8.3 Physicist Tests: MEE, Collimation, Spatial Resolution/MTF, SNR/CNR, AEC, AGD, kVp, HVL

What ARRT wants from you on physicist tests

You will not be asked to operate an ion chamber like a medical physicist. You will be asked why the physicist is there and how often. Image Production D is explicit: mammographer tests are detailed; physicist tests are purpose and frequency, referenced to the 2018 ACR Digital Mammography QC Manual. If the facility uses manufacturer QC, the physicist's list still has to be substantially the same as that manufacturer's program — tests, frequencies, and action limits can differ from ACR 2018.

Artifact evaluation, room illuminance, review of the technologist QC program, compression application, paddle alignment, and workstation QC continue in 8.4. This section is the beam, detector performance, and dose core: mammography equipment evaluation (MEE), collimation, spatial resolution and modulation transfer function (MTF), signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), automatic exposure control (AEC), average glandular dose (AGD), kVp, and half-value layer (HVL).

Mammography equipment evaluation versus the annual survey

MEE — purpose. An MEE is the full physics evaluation that asks, before patients, whether a unit (or a newly repaired major component) meets MQSA equipment standards and the QC program's pass criteria. Think of it as acceptance testing plus regulatory checkout.

MEE — frequency. Perform an MEE when a unit is installed, when it is disassembled and reassembled in the same or a new location (including used units and room-to-room moves), and when major components are changed or repaired. ACR and FDA language is blunt: all problems must be corrected before the equipment is used for examinations. There is no clinical workaround of imaging a few patients while waiting for a part.

Annual survey — purpose. Confirm that the unit that passed MEE is still performing, that technologist QC is actually being done, and that dose and image-quality metrics have not drifted. The physicist issues a written report to the responsible physician and to the people who service the equipment.

Annual survey — frequency. At least annually. Mobile units have extra post-move checks (MQSA 21 CFR 900.12(e)(7)) before imaging at a new site; that is not a substitute for the annual survey.

ACR 2018 uses the MEE as a baseline. Annual SNR and CNR are compared with MEE values for consistency (training materials use a 15% window versus baseline or the prior survey). Manufacturer CNR recipes differ; do not import a GE-only formula into an ACR stem.

Collimation assessment

Purpose. The x-ray field must cover the chest wall so posterior tissue is not clipped, yet it must not flood the room or miss the detector. Light-field alignment, if provided, must match the x-ray field. Collimation failure is a missed cancer on the chest wall or extra dose to the lung and stomach.

MQSA geometry (21 CFR 900.12). Beam limitation must let the entire chest-wall edge of the x-ray field extend to the chest-wall edge of the image receptor, and the field must not extend beyond any edge of the receptor by more than 2% of the source-to-image distance (SID). If a light field is provided, total misalignment along length or width must not exceed 2% of SID. Paddle alignment (1% SID) is in 8.4 but is measured in the same geometry session.

Worked example. SID is 65 cm. Two percent of SID is 1.3 cm. An x-ray field that extends 1.5 cm beyond a receptor edge fails the 2% test.

Frequency (ACR 2018). ACR training moved collimation to MEE-only for 2D, but DBT collimation stays annual because many tomosynthesis systems use dynamic or active collimation that can drift. Manufacturer programs may still require annual collimation on 2D. Read the stem.

Spatial resolution and MTF

Purpose. Spatial resolution is the system's ability to show fine detail — the difference between a sharp calcification cluster and a blur. Modulation transfer function (MTF) is the physicist's curve of how much contrast remains as spatial frequency (line pairs per millimeter) increases. A system can look pretty on large masses and still fail fine-detail transfer.

How it is done (basic). The physicist may use a line-pair plate, an edge (to compute MTF), or manufacturer software. Screen-film limiting resolution (historically about 11 lp/mm contact and 13 lp/mm magnification) is not the universal digital pass number. Digital criteria are program-specific. For DBT, physicists also measure z-resolution and volume coverage; those are annual ACR 2018 tests that answer whether slices are real and whether the reconstructed volume covers the breast support.

Frequency. Physicist test at MEE and typically again at the annual survey for spatial resolution in the ACR 2018 physicist set. Know that it is not a technologist weekly task.

