PET/CT and SPECT/CT Image Registration

In PET/CT and SPECT/CT, the CT and the emission study must be spatially aligned before they are combined — both for the fused image a physician reads and for the attenuation correction applied to the functional data. Because the CT is captured in seconds while the emission scan averages over minutes, the two datasets routinely disagree by a few millimeters, and at boundaries where attenuation changes sharply — the heart against the lung, the liver against the lung — that disagreement can create an artifactual defect that mimics disease.¹⁵

Image registration is therefore not a cosmetic overlay step. It is a quantitative requirement of attenuation correction and a recurring source of error that technologists, physicians, and physicists must recognize, correct, and verify.¹⁸

Introduction

Hybrid imaging pairs a functional modality — positron emission tomography (PET) or single-photon emission computed tomography (SPECT) — with computed tomography (CT) on a single system. The functional study shows where a radiopharmaceutical concentrates; the CT shows the anatomy and, critically, supplies the map of tissue attenuation used to recover quantitatively accurate emission images.⁵

For both of these jobs the two images must occupy the same coordinate space. Image registration is the process of finding the geometric transformation that brings one dataset into alignment with the other; fusion is the subsequent display of the aligned datasets together. On a hybrid scanner the detectors are mechanically co-registered, so a baseline alignment is built in, but patient motion between and during the two acquisitions means residual misregistration is the norm rather than the exception.¹⁷

This article explains why registration is required, the physics and mathematics of how it is performed, where misregistration comes from and how much it matters, the clinical consequences, practical steps to minimize it, the quality-control and regulatory context, and the questions clinical teams raise most often.

Topic Explanation

What registration means in hybrid imaging

Registration is finding the transformation that maps points in one image to the corresponding points in another. In hybrid imaging the "moving" image (often the CT, used as the attenuation map) is transformed to match the "fixed" emission image, or vice versa, so that a given anatomic location has the same coordinates in both.⁶

Key terms used throughout:

• Co-registration — alignment of two datasets acquired on the same system; on a hybrid scanner the hardware provides an initial co-registration.
• Fusion — the combined display of registered functional and anatomic images, typically with the functional data shown in color over a grayscale CT.
• Attenuation map (μ-map) — the spatial map of linear attenuation coefficients, derived by scaling the CT Hounsfield units to the photon energy of the emission study (511 keV for PET; the gamma energy of the radionuclide for SPECT).
• Misregistration — residual spatial disagreement between the emission and CT data after co-registration.

Why the CT does double duty

In modern PET/CT and SPECT/CT the CT replaced the older radionuclide transmission source for attenuation correction. The CT is fast, low-noise, and provides anatomic detail, but it is acquired at a diagnostic x-ray energy, so its Hounsfield units must be converted to attenuation coefficients at the emission energy before they can correct the functional data.⁵ This conversion is what makes the CT an attenuation map, and it is why any spatial mismatch between the CT and the emission data propagates directly into the corrected emission image. The broader mechanics of this conversion are covered in our guide to PET/CT attenuation correction and its SPECT counterpart, SPECT/CT attenuation correction.

The temporal mismatch at the root of the problem

A helical CT covers the chest in a few seconds, effectively freezing one phase of the respiratory cycle, while a PET or SPECT bed position integrates counts over minutes, averaging over many breaths. The two images therefore represent different average anatomy even when the patient is perfectly still otherwise. This temporal mismatch is the fundamental reason misregistration is intrinsic to hybrid imaging rather than merely a technical failure.²³

Key Technical Principles

Transformations: rigid, affine, and deformable

A registration is defined by the family of transformations it is allowed to use. The simplest is the rigid-body transformation, which permits only rotation and translation:

where $\mathbf{x}$ is a point in the original image, $R$ is a $3\times 3$ rotation matrix, and $\mathbf{t}$ is a translation vector. A rigid transformation has six degrees of freedom — three rotations and three translations — and assumes the anatomy does not deform. Adding uniform scaling and shear gives an affine transformation with twelve degrees of freedom. Neither can model the local, non-uniform warping produced by breathing or cardiac motion; that requires a deformable (elastic) transformation with many degrees of freedom, represented as a dense displacement field.⁶

