Digital Radiography MTF: The Edge Method

The modulation transfer function (MTF) is the single number-versus-frequency curve that tells you how faithfully a digital radiography detector reproduces fine detail, and the angled-edge method is the most practical way to measure it. From one image of a tilted, highly attenuating edge, a physicist can reconstruct the presampled MTF, compare it to baseline, and catch resolution loss long before it is visible on a clinical radiograph. ^{1, 3}

Spatial resolution is not a single number you can read off a spec sheet. A detector blurs an input in a way that depends on spatial frequency: coarse structures pass through almost untouched, while fine structures are attenuated or lost. The MTF captures exactly that frequency dependence, and measuring it correctly is a core skill in acceptance testing and routine quality control (QC). ^{1, 9}

This guide explains what the MTF is, why the edge method has largely replaced the slit method for digital radiography (DR), how the measurement is performed step by step, a worked Nyquist calculation, the clinical stakes, the practical pitfalls that corrupt a measurement, and how the MTF fits into the broader detective quantum efficiency (DQE) framework defined by IEC 62220-1-1. DRPS provides this analysis as part of its diagnostic radiography physics testing and accreditation support services.

Introduction

Every imaging detector is a low-pass filter. When an infinitely sharp line or edge is presented to a DR detector, the recorded image is not sharp: the signal spreads across neighboring pixels because of x-ray scatter in the converter, light diffusion in indirect-conversion detectors, charge sharing, and the finite size of the pixel aperture. The MTF quantifies how much of each spatial frequency survives this blurring. ^{1, 3}

Because the MTF is defined in the spatial-frequency domain, it connects directly to the question clinicians care about: can the system render the small, low-contrast structures that drive diagnosis, such as fine trabecular bone, small catheters and wires, or microcalcification-scale detail? A detector can pass a subjective "looks sharp" review and still show a measurable MTF drop that signals a degrading converter, a coil of charge-trapping, or a reconstruction change. ^{7, 9}

MTF testing belongs in two places: at acceptance, where it establishes the baseline against which the detector will be judged for its lifetime, and in periodic QC, where it is trended for drift. The edge method makes both practical because it needs only a small, robust test device and a single exposure at a clinically meaningful dose. ^{3, 10}

Topic Explanation

What the MTF actually represents

The MTF is the magnitude of the detector's spatial-frequency response, normalized to 1.0 at zero frequency. If you imaged a series of sinusoidal input patterns of increasing spatial frequency, the MTF at each frequency would be the output modulation divided by the input modulation. ^{1, 5}

In practice we do not expose the detector to true sinusoids, because producing sinusoidal x-ray patterns is difficult. Instead we measure the system's response to a sharp input and transform it into the frequency domain. Two equivalent inputs are used:

a narrow line, giving the line spread function (LSF) directly; and
a sharp edge, giving the edge spread function (ESF), whose derivative is the LSF.

The MTF is then the normalized magnitude of the Fourier transform of the LSF. ^{3, 4}

Presampled MTF versus the sampling problem

A digital detector does two things at once: it blurs the input (a continuous process), and it samples the blurred result onto a discrete pixel grid (a discrete process). The presampled MTF describes the blur and the pixel aperture before the effect of the sampling grid is folded in. It is the quantity acceptance testing reports because it characterizes the detector hardware rather than the particular phase at which a structure happens to land on the pixel lattice. ^{3, 4}

The sampling grid imposes a hard ceiling, the Nyquist frequency, above which frequencies cannot be represented correctly and instead alias back to lower frequencies. If you measured the MTF by simply sampling an on-axis edge at the pixel pitch, you could not see past Nyquist at all, and aliasing would distort what you did see. The angled-edge (and angled-slit) technique solves this by synthesizing a finely oversampled profile, so the presampled MTF can be characterized out to several times Nyquist. ^{3, 4}

For the companion noise and efficiency metrics that complete detector characterization, see our guide to detective quantum efficiency in digital radiography.

