ABSNM NMPI common mistakes
ABSNM NMPI · Understand the traps and how to avoid them.
When asked how a change in photomultiplier tube (PMT) gain or high voltage shifts the pulse-height spectrum, the candidate reasons qualitatively ("it gets bigger") instead of stating that increasing PMT gain/high voltage shifts the photopeak to a higher channel (higher pulse amplitude) without changing the number of detected gamma events.
Gain is a multiplicative amplifier property, but candidates conflate 'more amplification' with 'more counts' and never separate pulse amplitude (energy axis) from event rate (count axis).
What to do instead
Practise the rule 'gain moves the peak along the energy axis; activity moves the height of the peak'. For each electronics stage (PMT gain, preamp, amplifier coarse/fine gain, HV), write one sentence on whether it changes amplitude, count rate, or both.
The candidate computes energy resolution as a percentage of the wrong quantity, e.g. dividing FWHM by the full peak height or by the low-energy cutoff, or reporting FWHM in keV without converting to percent of the photopeak energy.
The formula (FWHM/peak energy x 100) is memorised as a string of symbols without attaching each symbol to a feature on a spectrum, and candidates rush the arithmetic under time pressure.
What to do instead
Drill on worked spectra: locate the photopeak centroid, measure FWHM at half maximum in channels, convert channels to keV using the calibration factor, then divide by centroid energy and multiply by 100. Sanity-check that a Cs-137 photopeak at 662 keV gives a resolution in the 6-10% range for a NaI system.
On dead-time and count-loss questions the candidate applies the observed (measured) count rate in the correction formula instead of the true rate, or ignores the paralyzable versus nonparalyzable distinction and assumes a single correction equation applies to all systems.
The correction equation is recalled by pattern-matching rather than by understanding that the model (paralyzable vs nonparalyzable) determines which expression is valid and that the true rate is the unknown being solved for.
What to do instead
Write out both models side by side (nonparalyzable: n = m/(1 - mtau); paralyzable: m = ne^(-n*tau)) and state which variable is measured versus true. Work one problem of each type to the numerical answer, and check that the corrected rate is always greater than the observed rate.
The candidate treats energy resolution and detection efficiency as interchangeable figures of merit, or claims that improving energy resolution automatically increases sensitivity, when resolution describes peak sharpness and efficiency describes the fraction of emitted radiation detected.
Both are quoted as percentages and both improve with 'better' detectors, so candidates collapse them into one concept and pick the distractor that sounds most like 'better performance'.
What to do instead
Build a two-column table: resolution (FWHM%, intrinsic vs system, affects scatter rejection and photopeak separation) versus efficiency (absolute vs intrinsic, depends on detector thickness, density, and geometry). Answer any comparison question by naming which quantity the stem is describing before choosing.
When asked what a specific gamma camera component does, the candidate assigns the function to the wrong element, e.g. saying the collimator determines energy or that the photomultiplier tubes provide spatial positioning by themselves without the positioning logic.
The signal chain (collimator -> crystal -> light guide -> PMT array -> positioning/preamp -> amplifier -> pulse-height analyser -> computer) is memorised as an ordered list of names rather than as a causal chain where each stage transforms a specific property.
What to do instead
Rehearse the chain as a story: collimator restricts the geometric origin of photons, crystal converts energy to light, PMTs convert light to charge, positioning logic computes X/Y from relative PMT signals, PHA accepts or rejects by energy. For each component, state its input and output in one line.
The candidate confuses spatial resolution with sensitivity when evaluating collimator choices, asserting that a high-resolution (long, small-hole) collimator also improves sensitivity, or that a high-sensitivity collimator improves resolution.
Both collimator properties are described as 'better when the design is optimised', so candidates assume all performance metrics improve together and miss the inherent trade-off.
What to do instead
Memorise the trade-off explicitly: longer septa and smaller holes improve resolution but reduce sensitivity; the reverse holds for high-sensitivity designs. When a stem changes one collimator parameter, predict the direction of change in both resolution and sensitivity before reading the options.
On uniformity and flood-field QC questions the candidate judges a flood image by eye as 'acceptable' without applying the specified quantitative criterion (integral and differential uniformity computed over the useful field of view), or confuses intrinsic with extrinsic uniformity testing.
The visual impression of a smooth flood image feels sufficient, and the distinction between intrinsic (collimator removed) and extrinsic (collimator in place) testing is not linked to which uniformity specification applies.
What to do instead
State the test definition before answering: intrinsic uniformity uses a point source with the collimator removed; extrinsic uses a flood source with the collimator on. Quote integral and differential uniformity as percentages over the useful field of view and compare against the stated acceptance limit rather than describing the image qualitatively.
The candidate answers SPECT reconstruction questions by naming a filter or algorithm without linking it to the effect on image noise and resolution, e.g. stating that a ramp filter alone produces clinically acceptable images or that more iterations always improve image quality.
Filter and iteration names are memorised as vocabulary, and the candidate does not connect cutoff frequency, filter order, or iteration count to the noise-resolution trade-off that the question is actually testing.
What to do instead
For every filter, state its effect in one sentence: ramp = sharp but noisy; Butterworth/Gaussian = suppresses high-frequency noise at the cost of resolution; cutoff frequency controls where suppression begins. For iterative reconstruction, note that too few iterations leave the image un-converged and too many amplify noise, then answer in terms of the trade-off.
On PET instrumentation questions the candidate mixes up the roles of the scintillator, the photodetector, and the coincidence/positioning electronics, or misstates what a time-of-flight or randoms correction actually measures.
PET is taught as a block of physics and electronics at once, and candidates memorise acronyms (LOR, TOF, randoms, scatter) without attaching each to the physical event it describes.
What to do instead
Map each term to an event: annihilation produces two 511 keV photons; a coincidence is two detectors firing within the timing window; a line of response is the line between them; randoms are uncorrelated coincidences; TOF uses arrival-time difference to localise along the LOR. Answer by identifying which event the stem describes.
The candidate treats QA frequency and tolerance limits as generic ('daily', 'within a few percent') instead of recalling the specific test, its required frequency, and its acceptance criterion, and cannot distinguish a regulatory requirement from a manufacturer recommendation.
QA schedules are memorised as a list of test names without the paired frequency and tolerance, and the candidate does not separate mandated performance standards from institutional or vendor protocols.
What to do instead
Build a three-column table (test, frequency, tolerance) for the camera and for the dose calibrator, and mark which entries come from a recognised performance standard versus local policy. When a stem asks 'how often' or 'what limit', answer only from the paired table entry.