Similarity metrics and the registration search

To decide when two images are aligned, the algorithm optimizes a similarity metric over the transformation parameters. Because PET or SPECT and CT are different modalities with no simple intensity correspondence, the standard choice is mutual information, an information-theoretic measure of how well the intensity of one image predicts the other:

where $H(A)$ and $H(B)$ are the marginal entropies of the two images and $H(A,B)$ is their joint entropy. Mutual information is maximized when the two images are aligned, because corresponding anatomic edges then coincide and the joint intensity distribution becomes most predictable. An optimizer adjusts the transformation parameters to maximize this metric.⁶ In cardiac PET/CT specifically, intensity-based automatic registration can struggle because the emission and CT contrast differ so strongly, and emission-driven or manual realignment of the μ-map has been shown to be reliable where fully automatic methods failed.³

How misregistration becomes an uptake error

Attenuation correction multiplies each emission measurement by a factor that undoes tissue attenuation. For a photon path of length $L$ through tissue of linear attenuation coefficient $\mu$, the correction factor is:

At the PET energy of 511 keV the narrow-beam linear attenuation coefficient of water is about $0.096{\text{cm}}^{-1}$, so a 20-cm soft-tissue path carries a correction factor of $e^{0.096\times 20}\approx 6.8$.⁵⁹ If misregistration shifts the CT so that a region truly overlying soft tissue is instead assigned lung density (roughly a quarter of water's coefficient), the applied correction factor collapses toward $e^{0.025\times 20}\approx 1.6$. The emission activity in that region is then under-corrected, and it appears artifactually cold. This is exactly the mechanism by which a breathing mismatch at the heart–lung border produces a false perfusion defect.¹⁴

The clinical magnitude has been measured. In myocardial PET/CT, published work found that misregistration between the emission and CT data on the order of 10 mm or more produced statistically significant artifactual reductions in apparent tracer uptake, with anterior-wall-to-septum uptake ratios falling from about 1.00 to about 0.80 as the shift increased.⁴ A larger series reported that misregistration caused artifactual defects in roughly 40% of cardiac PET/CT patients when standard helical CT attenuation correction was used, and that these defects normalized once the CT and emission data were correctly co-registered, for example by using an averaged CT acquired over the breathing cycle.¹³ Averaging the CT over respiration — a low-dose cine or "average CT" acquisition — is one of the most effective fixes because it matches the time-averaged anatomy that the emission scan actually sees.²

Sources of misregistration and their mitigations

Clinical Impact

The dominant clinical risk of misregistration is a false-positive or false-negative functional finding driven by an attenuation-correction error rather than true physiology. Cardiac imaging is the canonical example: an anterior or lateral "defect" created by a breathing mismatch can be misread as ischemia or infarction, prompting unnecessary downstream testing. Recognizing this pattern — and confirming it against the non-attenuation-corrected images, where a misregistration artifact typically disappears — is a core interpretive safeguard.¹³⁴

Misregistration also degrades quantitative accuracy in oncologic PET/CT. Because the standardized uptake value depends on attenuation-corrected activity, a lesion straddling the lung–liver or lung–chest-wall boundary can have its SUV biased by respiratory mismatch, which matters when SUV trends are used to assess treatment response. The same temporal-mismatch logic that creates cardiac defects produces lesion-quantification errors near moving boundaries, and respiratory motion management is used to control both. Our note on PET/CT respiratory gating and motion management addresses these techniques in depth.

Beyond attenuation correction, the fused display itself drives clinical decisions — localizing a focus of uptake to a specific lymph node, bone, or lesion depends on accurate overlay. A fusion that is off by a centimeter can assign activity to the wrong structure, with direct consequences for staging and for surgical or radiotherapy targeting.⁸

Practical Optimization Tips

Acquire the data to minimize mismatch

• Match breathing between the CT and the emission scan. Shallow free-breathing or an averaged/cine CT over the respiratory cycle generally aligns better with the time-averaged emission data than a breath-hold CT acquired at full inspiration.¹²
• Keep the interscan interval short and the patient still. Minimizing the delay between CT and emission, and immobilizing the patient, reduces bulk-motion mismatch.
• Use motion management where it matters. Respiratory gating and, for cardiac studies, ECG gating reduce the averaging mismatch at moving boundaries.

Review and correct before signing

• Always inspect the overlay, especially over the heart and the diaphragm. Confirm that the myocardium sits on soft tissue in the μ-map, not on lung.
• Compare attenuation-corrected and non-attenuation-corrected images. A defect that is present on the corrected images but absent on the non-corrected images is a hallmark of a misregistration artifact.¹³
• Realign and reconstruct when needed. Manual or emission-driven realignment of the CT, or reconstruction with an averaged CT, resolves most cardiac misregistration artifacts.¹²³ Related QC habits are summarized in our cardiac SPECT MPI quality control guide.