Key Technical Principles

The geometry of the angled edge

The core trick of the edge method is deliberate misalignment. An opaque, polished edge is placed on the detector and tilted a few degrees — typically about 1.5° to 3° — relative to the pixel rows or columns. Because the edge is slightly tilted, each successive row of pixels crosses the edge at a slightly shifted position. Projecting all rows onto an axis perpendicular to the edge interleaves their samples into a single, finely spaced oversampled ESF. ^{1, 3}

The oversampling factor is set by the tilt. If the edge is tilted by an angle $θ$ from the pixel columns, each row shifts by $tan θ$ of a pixel relative to the next, so the synthesized sample spacing is

Δ x_{eff} = a \cdot tan θ

where $a$ is the native pixel pitch. A tilt of about 2.3° gives $tan θ \approx 0.04$ , which is roughly a 25-fold increase in sampling density:

\frac{a}{Δ x _{eff}} = \frac{1}{tan θ} \approx \frac{1}{0.04} = 25.

That is why a small, well-measured tilt angle is essential: too small an angle gives too few interleaved samples, and too large an angle (or a poorly estimated angle) smears the synthesized ESF. Robust algorithms estimate the edge angle automatically to better than a few hundredths of a pixel even in the presence of quantum noise. ⁸

From edge to MTF

Once the oversampled ESF is built, the processing chain is:

Determine the edge angle from the image, usually by fitting the edge location row by row.
Reproject and bin the pixel values into the oversampled ESF with equal spacing.
Differentiate the ESF to obtain the LSF (an edge is the integral of a line).
Window the LSF to suppress noise in its tails.
Fourier transform the LSF and take the magnitude.
Normalize to 1.0 at zero frequency.

The result is the presampled MTF as a function of spatial frequency in cycles per millimeter. ^{1, 3, 4}

The Nyquist frequency and why it matters

The Nyquist frequency is fixed by the sampling pitch alone:

f_{Ny} = \frac{1}{2 a} .

For a detector with a 0.143 mm pixel pitch:

f_{Ny} = \frac{1}{2 \times 0.143 mm} = 3.50 cycles/mm .

A finer 0.100 mm pitch raises Nyquist to 5.0 cycles/mm, and a coarser 0.200 mm pitch lowers it to 2.5 cycles/mm. The presampled MTF is routinely reported at and beyond Nyquist precisely because the behavior there — how fast the MTF falls, and how much response remains above Nyquist to alias into images — is diagnostic of detector design and condition. ^{3, 4}

Standardized beam quality

To make results reproducible and comparable across sites and vendors, the measurement uses standardized radiation qualities defined in IEC 61267:2005. For general radiography, the RQA5 quality (approximately 70 kV with added aluminum filtration) is a common choice. IEC 62220-1-1:2015 specifies how the MTF is combined with the normalized noise power spectrum (NNPS) and the incident air kerma to compute the DQE of radiographic detectors; mammography and dental systems are handled under separate parts of the standard. ^{1, 2}

Comparing resolution-measurement methods

Method	Test object	Exposure / practicality	Output	Key limitation
Edge method	Polished, highly attenuating straight edge, tilted ~1.5°–3°	Single exposure at clinical air kerma; tolerant of small misalignment	Presampled MTF to several × Nyquist	Sensitive to edge scatter, air gap, and K-edge effects ⁶
Slit method	Narrow precision slit (e.g., ~10 µm), tilted	High exposure; demanding fabrication and alignment	Presampled MTF (LSF directly)	Alignment and slit-width errors; low signal ⁴
Bar-pattern	Line-pair resolution gauge	Quick, visual	Limiting (aliased) resolution only	Not a true MTF; reads an aliased limit ³
Sine-pattern	Sinusoidal test chart (optical/CCD)	Direct modulation at set frequencies	MTF at discrete frequencies	Practical mainly for optical stages of a system ⁵

The edge and slit methods agree closely when both are done carefully — one landmark comparison found a mean MTF difference of only about 0.008 between them — but the edge method's single-exposure simplicity is why it dominates DR acceptance testing and QC. ³

