Three-D postprocessing shines in abdomen, pelvis

Moviegoers may relegate those red- and green-lensed cardboard glasses worn while watching 3D movies to a bygone era. But the release of a spate of animated movies in true digital 3D format over the past several years proclaims the comeback of this exciting genre.

In an unlikely comparison, 3D diagnostic imaging, buoyed by recent technological advances, is also beginning to hit its stride and could heralda new imaging paradigm in multi­slice CT.

The technology leap made with 64-slice CT scanners has resulted in excellent spatial and temporal resolution. Gargantuan numbers of images are generated per data set, however, reaching up to 3000 images in certain cases. Axial interpretation, which has been the norm since the introduction of CT scanning, can be cumbersome, tiring, and time-consuming with such large data sets. Yet quick report turnaround times are considered vital for patient management.

The ability to achieve near-isotropic imaging using 64-slice CT scanners has been a driving force for 3D imaging to become more of a necessity than a luxury. Scanning speeds three to four times faster than previous generation scanners have obviated motion and respiratory artifacts, and enhanced z-axis coverage has resulted in 3D reconstructions of superior quality. Significant improvements in postprocessing software have led to the development of faster and more interactive workstations that can handle a large volume of images.

TEAM EFFORT

Referring physicians can request additional 3D analysis with multislice CT scans if they are already aware of the benefits of the added information portrayed in 3D images. Alternatively, a radiologist can offer 3D reconstruction as a problem-solving option after viewing axial images. These options have led to the creation of 3D labs with stand-alone workstations that provide the processing power to handle gargantuandata sets.

A special breed of radiology technologists referred to as 3D postprocessing technologists or 3D imaging specialists has emer­ged. The work profile of a 3D postprocessing technologist is a unique mix of art and science that encompasses good computer skills and a background in CT, MR, vascular, and basic diagnostic imaging.

To ensure a smooth and efficient workflow, the radiologist must oversee all 3D reconstructions generated by the technologist in the 3D lab. He or she must be prepared to answer queries and mentor the 3D imaging technologists in addition to providing referring physicians with valuable input regarding complex cases.

The table lists applications in the abdomen and pelvis. Optimization of MSCT protocols is the key to success. Scanning parameters with reference tointerscan spacing, timing of contrast injection, pitch, and slice thickness are integral to creating superior quality 3D reconstructions and image postprocessing.

POSTPROCESSING TECHNIQUES

Various techniques offer advantages but face limitations, which may be minimized by refinements.

• Multiplanar reconstruction. Source axial images are reconstructed to generate images along arbitrary imaging planes. Data set quality is similar to that of source axial images. Before the advent of 16- and 64-slice CT, MPR suffered from stair-step artifacts due to nonvolumetric scans and motion between slice acquisitions. The near-isotropic images generated by 64-slice CT scanners, with negligible motion or respiratory artifacts, are ideal for generating MPR images (Figure 1). This is a simple and widely used technique. It can be useful as a quick problem-solving tool and adds very little to the reading time, sometimes actually decreasing reading time.
• Curved planar reformation. This technique is used to visualize tubular structures such as ducts and blood vessels. Longitudinal cross-sections of tubular structures are obtained to show their lumen, wall, and surrounding tissue in a curved plane (Figure 2). CPR does have limitations, however. Inaccurate centerline detection can give an impression of pseudostenosis, and longitudinal sections may not cover the vessel in its entirety.
• Maximum intensity projection. The highest attenuation voxels along lines projected through the volume data are selected. Parallel rays are cast from the projection plane through stacks of reconstructed sections that make up data volume.

The maximum intensity encountered along each ray is placed into the projection plane to construct the MIP. The important aspect to remember is that serious diagnostic errors can occur if MIP images are used alone. Radiologists should use MIPs in correlation with volume-rendered images.The advantages of MIPs include that they are computationally fast, have less variability in image reconstruction than volume rendering, and are valuable for 3D rendering of vasculature. The technique is limited, however, because it cannot accurately depict actual 3D relationships, it overestimates stenosis, and it loses depth information due to a lack of shading. In addition, large calcifications prevent accurate visualization of vascular lumen in atherosclerotic vessels, and small vessels with lesser decreased intensity than the background can become invisible due to volume averaging.
Radiologists can minimize these limitations by applying multiple viewing angles with an incorporated cine loop, using curved-slab MIP, and employing depth-shaded MIP or closest vessel projection MIP.

