Multiplanar Reformation (MPR) & Curved Planar Reformation (CPR)

Reformation is a new view of existing image data

Multiplanar reformation (MPR) resamples a reconstructed volume to display a different plane. It differs from primary reconstruction from projection data. A coronal or sagittal MPR generally uses the stored thin axial images; a new primary reconstruction with a different acquisition-dependent kernel generally needs appropriate raw data. Knowing which operation is requested prevents an unnecessary repeat scan or a promise that the archive cannot fulfill.

The usual anatomical planes are transverse, coronal and sagittal. Transverse separates superior from inferior, coronal anterior from posterior, and sagittal left from right. Oblique planes can align with a joint, vessel or other structure. Verify the patient's position and DICOM orientation rather than assuming the screen's left edge always represents the same anatomical side in every derived view.

Sampling and slice intervals

A cubic voxel has equal nominal dimensions in all three directions. For a 320 mm reconstruction field and a 512 matrix, the in-plane pixel width is 320/512 = 0.625 mm. A 0.625 mm reconstructed thickness gives cubic sampling, provided the acquisition and reconstruction support it. A 5 mm thickness instead gives 0.625 × 0.625 × 5 mm elongated samples.

Fine longitudinal data support smoother oblique surfaces than thick, widely spaced data. However, cubic dimensions alone do not guarantee identical physical resolution in every direction. Focal spot, detector aperture, reconstruction and slice sensitivity contribute. Interpolation between thick slices cannot recover detail that was never adequately sampled.

For 0.625 mm thickness and 0.4 mm center-to-center interval, nominal overlap is (0.625 − 0.4)/0.625 × 100 = 36%. It is not 50%. Overlap can improve reformation continuity without adding exposure when made from existing data, but adjacent images contain correlated information. A smaller interval is not the same as a thinner independently resolved slice.

Thin and thick displays

A thin MPR displays a selected plane with interpolation appropriate to the software. A thick averaged slab combines information over a chosen depth and can reduce noise while increasing partial volume. This differs from maximum-intensity projection, which selects the highest value along each ray, or minimum-intensity projection, which selects the lowest. Label the operation and thickness so that the reader knows what information has been combined or selected.

For spine imaging, align a plane with the vertebral bodies or disc space as ordered. For an extremity, align with the joint rather than accepting a generic body-axis plane that cuts the articular surface obliquely. Review source slices when an apparent discontinuity could reflect motion, spacing or reformation rather than a true fracture.

Curved planar reformation follows a structure

Curved planar reformation (CPR) follows a selected curved path, commonly a vessel or duct, to show its course in one display. A centerline may be drawn manually or estimated automatically. It must be inspected because the algorithm can follow a vein, branch, calcification or other high-attenuation structure instead of the intended lumen.

A longitudinal CPR is useful for tracing the vessel, but stenosis measurements should use appropriate cross-sections perpendicular to the local course. An oblique cut can elongate the lumen and distort a diameter measurement. Irregular lumens are not necessarily circular or elliptical, so an ellipse formula is only an approximation when that shape is assumed.

Vascular and nonvascular applications

In coronary CTA, CPR follows a coronary artery around the heart and relates narrowing to branch anatomy. Motion, calcium blooming and a poorly chosen centerline can create pseudostenosis. In carotid or peripheral runoff imaging, follow the vessel continuously and compare cross-sectional measurements with source images. A visually narrow point on one stretched display does not prove the percentage of stenosis.

For pancreaticobiliary or urinary questions, a curved view can trace a duct or ureter, but overlapping vessels, peristalsis and incomplete enhancement may limit automatic tracking. Dental or maxillofacial views can follow the jaw. Each application benefits from a recognizable path and correctly labeled derived series, not from a generic curved image with no anatomical verification.

Measurements and display limitations

Use the specified reference method for a measurement. Carotid percentage stenosis, minimum luminal area and vessel diameter are different metrics and should not be substituted for one another. CT plaque attenuation also depends on spectrum, enhancement, reconstruction and partial volume; one rigid HU range does not guarantee a histologic plaque composition or stability.

An apparent branch separation may be caused by motion between source images. A calcified plaque can obscure an enhancing channel through blooming. A curved display may foreshorten or stretch distances depending on its layout. Use orthogonal views and source data to test whether the finding survives a different display.

A practical quality review

Suppose a carotid CPR appears to show a severe short stenosis. Inspect the centerline and perpendicular section. If the line has entered an adjacent calcified wall or another vessel, correct the path under the approved workflow and reassess. If the source data still show substantial narrowing, preserve the evidence for interpretation rather than treating a prettier CPR as a reason to dismiss it.

Before exporting, verify series identity, side, orientation, thickness and labels. Preserve the source data and avoid overwriting clinically relevant series. Postprocessing should reveal the acquired information in a useful form, with limitations communicated, rather than create unsupported certainty.

Reference: AAPM reconstruction and display terminology.