than a single pass/fail snapshot.

Common pitfalls to avoid

• Confusing lag with ghosting and applying the wrong correction.
• Ignoring timing. A detector that passes with proper spacing can fail when exposures are rushed.
• Skipping the test after service. Detector work is exactly when temporal behavior changes.
• Treating a passing result as permanent. Trapped-charge behavior drifts with detector age, temperature, and bias.
• Using a generic limit. The action limit and method are unit- and manual-specific; use the procedure the facility has adopted. ¹

Regulatory Considerations

Digital mammography in the United States is regulated under the Mammography Quality Standards Act (MQSA), and QC — including detector artifact testing — is performed to the quality control manual the facility has adopted. Under MQSA, a facility must follow the QC program established by the image receptor manufacturer or an FDA-approved alternative such as the ACR Digital Mammography QC Manual. ⁶

• FDA MQSA, 21 CFR Part 900. Sets the federal quality standards for mammography, including the requirement that QC tests be performed and that failures be corrected before the affected component is used clinically, or within a defined period for other tests. ⁶
• ACR Digital Mammography QC Manual (2018). The current ACR manual defines the technologist, radiologist, and medical physicist tests, including the temporal-artifact (ghost image) evaluation, the method, and the action limits for units that have adopted it. The FDA has determined that facilities may use this manual for QC of both 2D and digital breast tomosynthesis applications of compatible units. ¹
• ACR–AAPM–SIIM Practice Parameter for Digital Mammography. Describes the determinants of image quality and the role of the qualified medical physicist in the acceptance testing and annual survey that bracket detector performance. ⁷
• IEC 62220-1-2. The international methodology standard for measuring DQE of mammographic detectors, which underpins the acceptance-testing metrics against which detector performance (and its degradation) is judged. ⁸

A board-certified medical physicist performs the artifact evaluation as part of the acceptance test and the annual survey, documents the result against the adopted action limit, and recommends recalibration or service when a unit fails. Of the states DRPS serves, X-ray mammography units are regulated by the FDA under MQSA and by the state radiation-control program; the physicist's survey satisfies both the federal and the state requirement. ^{6, 7} For the broader mammography QC picture, see our guides to mammography quality control and MQSA and ACR digital mammography phantom QC.

Frequently Asked Questions (FAQs)

What is the difference between image lag and ghosting in a mammography detector?

Image lag is residual signal: a fraction of the charge from one exposure is read out in later frames, so a faint copy of the previous image is added to the next one. Ghosting is a change in local detector sensitivity (gain) caused by trapped charge, so the region that was previously exposed responds differently to the next uniform exposure. Lag adds a signal; ghosting changes the gain.

What causes ghosting in amorphous selenium mammography detectors?

Trapping of charge in the bulk of the amorphous selenium layer. X-ray-generated electrons become trapped and then recombine with subsequently generated holes, reducing x-ray sensitivity in the previously exposed region. Ghosting decreases as the applied electric field across the selenium is increased, so detector design and bias matter. ²

How is the ghost image ratio measured in mammography QC?

The detector is exposed with a partial attenuator covering part of the field, read out, and then exposed uniformly. The ghost image ratio compares the residual signal difference between the previously blocked and previously exposed regions to the signal of the uniform exposure. The exact phantom, method, and action limit are specified in the ACR Digital Mammography QC Manual and the manufacturer's QC procedures. ¹

Is ghosting a patient safety problem or only an image quality problem?

Both. A residual or sensitivity artifact can mimic a lesion (false positive) or reduce the contrast of a real finding (false negative), and it can corrupt flat-field calibration and automatic exposure control. In a screening examination of asymptomatic patients, the tolerance for such artifacts is low.

How often is ghosting or image lag tested in digital mammography?

Temporal-artifact tests are part of the medical physicist's tests under the ACR Digital Mammography QC Manual, performed at least annually and after detector service, with manufacturer-specified routine checks in between. The exact frequency and limits follow the adopted QC manual.

Can ghosting be corrected without replacing the detector?

Often yes — with periodic flat-field recalibration, adequate time between exposures, and firmware corrections. Persistent ghosting above the action limit after recalibration points to a detector or bias problem requiring service, and the unit should not be used clinically until it passes.