Clinical Impact

A degraded MTF is a quiet failure. Unlike a dead pixel line or a gross artifact, a gradual loss of high-frequency response does not announce itself; it just makes small, low-contrast structures slightly harder to see on every image the detector produces. Over a fleet and a year, that is a large number of exposures. ^{7, 9}

The clinical stakes map to specific frequency bands:

Mid and high frequencies govern the rendering of edges, fine lines, catheters and guidewires, and trabecular detail. A high-frequency MTF drop blurs exactly the structures radiologists scrutinize most.
Low frequencies govern large-area contrast. An artificially elevated low-frequency MTF — often a measurement artifact from edge scatter or glare rather than true performance — can mask a real problem or inflate a DQE result. ⁶

Because detectability of a subtle structure depends jointly on the MTF and the noise power spectrum, the MTF is not interpreted alone. Model-observer studies show that a detectability index built from the MTF, the noise power spectrum, and the object contrast tracks human contrast-detail performance, and that this index follows measured DQE differences between detectors. In other words, a careful MTF measurement is a direct input to how well the system will let a radiologist see a real lesion. ⁷

Practical Optimization Tips

A defensible edge-method MTF follows the same disciplined workflow every time.

1. Prepare and align the edge device

Use a straight, polished, highly attenuating edge — for example a laminated lead or tungsten foil — free of nicks and surface contamination. Place it directly on the detector cover, remove any grid, and tilt it a few degrees (about 1.5° to 3°) to the pixel matrix. ^{1, 3}

2. Set a standardized beam quality

Use an IEC 61267 quality such as RQA5 for general radiography, with the tube and added filtration documented. Expose to a clinically relevant detector air kerma so the measurement reflects normal operation, not an artificially high- or low-dose regime. ^{1, 2}

3. Control geometry and scatter

Keep the air gap between edge and detector small, and keep the edge aligned with the central ray to minimize focal-spot and obliquity blur. Scattered radiation from a semitransparent edge, an excessive air gap, and exceeding the K-edge of the edge material all distort the measured MTF, usually by raising the low-frequency portion. Choose a beam energy below the edge material's K-edge where feasible. ⁶

4. Process consistently

Use the same edge-angle estimation, binning, ESF windowing, and extrapolation each time, ideally with validated software, so trends reflect the detector and not the analysis. Accurate automated edge-angle estimation (to better than ~0.01 pixel) keeps the synthesized ESF clean even with quantum noise. ⁸

5. Baseline and trend

Record the full MTF curve at acceptance and compare periodic measurements to it, not just a single number. Re-baseline after a detector swap, a converter or coil replacement, a software or reconstruction change, or any service that could alter the signal chain, because a stale baseline hides a real change. ¹⁰

Common pitfalls to avoid

Reading a bar pattern and calling it an MTF. A line-pair gauge gives an aliased limiting resolution, not the frequency-by-frequency response.
Ignoring processing. Many DR systems apply frequency-dependent enhancement; measure the MTF on raw ("for-processing") data, not processed images.
Letting the low frequencies drift up. An elevated low-frequency MTF is usually scatter or glare, not real performance. ⁶
Over- or under-tilting the edge. Too small an angle starves the oversampled ESF; too large an angle or a bad angle estimate smears it. ⁸
Comparing across beam qualities. An MTF at one beam quality is not directly comparable to one at another; standardize. ²

Regulatory Considerations

Spatial-resolution testing of DR detectors sits inside the facility's acceptance-testing and routine-QC obligations rather than under radioactive-material licensing. Diagnostic x-ray equipment is regulated by the U.S. Food and Drug Administration and by state radiation-control programs, not by the Nuclear Regulatory Commission (NRC), which governs byproduct (radioactive) material. Of the states DRPS serves, Florida, Maryland, Virginia, California, Nevada, Pennsylvania, New York, and New Jersey are NRC Agreement States, while Washington, DC and Delaware are regulated directly by the NRC for radioactive material; for x-ray machines, the state (or the relevant direct-NRC-area program) sets registration and inspection requirements. ²

The applicable technical frameworks for DR resolution measurement are consensus documents rather than rules:

IEC 62220-1-1:2015 — the method for determining the DQE of digital x-ray imaging devices in radiography, including the MTF and NNPS inputs. ¹
IEC 61267:2005 — the standardized radiation qualities (RQA, RQR series) used so measurements are reproducible. ²
ICRU Report 54 — foundational guidance on the assessment of medical image quality, including MTF, noise, and detectability. ⁹
AAPM Report No. 93 — acceptance testing and QC guidance historically applied to photostimulable storage-phosphor (CR) systems, with resolution testing as a core element. ¹⁰

Accreditation programs and the ACR–AAPM technical standards for diagnostic imaging expect a qualified medical physicist to perform acceptance testing and periodic performance evaluation, of which resolution assessment is part. A documented MTF baseline and trend is exactly the kind of objective evidence that supports accreditation and defends image-quality decisions. See our overview of ACR accreditation physics requirements and CT acceptance testing and commissioning for how the same acceptance-then-trend philosophy applies across modalities.

Frequently Asked Questions (FAQs)

What does the MTF of a digital radiography detector measure?

The modulation transfer function measures how much image contrast a detector preserves at each spatial frequency. It is normalized to 1.0 at zero frequency and falls toward zero as detail gets finer, so a higher MTF at a given spatial frequency means sharper rendering of small structures and edges.

Why is the edge method preferred over the slit method for DR?

The edge method uses a simple attenuating edge, needs only a single exposure at clinically relevant air kerma, and tolerates modest alignment error. The slit method requires a precisely fabricated narrow slit, exact alignment, and high exposure. Both can give equivalent presampled MTF, but the edge method is far more practical for acceptance testing and routine QC.

What is the presampled MTF and why does the edge have to be tilted?

The presampled MTF is the detector's resolution before the sampling grid is applied, including detector blur and the sampling aperture but excluding aliasing. Tilting the edge a few degrees makes successive detector rows sample the edge at slightly different positions, which synthesizes a finely oversampled edge-spread function and lets you characterize the response beyond the pixel Nyquist frequency.

What is the Nyquist frequency and how is it related to pixel pitch?

The Nyquist frequency is one over twice the sampling pitch. For a detector with 0.143 mm pixels it is about 3.5 cycles per millimeter. Spatial frequencies above Nyquist cannot be sampled correctly and alias back into the image, which is why the presampled MTF is used to describe behavior above that limit.

What beam quality should be used to measure DR MTF?

Standardized radiation qualities from IEC 61267 are used so results are reproducible and comparable. For general radiography the RQA5 quality (about 70 kV with added aluminum filtration) is common. IEC 62220-1-1 pairs the MTF with the noise power spectrum and air kerma to report detective quantum efficiency.

What are the most common errors that corrupt an edge-method MTF?

The frequent problems are scattered radiation from a semitransparent edge raising the low-frequency MTF, an excessive air gap, exceeding the K-edge of the edge material, an edge that is not polished or is contaminated, a poorly estimated edge angle, and veiling glare. Controlling geometry, edge quality, and beam energy keeps the measurement defensible.

How does MTF relate to detective quantum efficiency (DQE)?

MTF is one of the inputs to DQE. DQE combines the squared MTF, the noise power spectrum, and the incident photon fluence to express how efficiently a detector converts incident x-ray quanta into usable signal-to-noise ratio at each spatial frequency. A good MTF with poor noise behavior can still yield a low DQE.

Key Takeaways

The MTF is the detector's frequency response. It is normalized to 1.0 at zero frequency and falls with increasing spatial frequency; higher is sharper.
The edge method is the practical standard. A single image of a tilted, highly attenuating edge yields the presampled MTF via ESF → LSF → Fourier transform.
Tilt creates oversampling. A few-degree tilt interleaves row samples into a finely spaced ESF, letting the presampled MTF be characterized past Nyquist.
Nyquist is set by pixel pitch. $f_{Ny} = 1/ (2 a)$ ; a 0.143 mm pixel gives 3.5 cycles/mm.
Standardize the beam. Use IEC 61267 qualities (e.g., RQA5) so results are comparable, and measure on raw data.
Guard the low frequencies. Edge scatter, air gaps, K-edge effects, and glare inflate the low-frequency MTF and must be controlled.
Baseline and trend. The MTF is QC evidence only if it is compared to an acceptance baseline and re-baselined after service.