Verify the system, not just the study

• Confirm hybrid alignment at acceptance and periodically. A phantom with markers visible to both subsystems verifies that the emission and CT fields of view coincide within tolerance, independent of any particular patient. This is part of routine hybrid-system QC, as described in our overview of PET/CT daily QC and calibration.

Regulatory Considerations

Hybrid imaging sits at the intersection of radioactive-material regulation for the radiopharmaceutical and machine regulation for the CT, with registration accuracy addressed through accreditation and professional quality standards rather than a single numeric regulatory limit.

• Radioactive material (the PET/SPECT side). Possession and medical use of the radiopharmaceutical fall under 10 CFR Part 35 (or the equivalent Agreement State program), with dose limits under 10 CFR Part 20. These govern the functional study, the authorized user, and the medical physicist's role.
• The CT subsystem. The CT is regulated as a radiation-producing machine under state radiation-control programs, with image-quality and dose performance confirmed by a qualified medical physicist.
• Accreditation and professional standards. The ACR nuclear medicine and PET accreditation programs and SNMMI procedure standards establish image-quality and QC expectations for hybrid systems, including the spatial alignment of the emission and CT data. International guidance from the IAEA — including the Human Health Series atlas of PET/CT quality control and image artefacts and the companion quality-assurance guidance that defines a PET/CT image-registration accuracy test — catalogs misregistration among the artifacts a program must recognize and control.⁸⁹
• State jurisdiction. DRPS serves Florida, Maryland, Virginia, Washington DC, California, Nevada, Pennsylvania, New York, New Jersey, and Delaware. Radioactive-material medical use is administered by the NRC or the Agreement State, while the CT machine is regulated by the state radiation-control program; Washington DC and Delaware are direct-NRC jurisdictions for radioactive material. Always confirm requirements with the authority having jurisdiction.

Frequently Asked Questions (FAQs)

Why do the PET or SPECT images and the CT need to be registered?

The CT serves two purposes in a hybrid study: it provides the anatomic image that is fused with the functional data for interpretation, and it provides the attenuation map used to correct the emission data. Both uses require the CT and the emission data to represent the same anatomy in the same position. Because the two scans are acquired at different times and on different timescales, they must be aligned — registered — before they are combined.⁵

What causes misregistration in PET/CT and SPECT/CT?

The most common cause is motion between or during the scans: respiratory motion, cardiac motion, and bulk patient movement. A CT is acquired in seconds while an emission scan averages over minutes, so the two capture the anatomy in different phases. Bowel peristalsis, a full bladder, changes in position between scans, and truncation or metal in the CT can also contribute.¹²

How much misregistration causes a problem?

It depends on where it occurs. At the heart–lung or liver–lung boundary, where attenuation changes sharply, even small shifts matter. Published cardiac PET/CT studies found that misregistration on the order of 10 mm or more produced statistically significant artifactual reductions in apparent tracer uptake in the anterior and lateral walls, and that misregistration artifacts affected a large fraction of cardiac cases when uncorrected.¹⁴

What is the difference between rigid and deformable registration?

Rigid registration allows only translation and rotation — six degrees of freedom — and treats the body as a single non-deforming object. Affine registration adds scaling and shearing. Deformable (elastic) registration allows local, non-uniform warping to accommodate motion such as breathing. Hybrid scanners perform hardware co-registration by acquiring both modalities on one gantry, and software registration refines or corrects the alignment afterward.⁶

Does hardware co-registration on a hybrid scanner eliminate the problem?

No. Acquiring PET or SPECT and CT on the same gantry with the patient on one table removes gross positioning differences and fixes the spatial relationship between the detectors, but it does not stop the patient from breathing or moving between the two acquisitions. Respiratory and cardiac motion can still misalign the emission and CT data, so software correction and good acquisition technique remain necessary.¹⁷

How is registration accuracy checked during quality control?

Hybrid-system acceptance testing includes a check of the spatial alignment between the emission and CT subsystems using a phantom with sources visible in both modalities, confirming that the fused images coincide within tolerance. Clinically, technologists and physicians review the overlay of emission and CT data — particularly over the heart — and use the non-attenuation-corrected images to confirm that a suspected defect is not a misregistration artifact.⁸

