Building a gamma camera quality assurance programme that survives an inspection

A working QA schedule for a gamma camera department: what the daily flood and photopeak checks actually detect, how weekly resolution and linearity tests differ, and how to keep records a regulator will accept.

A gamma camera that has been imaging patients for three years usually has a folder of printed flood images somewhere near the console. That folder is not a quality assurance programme. A programme is a written schedule, tolerances agreed in advance, a named person responsible for each test, and a record complete enough that somebody can reconstruct what the detector was doing on any given day. The difference shows up on two mornings: the one when an inspector asks to see your records, and the one when a cardiologist tells you a perfusion defect looks wrong and you have to decide whether the problem is the patient or the detector.

Acceptance testing sets the baseline for everything after it

The measurements taken when a camera is first installed are the only ones you will ever make on a system that is known to be at its best. Every result for the next ten years is compared against that baseline, so it is worth doing properly, and worth doing with your own medical physicist in the room rather than only the vendor engineer.

The NEMA NU 1 standard defines the measurement conditions for gamma camera performance, which is what makes numbers comparable between systems, between vendors and between years. The parameters that matter most in routine practice are intrinsic and system spatial resolution, energy resolution, integral and differential uniformity over the useful and central fields of view, system planar sensitivity, count rate performance, multiple window spatial registration, and for SPECT systems the centre of rotation and tomographic uniformity. The IAEA Human Health Series guidance on quality assurance for SPECT systems is a useful companion, because it explains what to do when a measurement fails rather than only how to take it.

• Witness the tests. Acceptance is a contractual event. If the hospital physicist did not see the measurement being taken, the hospital has bought a certificate, not a baseline.
• Keep the raw data. Ask for the acquisition files, not only the printed report. When a fault appears in year four you will want to reprocess the original flood, and the console may have been replaced by then.
• Record the conditions. Isotope, activity, source distance, collimator, energy window, matrix size, counts and room temperature. A uniformity figure without its conditions is not a measurement.
• Agree tolerances in writing. Decide at acceptance what change from baseline triggers a service call, and put the numbers in the QA manual before anyone has a reason to argue about them.

What the daily tests are actually detecting

Photopeak and energy window

The first check of the day confirms that the energy window is still centred on the photopeak. Photomultiplier gain drifts with temperature, and in a room where the air conditioning was off overnight the detector starts the morning warmer than it finished. A window that has walked off the peak reduces counts, changes the scatter fraction and degrades contrast, and none of that is obvious on a clinical image until it is severe.

Uniformity flood

The daily flood is the most informative single picture you can take of a gamma camera, because almost every fault appears in it. A failing photomultiplier tube shows as a cold or hot patch roughly the size of the tube. A cracked crystal shows as a dark line. Moisture ingress into the crystal assembly shows as a blotch that spreads slowly week by week, which is exactly why flood images have to be kept and compared rather than glanced at and thrown away. A mis-set energy window makes the whole field look mottled.

There are two ways to acquire it. An intrinsic flood is taken with the collimator removed and a point source at distance, and it tests the detector alone. An extrinsic flood is taken with the collimator fitted and a flood source across the face, and it tests the detector plus the collimator you actually image with. A refillable flood phantom such as the Leeds Test Objects Gamma Flood Phantom, filled with technetium eluate, is the practical choice for a department that already runs a generator, because it avoids importing and later disposing of a sealed sheet source.

Record both integral uniformity and differential uniformity, over the useful field of view and the central field of view. Integral uniformity catches a large slow change. Differential uniformity catches a sharp local one, which is the pattern a dying tube makes, and it is usually the number that moves first.

Weekly and monthly: resolution, linearity and centre of rotation

Spatial resolution

A bar pattern phantom such as the Leeds Test Objects Gamma Resolution Phantom gives a visual, scored assessment of how fine a detail the system can still separate. Acquire it the same way every week, in the same orientation, with the same counts, then rotate it by ninety degrees the following week so that both axes get tested over the month. Scoring is subjective, so have the same two people do it and record who scored it.

Linearity and distortion

Resolution tells you how small a detail you can see. Linearity tells you whether what you see is in the right place. A camera can pass a resolution check and still bend straight lines, and that quietly damages SPECT reconstruction because the projection data no longer maps correctly into the volume. An orthogonal pattern such as the Gamma Distortion Phantom makes the fault visible: you are looking for curvature, and for lines that are straight in one region and bowed in another, which points at the positioning circuit or at a correction map that is out of date.