• Minimum intensity projection. Each voxel along a line from the viewer's eye through the volume of data is evaluated. The minimum voxel value is selected on the basis of minimum intensity to obtain the 2D image. The main drawbacks of minimum intensity projection are lack of depth information, superimposition of complex an­ atomic structures, and loss of density information.

RENDERING TECHNIQUES

A computer algorithm is used to transform serially acquired axial CT image data into 3D images. Types of rendering are thresholding/surface-based (binary) and percentage/semitransparent volume-based (continuous) techniques. The type of rendering determines the quality of 3D reconstructed images.

• Surface rendering. Object surfaces are portrayed in the rendition directly from the given scene. Usually, thresholding is used to specify the structure of interest in the given scene. Surface rendering forms the bulk of 3D imaging in the emergency room. It is excellent for fractures and fracture-dislocations, vascular trauma, and vascular emergencies.
• Shaded surface display. In this thresholding-based, binary classification technique, each tissue is designated two absolute numbers: a low and a high threshold. Each tissue is assigned a color, and once the volume is classified, thresholding-based algorithms extract surfaces from the classified data. Volume rendering has essentially replaced SSD.
  Shaded surface display is advantageous because it is less computationally intensive than other methods and, therefore, requires less time to generate images. It also has the clearest volume depth cues of all 3D images. The technique is limited because it is prone to volume averaging and artifacts.
• Volume rendering. This percentage/semitransparent volume-based technique displays data without classifying it into rigid all-or-nothing categories. It assumes that a voxel may represent one or more tissue types. The amount of each tissue is between 0% and 100% in each voxel. Each tissue type is given a color and transparency. The volume-rendered image closely approximates the actual contents of voxels (Figure 3).

Volume rendering offers an excellent anatomic overview and has high-fidelity images with significantly fewer artifacts than other methods. It is, however, time consuming and does not grade stenosis accurately.

THE TIME IS NOW

High-quality 3D images provide quantitation of various lesions, including stenosis, diameter, and circumference. Surgeons could use 3D images as a virtual road map to access lesions or tumors in complex anatomic relationships. This capability will translate into better patient care and improved visualization of vascular stenoses, tumor vessel relationships, and traumatic injuries.

Such visual information allows better preprocedural planning, intraprocedural guidance, and postintervention assessment. A busy 3D lab can be a significant source of revenue, where the primary focus is to generate useful and not just pretty 3D images that provide relevant clinical information.

Radiologists face constant pressure to interpret more images in less time. Interactive 3D real-time display allows superior and quicker visualization of anatomy and pathology that may have been difficult to interpret in 2D with axial images. A modification of the established axial-based workflow pattern could lead to a new paradigm wherein 3D imaging techniques enable one to work smarter instead of harder.

Today, a radiologist could probably comprehend our bodies' complex choreography using an interactive 3D realm quicker than one who doesn't. In the words of science fiction writer J.G. Ballard, "Science and technology multiply around us. To an increasing extent, they dictate the languages in which we speak and think. Either we use those languages, or we remain mute."

CLAMORING COLLEAGUES

Radiologists can't afford to remain mute. Our clinical colleagues clamor for 3D images. Our patients want to see them as well. And data increasingly show that 3D images, either alone or in some combination with axial source images, provide a higher confidence in diagnosis.Let's all put on our red- and green-colored glasses and step into the 21st century. This way we'll not only be prepared to serve our patients better, but we'll also be able to watch the much-awaited 3D Star Wars that will be coming soon to a multiplex near you.

Dr. Sebastian is a radiologist in the abdominal imaging section, and Dr. Small is director of abdominal imaging, both at Emory Uni­versity School of Medicine in Atlanta, Georgia.