Key Takeaways

• Lag and ghosting are different artifacts. Lag is residual signal added to the next frame; ghosting is a change in local sensitivity. The distinction drives the correct fix. ^{2, 3}
• Both come from trapped charge. In a-Se detectors, bulk charge trapping dominates ghosting and is reduced at higher applied field; in CsI/a-Si detectors, photodiode trapping and scintillator afterglow dominate. ^{2, 3}
• The ghost image ratio is the QC metric. A partial-field exposure followed by a uniform exposure reveals any sensitivity imprint; the method and limit come from the adopted QC manual. ¹
• The clinical stakes are real. A temporal artifact can create a false recall or hide a cancer, and it can corrupt calibration and AEC. ¹
• Routine controls work. Adequate inter-exposure timing and current flat-field calibration clear most artifacts; persistent failures mean service.
• It is a documented physicist test. Temporal-artifact evaluation is part of acceptance testing and the annual MQSA survey under the ACR DM QC Manual. ^{1, 6, 7}

Conclusion

Ghosting and image lag are the price of using a real, charge-trapping detector instead of an ideal one. The physics is well understood — trapped charge that releases as residual signal (lag) or that modulates local sensitivity (ghosting) — and it behaves predictably with exposure history, time, and applied field. What turns that understanding into safe imaging is a disciplined QC program: separate lag from ghosting, measure the ghost image ratio with a defined method and limit, trend the result, and act on drift before it reaches the clinical image. In mammography, where the diagnostic target is small and the consequences of a missed or false finding are large, that discipline is not optional. A board-certified medical physicist's acceptance test and annual survey are where it gets done and documented. ^{1, 6, 7}

How DRPS Can Help

Diagnostic Radiation Physics Services helps mammography facilities keep detector performance defensible and compliant. Our work includes mammography physics and MQSA surveys, acceptance testing of new and replacement detectors, temporal-artifact and image-quality evaluation, troubleshooting of ghosting and lag complaints, and accreditation support aligned with the ACR Digital Mammography QC Manual and FDA MQSA requirements.

DRPS supports facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, Nevada, New York, Pennsylvania, New Jersey, and Delaware. If your unit is showing a residual or latent image, or you are bringing a new detector online, a documented physicist evaluation is the fastest way to know whether it is fit for clinical use.

Related Resources

• Digital radiography detector lag and ghosting QC
• Mammography CNR and SDNR QC
• ACR digital mammography phantom QC
• Digital radiography detector: direct vs indirect conversion
• Detective quantum efficiency in digital radiography
• Mammography physics and MQSA surveys
• Accreditation support

References

1. American College of Radiology. 2018 ACR Digital Mammography Quality Control Manual. Reston, VA: ACR; 2018. acr.org
2. Zhao W, DeCrescenzo G, Kasap SO, Rowlands JA. Ghosting caused by bulk charge trapping in direct conversion flat-panel detectors using amorphous selenium. Med Phys. 2005;32(2):488-500. doi:10.1118/1.1843353. PubMed
3. Siewerdsen JH, Jaffray DA. A ghost story: spatio-temporal response characteristics of an indirect-detection flat-panel imager. Med Phys. 1999;26(8):1624-1641. doi:10.1118/1.598657. PubMed
4. Saunders RS, Samei E, Jesneck JL, Lo JY. Physical characterization of a prototype selenium-based full field digital mammography detector. Med Phys. 2005;32(2):588-599. doi:10.1118/1.1855033. PubMed
5. Marshall NW, Lemmens K, Bosmans H. Physical evaluation of a needle photostimulable phosphor based CR mammography system. Med Phys. 2012;39(2):811-824. doi:10.1118/1.3675403. PubMed
6. U.S. Food and Drug Administration. Mammography Quality Standards Act (MQSA) regulations, 21 CFR Part 900. fda.gov
7. American College of Radiology, American Association of Physicists in Medicine, Society for Imaging Informatics in Medicine. ACR–AAPM–SIIM Practice Parameter for Determinants of Image Quality in Digital Mammography. acr.org
8. International Electrotechnical Commission. IEC 62220-1-2: Medical electrical equipment — Characteristics of digital X-ray imaging devices — Determination of the detective quantum efficiency — Detectors used in mammography. Geneva: IEC. iec.ch
9. Zhao W, Ji WG, Debrie A, Rowlands JA. Imaging performance of amorphous selenium based flat-panel detectors for digital mammography: characterization of a small area prototype detector. Med Phys. 2003;30(2):254-263. doi:10.1118/1.1538233. PubMed
10. Marshall NW, Monnin P, Bosmans H, Bochud FO, Verdun FR. Image quality assessment in digital mammography: part I. Technical characterization of the systems. Phys Med Biol. 2011;56(14):4201-4220. doi:10.1088/0031-9155/56/14/002. PubMed