A breast imaging unit needs different test objects for its weekly check, its annual survey and its tomosynthesis mode. Which phantom answers which question, and what else the programme needs besides phantoms.
Breast imaging is unforgiving of small losses in image quality in a way that most radiography is not. The features that matter are micro-calcifications a few hundred micrometres across and low contrast masses that sit only slightly above the surrounding tissue. A detector that has drifted, a filter that has been changed, an automatic exposure control that is reading a different thickness than it thinks: none of those produce an image that looks broken. They produce an image that looks fine and no longer shows the smallest detail it used to.
That is why a mammography quality control programme is built around scoring, not around inspection. Somebody counts details, writes the number down, and the number is compared with last month's. The phantom set is what makes that count mean anything.
Mammography test objects answer four separate questions and it is worth keeping them apart, because departments routinely buy one object and expect it to cover all four.
A routine weekly or monthly test answers the first two quickly. An annual or commissioning survey answers all four with more effort. Buying for the annual survey and then using it weekly is how a programme quietly stops being done.
For a department that will do one test per week or per month, the routine objects are the right purchase. TOR MAS is supplied with a 7 cm attenuator stack and covers sensitometry, resolution limit, low and high contrast detail detectability and simulated micro-calcifications, so a single score can be recorded at each session. TOR MAX is the same idea with two resolution limit patterns rather than one, which is useful where the system is being compared against another unit and the single pattern is not discriminating.
Neither of those produces an image that looks like a breast, and that is deliberate: a geometric target is easier to score consistently. Where a more natural looking image is wanted alongside the geometric one, TOR MAM supplements them. One half carries filaments, micro-particles and low contrast details, and the other simulates breast tissue with micro-calcifications, so a radiographer can see what a real detail looks like at the current performance level rather than only counting discs.
Scoring is where routine programmes fail in practice, and for a predictable reason: two observers disagree, an observer's threshold changes with the reading conditions, and the record becomes noisy enough that a genuine drift hides inside it. AutoPIA is the answer to that. It recognises and analyses test images of Leeds Test Objects phantoms automatically, calculates reproducible quality indices, stores the results and plots control charts with baseline and threshold values, so a trend is visible as a trend rather than as a column of numbers in a logbook. For a Bangladeshi department where the radiographer doing the weekly test changes with the roster, that is the difference between a programme and a filing exercise.
Commissioning and annual work needs a full set. DMAM2 is built to the fourth edition European guidelines and brings together a threshold contrast detail plate, filaments, a graded set of PMMA plates, mesh and grid targets, a steel straight edge, a resolution pattern and the filters needed for homogeneity and ghosting tests. Once the department has that set, DMAM2 Gold is available separately as just the gold contrast detail plate, carrying 78 details in six sizes across a wide contrast range, which is what a routine threshold contrast score is read from without unpacking the whole case.
For the physical measurements, PIXMAM is a stack of finely toleranced PMMA plates, one of which carries an encapsulated aluminium foil, used for automatic exposure control checks, detector homogeneity, signal to noise ratio and contrast to noise ratio. The graded stack is the important part. Presenting the system with a series of thicknesses is the only way to find out whether the exposure control behaves across the range of patients it will actually see, rather than at one convenient thickness.
One point worth stating plainly for departments building a programme from scratch. Signal to noise and contrast to noise are computed from pixel values in a defined region of the raw image, so the department needs access to unprocessed images and a means of drawing a region of interest on them. If the workstation only presents processed images for display, those measurements cannot be made, and that limitation should be discovered during acceptance rather than a year later. The same argument for testing what the department can actually access applies across nuclear medicine, and is set out in choosing QA phantoms for a nuclear medicine department.
A tomosynthesis system reconstructs a stack of planes from projections taken across a limited angular range. That introduces two things a two dimensional phantom cannot test: how well the system resolves in the depth direction, and whether the reconstruction is geometrically faithful.
PIXMAM 3D is the routine answer. It adds a plate holding twenty five embedded spheres for the Z resolution test to the standard PIXMAM stack, so geometric distortion and image resolution can be checked alongside the usual physical measurements. For commissioning and annual work, TOMO IQ is the complete set, designed to the EUREF breast tomosynthesis protocol, combining the graded PMMA stack and Z resolution plate with a TORMAM phantom, an NPS aluminium filter, an MTF steel edge and x-ray rulers.
The practical warning about tomosynthesis testing is time. A full EUREF style survey on a combined two dimensional and tomosynthesis unit is a long session, and it needs the room booked, the physicist present and the system out of clinical use. Departments in Bangladesh that add tomosynthesis to an existing unit frequently do not add anything to the physics time budget, and the annual survey then gets truncated at the point where the tomosynthesis tests would have started. Book the time before buying the phantom.
Several tests in a breast imaging programme have nothing to do with image quality targets, and they are the ones most often missing from a Bangladeshi department's equipment list.
Two items that look like packaging are worth budgeting for. Protective Cases with foam inserts are available separately, and in a service that shares one physicist across several sites, phantoms travel by road in heat and humidity and get dropped. And a damaged PMMA plate is not repairable: a scratch on a contrast detail plate becomes a permanent artefact scored as a detail, which quietly biases every result afterwards.
Finally, decide who owns the programme before the equipment arrives. The weekly count is a radiographer task and it takes minutes. The annual survey is a medical physicist task and it takes a day. If the department has no physicist of its own, that arrangement has to be contracted rather than assumed, because a phantom set with nobody scheduled to use it is the commonest way a quality control programme fails, and it fails silently.