Visualization in Medicine

Introduction

The combination of computerized imaging and medicine has resulted in the use of the techniques of Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET), Single Photon Emission Computed Tomography (SPECT) and Ultrasonography, all of which produce volume imaging data. These data may be displayed as acquired, generally as a series of slices oriented in the acquisition plane. This type of display has proven quite useful for diagnostic purposes; it is simple and requires unsophisticated computer resources. Displaying the slices as they were acquired has one serious limitation, the three dimensional relationships between adjacent slices cannot be visualized directly. This limitation is most evident when the information is being used for therapy planning.

Multiplanar reconstruction (MPR) is a technique for computing arbitrarily oriented planar slices from volume imaging data. By examining arbitrarily oriented slices it is possible to elucidate information about 3D relationships that is not evident when examining slices oriented in the acquisition plane. MPR is routinely used with CT; programs to display MPR's exist for most current CT scanners. It is less often used with other modalities. Three dimensional imaging techniques produce 3D images of the structures being studied and requires more sophisticated computer hardware and software.

There are numerous applications for visualization in medicine. The first major use was in planning orthopedic surgery using CT scans. The orthopod who is managing a patient with a bone injury needs to answer a number of different questions. Should the patient receive surgery? If so, what type of approach is to be used? What type of hardware (plates, screws, etc.) need to be used? All of these questions are addressed most easily in a three-dimensional format. Bone imaging from CT is probably one of the easiest applications from a technical point of view. Bones are high density objects and this results in high contrast between bones and other tissue on a CT scan. The obvious need for 3D visualization in orthopedics, and the relative ease of picking out bones from CT scans resulted in the first attempts at 3D visualization. Although orthopedics continues to be an important application for visualization, there are many other fields that could and do benefit.

Oncologic (cancer) imaging is a prime candidate. Treatment of oncologic patients generally falls into three categories:chemotherapy, radiation therapy, or surgical resection. All three of these treatments can benefit from visualization. A key question in chemotherapy is whether the tumor is growing or shrinking. This can be best seen using visualization. Radiation therapy is done by configuring X-ray beams to target the tumor and spare other tissue. This problem is best addressed in three dimensions. Surgical resection relies upon removing tumor and keeping surrounding tissue intact and functional. Other applications include: multimodality imaging, neurological imaging for brain surgery, tissue characterization, medical school teaching, plastic surgery, etc.

There are several important research areas in this field. Segmentation (classification) of data is a necessary step in almost all visualization techniques. In order to image a specific structure, it is generally necessary to identify it some way in the computer. This problem has proved to be very difficult to do in a general way. Rendering speed is another area that continues to be an obstacle for future research. Many applications require near real-time response from an imaging system to allow interactive manipulation. Although imaging methods exist that allow for this, they generally produce poorer quality images. The high quality rendering algorithms produce very good images, but take a long time.

The registration (spatial alignment) of multiple datasets for multimodality imaging is difficult. A patient may receive multiple studies using different acquisition techniques or the same technique at different times. Optimally, these studies should be examined together. This requires that the studies be aligned spatially. A number of good methods for registering rigid structures (such as the head) are available. But, the registration of non-rigid structures (the liver, kidneys, etc.) is an unsolved problem.


Visualization in Medicine
HyperVis Table of Contents

Last modified on March 02, 1999, G. Scott Owen, owen@siggraph.org