SNR and CNR (low-contrast performance)

Purpose. SNR asks whether the useful signal is large compared with random noise. CNR asks whether a low-contrast object (the mass-like disk in the phantom) stands out from background noise. Digital mammography dies from noise as often as from blur: an underexposed AEC cell can still look gray on the AWS while CNR has collapsed.

Frequency and ACR 2018 idea. Measure at MEE (establish a lower-limit SNR at a reference thickness, often 4 cm, and a CNR from the phantom disk). At the annual survey, remeasure and compare with the MEE baseline. ACR training describes keeping annual values within about 15% of the reference rather than chasing vendor-specific correction factors. Technologists do not compute CNR as part of weekly phantom scoring; they score objects and artifacts.

AEC system performance

Purpose. Automatic exposure control should choose mAs (and often kVp/filter) so a 2 cm fatty breast and an 8 cm dense breast both land at appropriate detector exposure. If AEC is deaf, thin breasts are overdosed and thick breasts are noisy.

Frequency. MEE is the thorough test: ACR 2018 training describes imaging about 2, 4, 6, and 8 cm plus a magnification thickness, with a pass/fail SNR floor at the 4 cm reference. Annual testing checks that those SNRs still match the MEE within the consistency window. Manufacturer AEC tests may still talk in optical density from the film era; digital stems should talk SNR/CNR or the vendor's exposure index, not OD 1.20.

Average glandular dose

Purpose. AGD estimates dose to glandular tissue, the tissue at risk. It is the regulatory dose metric for mammography, not entrance skin exposure alone.

Performance criterion (MQSA). 21 CFR 900.12(e)(5)(vi): the AGD for a single cranio-caudal view of an FDA-accepted phantom simulating a standard breast shall not exceed 3.0 mGy (0.3 rad) per exposure, using clinical technique factors for a standard breast. That limit applies to digital systems as well; other-modality QC cannot exceed the screen-film dose cap.

Frequency. Physicist measurement at MEE and at least annually. ACR 2018 uses a calculation method (Dance) that covers 2D and DBT and different target-filter pairs. A weekly technologist phantom pass does not prove AGD is under 3.0 mGy; dose is a physics calculation from output and beam quality.

Worked example. Standard-phantom AGD is 3.1 mGy on the CC technique used clinically. That exceeds 3.0 mGy. Dosimetry is on MQSA's list of results that require corrective action before further examinations using the failed component (21 CFR 900.12(e)(8) citing (e)(5)(vi)). The unit does not image patients until technique or filtration is corrected and dose is remeasured.

kVp accuracy and reproducibility

Purpose. Selected kilovoltage must be the kilovoltage delivered. Contrast and dose both move when kVp is wrong. Reproducibility asks whether the same station fires the same kVp every exposure.

MQSA numbers (know these as the federal equipment standard). Accuracy within ±5% of indicated kVp at the lowest measurable clinical kVp, the most commonly used clinical kVp, and the highest available clinical kVp. At the most commonly used clinical kVp, the coefficient of variation for reproducibility shall be ≤ 0.02.

Worked example. Selected 28 kVp, measured 26.5 kVp. Relative error is (28 − 26.5) / 28 = 5.4%, which is outside ±5%.

Frequency (ACR 2018). MEE only for kVp (a reduction from annual film-era testing). Manufacturer QC may still require periodic kVp. Either way, it is a physicist test, not a weekly tech test.

HVL and beam quality

Purpose. Half-value layer is the thickness of aluminum that cuts beam intensity in half. Too little filtration means a soft beam, extra skin dose, and a beam that does not match the dose tables. MQSA sets a minimum HVL, not a beauty contest for the highest HVL (excessively hard beams can cost subject contrast).

MQSA table (designed operating range below 50 kV). Minimum HVL is 0.20 mm Al at 20 kV, 0.25 mm Al at 25 kV, and 0.30 mm Al at 30 kV, with interpolation in between — roughly kVp/100 in millimeters of aluminum. The physicist measures this with aluminum sheets and a dosimeter.

Frequency (ACR 2018). MEE only, paired with kVp. Annual surveys still need a legal beam, but ACR 2018 does not make HVL a yearly busywork item unless the program or a repair puts it back in scope. After a tube or filter change, expect an MEE that includes HVL.

Purpose and frequency snapshot

If a question says manufacturer QC, the right-hand column may change; the purposes and the 3.0 mGy dose cap do not.