Conclusion

The MTF turns the vague question "is this detector sharp enough?" into a reproducible, frequency-resolved measurement. The edge method makes that measurement practical: one exposure of a tilted, attenuating edge, processed into an oversampled ESF, its derivative, and a Fourier transform, yields a presampled MTF that can be trended for the life of the detector.

The value is in the discipline. Standardized beam quality, controlled geometry, a clean edge, consistent processing, and an honest baseline are what separate a defensible resolution measurement from a number that drifts with the analyst. Paired with the noise power spectrum in the DQE framework, a careful MTF measurement is one of the most direct physical links between a detector's condition and a radiologist's ability to see the structures that matter.

How DRPS Can Help

Diagnostic Radiation Physics Services helps imaging facilities build acceptance-testing and QC programs that produce objective, defensible evidence. For digital radiography that includes presampled MTF and DQE measurement, noise and uniformity assessment, exposure-index verification, and trend documentation aligned with accreditation expectations, delivered as part of diagnostic radiography physics testing, accreditation support, and broader medical physics consulting.

DRPS supports facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, Nevada, New York, Pennsylvania, New Jersey, and Delaware.

Good image quality is not an accident of a good purchase; it is the result of measuring, baselining, and trending the physics that drives it.

Related Resources

References

International Electrotechnical Commission. IEC 62220-1-1:2015 — Medical electrical equipment. Characteristics of digital X-ray imaging devices. Part 1-1: Determination of the detective quantum efficiency — Detectors used in radiographic imaging. Geneva: IEC; 2015. webstore.iec.ch
International Electrotechnical Commission. IEC 61267:2005 — Medical diagnostic X-ray equipment. Radiation conditions for use in the determination of characteristics. Geneva: IEC; 2005. webstore.iec.ch
Samei E, Flynn MJ, Reimann DA. A method for measuring the presampled MTF of digital radiographic systems using an edge test device. Med Phys. 1998;25(1):102-113. doi:10.1118/1.598165. PubMed
Fujita H, Tsai DY, Itoh T, Doi K, Morishita J, Ueda K, Ohtsuka A. A simple method for determining the modulation transfer function in digital radiography. IEEE Trans Med Imaging. 1992;11(1):34-39. doi:10.1109/42.126908. PubMed
Boone JM, Yu T, Seibert JA. Sinusoidal modulation analysis for optical system MTF measurements. Med Phys. 1996;23(12):1955-1963. doi:10.1118/1.597840. PubMed
Maruyama S, Shimosegawa M. The effects of scattered radiation from a semitransparent edge on MTF measurement: verification of several factors by Monte Carlo simulation. Phys Eng Sci Med. 2020;43(2):547-556. doi:10.1007/s13246-020-00855-6. PubMed
Van Peteghem N, Bosmans H, Marshall NW. NPWE model observer as a validated alternative for contrast detail analysis of digital detectors in general radiography. Phys Med Biol. 2016;61(21):N575-N591. doi:10.1088/0031-9155/61/21/N575. PubMed
Honda M. A method of measuring presampled modulation transfer function using a rationalized approximation of geometrical edge slope. Nihon Hoshasen Gijutsu Gakkai Zasshi. 2014;70(4):346-358. doi:10.6009/jjrt.2014_jsrt_70.4.346. PubMed
International Commission on Radiation Units and Measurements. ICRU Report 54: Medical Imaging — The Assessment of Image Quality. Bethesda, MD: ICRU; 1996. icru.org
American Association of Physicists in Medicine, Task Group 10. AAPM Report No. 93: Acceptance Testing and Quality Control of Photostimulable Storage Phosphor Imaging Systems. College Park, MD: AAPM; 2006. aapm.org