Centre of rotation

A centre of rotation error produces a ring artefact in reconstructed slices, and it is one of the few faults a technologist will blame on the patient. Check it monthly with a point source, with the collimator you use clinically fitted, and repeat it after any mechanical work on the gantry or after the heads have been moved for service. On dual head systems, check head to head registration at the same time.

Quarterly and annual work

• Tomographic image quality. A SPECT image quality phantom such as SPECT IQ tests the whole chain: acquisition, reconstruction filter, attenuation correction, scatter correction and display. A planar flood cannot detect a bad reconstruction setting.
• System sensitivity. Counts per unit activity, measured per collimator. A gradual fall points at collimator damage or detector ageing, and it is the number that tells you whether your acquisition times are still appropriate.
• Uniformity correction map update. Follow the manufacturer procedure and the manufacturer interval. Updating a correction map to hide a failing tube is a well known way to lose a detector.
• Collimator inspection. Physical examination for dents and septal damage, plus a flood taken through every collimator in the store, not only the one currently on the head.
• Mechanical and safety checks. Couch indexing accuracy, collision sensors, emergency stops, radial motion, and the physical condition of cabling at the drag chain.
• Repeat measurements against acceptance. Once a year, repeat a defined subset of the NEMA style acceptance measurements so that ageing is quantified rather than assumed.

A schedule that a real department can staff

A QA programme collapses when it has been designed for a department with a full time physicist. Build it around the staff you actually have, and keep the daily items short enough to survive a busy morning list.

Frequency	Test	Phantom or tool	A failure usually means
Daily	Photopeak and energy window	Point source or a dispensed vial	PMT gain drift, overnight temperature swing
Daily	Extrinsic uniformity flood	Gamma Flood Phantom	Failing PMT, crystal damage, wrong energy window
Weekly	Spatial resolution	Gamma Resolution Phantom	Collimator damage, detector tuning drift
Weekly	Contamination and background survey	Survey meter and wipe tests	Spill, leaking vial, contaminated collimator
Monthly	Linearity and distortion	Gamma Distortion Phantom	Positioning circuit drift, correction map out of date
Monthly	Centre of rotation and head registration	Point source, SPECT acquisition	Gantry wear, head alignment error
Quarterly	Tomographic image quality	SPECT IQ phantom	Reconstruction, attenuation correction or collimator fault
Annual	Sensitivity, correction map update, mechanical safety	Manufacturer procedure with physicist	Cumulative detector and collimator ageing

Documenting the programme so an inspector accepts it

A nuclear medicine department in Bangladesh operates under a licence issued by the Bangladesh Atomic Energy Regulatory Authority, and the practical test of your records is simple: could an inspector who has never met your staff follow them without you in the room. If the answer needs a verbal explanation, the record is incomplete.

• The test name, and a reference to the written procedure it follows.
• Date, time, operator name and signature. Initials on a sheet with no signature list anywhere is a common gap.
• Isotope, activity and phantom identity, including the phantom serial number where it has one.
• The numerical result, not only a tick in a pass column. A tick records an opinion; a number records a measurement.
• The tolerance that was in force on that date, because tolerances change when equipment is serviced or upgraded.
• The action taken when a result was out of tolerance, who was informed, and the result of the repeat test.
• An annual review page signed by the medical physicist, confirming the programme was followed and the tolerances are still appropriate.

Two things fail an inspection more often than a bad measurement: unsigned pages, and an out of tolerance result with nothing written after it. A failed test that was investigated, fixed and re-tested is evidence that the programme works. A failed test followed by silence is evidence that nobody was reading the sheets.

If the only person who can explain your QA file is the person who wrote it, you do not have a QA file. You have a private notebook.: A recurring finding at radiation safety inspections

Where QA programmes quietly fail in Bangladesh

• Interrupted acquisitions. A flood that lost forty seconds to a power cut and was restarted has fewer counts than the protocol requires, so the figure it produces is noise. Record the interruption and repeat the test.
• Air conditioning switched off overnight. The saving on electricity is small. The cost is a detector that starts every morning in a different thermal state, which appears as photopeak drift and uniformity variation that nobody can explain.
• Humidity. Sodium iodide crystals are hygroscopic and the assembly is sealed, but seals age, and monsoon humidity combined with daily temperature cycling is what breaks them. Log room temperature and relative humidity, and keep the log with the QA file so that a crystal failure can be traced.
• One trained person. When the only technologist who knows the procedure goes on leave, the schedule stops. Train at least two, and write the procedure at a level a competent newcomer can follow.
• Tolerance drift. A result creeps out of tolerance, somebody widens the tolerance instead of calling service, and six months later the detector is beyond economical repair. Changing a tolerance should require the physicist to sign, with a reason.

Back to all Insights