Key Takeaways

• The CT in a hybrid study is both the fusion anatomy and the attenuation map, so emission–CT alignment is a quantitative requirement, not just a display preference.⁵
• Misregistration is intrinsic because a CT freezes one phase while the emission scan averages over minutes.²
• Attenuation correction multiplies by $e^{\mu L}$, so assigning the wrong tissue to a path — soft tissue versus lung — produces large, artifactual uptake changes.⁹
• In cardiac PET/CT, shifts of about 10 mm or more cause significant artifactual uptake reductions, and uncorrected misregistration affected roughly 40% of cases in a large series.¹⁴
• Rigid registration has six degrees of freedom; deformable registration is needed for respiratory and cardiac motion, with mutual information the standard cross-modality similarity metric.⁶
• Averaged/cine CT, motion management, overlay review, and comparison with non-attenuation-corrected images are the practical defenses.¹²³

Conclusion

Image registration is the quiet workhorse of hybrid imaging. When it succeeds, the reader sees a seamless fusion and trusts the attenuation-corrected numbers; when it fails, a few millimeters of breathing mismatch can manufacture a defect that looks exactly like disease. Because the mismatch is built into the physics of acquiring a fast CT alongside a slow emission scan, it cannot be engineered away entirely — it must be managed through acquisition technique, software correction, disciplined overlay review, and periodic verification that the system's two halves still agree. A nuclear medicine program that treats registration as a core quality-control concern, not an afterthought, is one that keeps its functional findings trustworthy.¹⁵⁸

How DRPS Can Help

Diagnostic Radiation Physics Services (DRPS) supports PET/CT and SPECT/CT programs across Florida, Maryland, Virginia, Washington DC, California, Nevada, Pennsylvania, New York, New Jersey, and Delaware with PET/CT and nuclear medicine physics, acceptance testing of hybrid systems, attenuation-correction and registration-accuracy verification, accreditation support, and routine quality-control program design — all performed by board-certified medical physicists. We help facilities confirm that the emission and CT subsystems align within tolerance, establish the overlay-review and averaged-CT workflows that prevent cardiac misregistration artifacts, and document the QC that accreditation and state programs expect.

Related Resources

• PET/CT attenuation correction
• SPECT/CT attenuation correction
• PET/CT respiratory gating and motion management
• Cardiac SPECT MPI quality control
• PET/CT daily QC and calibration
• PET/CT and nuclear medicine physics services

References

1. Gould KL, Pan T, Loghin C, Johnson NP, Guha A, Sdringola S. Frequent diagnostic errors in cardiac PET/CT due to misregistration of CT attenuation and emission PET images: a definitive analysis of causes, consequences, and corrections. Journal of Nuclear Medicine. 2007;48(7):1112-1121. doi:10.2967/jnumed.107.039792. doi.org
2. Pan T, Mawlawi O, Luo D, et al. Attenuation correction of PET cardiac data with low-dose average CT in PET/CT. Medical Physics. 2006;33(10):3931-3938. doi:10.1118/1.2349843. doi.org
3. Martinez-Möller A, Souvatzoglou M, Navab N, Schwaiger M, Nekolla SG. Artifacts from misaligned CT in cardiac perfusion PET/CT studies: frequency, effects, and potential solutions. Journal of Nuclear Medicine. 2007;48(2):188-193. pubmed.ncbi.nlm.nih.gov
4. Tomita Y, Ishida M, Ichikawa Y, et al. The effect of misregistration between CT-attenuation and PET-emission images in ¹³N-ammonia myocardial PET/CT. Journal of Nuclear Medicine Technology. 2016;44(2):73-77. doi:10.2967/jnmt.116.172742. doi.org
5. Kinahan PE, Hasegawa BH, Beyer T. X-ray-based attenuation correction for positron emission tomography/computed tomography scanners. Seminars in Nuclear Medicine. 2003;33(3):166-179. doi:10.1053/snuc.2003.127307. doi.org
6. Sotiras A, Davatzikos C, Paragios N. Deformable medical image registration: a survey. IEEE Transactions on Medical Imaging. 2013;32(7):1153-1190. doi:10.1109/TMI.2013.2265603. doi.org
7. Schaefferkoetter J, Shah V, Hayden C, Prior JO, Zuehlsdorff S. Deep learning for improving PET/CT attenuation correction by elastic registration of anatomical data. European Journal of Nuclear Medicine and Molecular Imaging. 2023;50(8):2292-2304. doi:10.1007/s00259-023-06181-9. doi.org
8. International Atomic Energy Agency. PET/CT Atlas on Quality Control and Image Artefacts. IAEA Human Health Series No. 27. Vienna: IAEA; 2014. iaea.org
9. International Atomic Energy Agency. Quality Assurance for PET and PET/CT Systems. IAEA Human Health Series No. 1. Vienna: IAEA; 2009. iaea.org
10. U.S. Nuclear Regulatory Commission. 10 CFR Part 35, Medical Use of Byproduct Material. nrc.gov