ALGOR, Inc
ALGOR, Inc
 
ALGOR, IncALGOR, Inc Click Here For Printer Friendly VersionALGOR, Inc
ALGOR, IncALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc ALGOR, Inc

NEW MEDICAL IMAGING TECHNIQUE IMPROVES CHANCES OF EARLY CANCER DETECTION

By Dr. Jonathan Ophir


This ovine (sheep) kidney was imaged using both traditional sonogram and the experimental elastogram. Although the sonogram (left) sufficiently shows the general kidney's outline, the elastogram (center) reveals more of the kidney's interior structure, as compared to the anatomy diagram of the kidney (right).

June 16, 1998, Pittsburgh, Pennsylvania - Physicians use a variety of technologies to detect tumors and determine which are malignant and which are benign. These techniques include x-ray, ultrasound, biopsy and physical examination. While each of these tests can provide valuable information, my colleagues and I at the University of Texas Medical School (UTMS) at Houston, the University of Kansas Medical Center (UKMC), Baylor College of Medicine (BCM) and the Ecole Polytechnique in Montreal are developing a new method for detecting and differentiating cancer using a new type of medical imaging we call elastography.

Elastography relies on imaging the strains induced in the tissue as a result of a small external mechanical compression. To develop this imaging technique, we have employed a valuable modeling and analysis tool from the computer-aided engineering (CAE) field: the Finite Element Analysis (FEA) software of Pittsburgh-based Algor, Inc.

How Elastography Works

The elasticity of soft tissues depends to a large extent on their molecular building blocks (fat, collagen, etc.), and on the microscopic and macroscopic structural organization of these blocks. In the normal breast, for example, the glandular structure may be firmer than the surrounding connective tissue, which in turn is firmer than the subcutaneous fat. The standard medical practice of soft tissue palpation is based on qualitative assessment of the stiffness of tissue.

This new technology allows the hardness or stiffness of biological tissues to be estimated and imaged using modified conventional ultrasound instruments. It is known that certain pathologic conditions, such as malignant tumors, often manifest themselves as changes in the mechanical properties of tissue; this, in fact, is the principle behind tissue palpation. In many cases, despite the differences in stiffness, the small size of a pathological lesion and/or its location deep in the body preclude its detection and evaluation by palpation. In many cases, the lesion may not be ultrasonically detectable.


These cross-sectional images were taken from a cranio-caudal scan of a breast of a volunteer patient using a modified Diasonics Spectra™ ultrasound scanner operating with a 5 MHz transducer array. The sonogram and the corresponding elastogram were taken simultaneously from the identical anatomical site. The sonogram shows the presence of a solitary hypoechoic (echo-free, or dark) lesion. The elastogram displays the hard tissues as dark, and the soft tissues as light. It shows the same lesion as being hard (typical of most breast cancers) and larger, most likely due to desmoplasia that causes hardening only around cancerous lesions. It also shows a soft core, suggestive of a necrotic center. Additionally, a second small (~6mm) lesion is detected on the elastogram at 10 o'clock relative to the main lesion. The subcutaneous soft fat layer (thick white band on the elastogram) is highly visible and its normal flat shape is distorted by the presence of the two hard lesions. This anatomical structure is not visible on the sonogram. The elastogram’s ability to display the smaller lesion demonstrates its capability of detecting tumors in earlier stages of development.

For example, tumors of the prostate or the breast could be invisible or barely visible in standard ultrasound examinations, yet be much harder than the embedding tissue. Diffuse diseases such as cirrhosis of the liver are known to significantly increase the stiffness of the liver tissue as a whole, yet they may appear normal in a conventional ultrasound examination. We further believe that the elastic properties of benign lesions are fairly uniform throughout a benign tumor. Cancerous tumors, on the other hand, grow in a very disorganized way. Therefore, within a given malignant tumor, the elastic properties of one area of a tumor may be significantly different from those in another area.

The concept relating to the measurement of these tissue changes is an extension of the basic principles associated with traditional medical ultrasonic imaging. The principle is based upon the fact that tissues are deformed slightly when a small displacement is externally applied. Tissues that are more elastic will deform more than those tissues that are harder or less elastic. These internal deformations can then be detected and characterized with elastography. Since most cancerous tissues are much harder than normal tissues, it is expected that they will show up well on the image known as an elastogram.

To create an elastogram, two ultrasound images of the same breast tissue are taken: one of the tissue in its normal, uncompressed state, and another one of the tissue in a slightly compressed state. These images are compared point-by-point by signal processing methods. Signal processing determines how the tissue elements moved when compressed, and then converts this information into an elastogram.

National Cancer Institute Supports Elastography Research

The development of this technique is currently funded by the National Cancer Institute through a 5-year, multi-million dollar Program Project Grant (grant # P01-CA54597). The work involves basic as well as clinical research. In order to study the behavior of hard tumors that are embedded in softer tissues, we have found it useful to model certain complicated tumor geometries with Algor software.

Algor's finite element analysis capabilities enable us to predict the mechanical behavior of the tissue under compression at every location. We then use this information, along with additional models of the ultrasonic properties of the tissue, to create simulated elastograms. From these simulated elastograms we can learn much about what we should expect from real tissues, and how we should optimize the experiment and develop new elastography software algorithms.

To test the elastography technology, we need to create models based on varying conditions from a malignant tumor near the chest cavity to a cyst near glandular tissue. It is impossible to find human subjects to meet all the criteria that we want to test.

That is where Algor comes in. We use Algor software to run analyses on various hypothetical tissue arrangements to see how different types of tissues arranged in different geometries move when pressure is applied. We chose Algor because of its ability to run efficiently on a desktop personal computer and reasonable cost, as well as its proven accuracy, modeling and analysis capabilities.

The Analysis Procedure

For each hypothetical placement of tissues, we use Algor's Superdraw to create a computer model of the tissue in its normal state. Most of our FEA models are two-dimensional. Building and analyzing three-dimensional models for this application does not offer significant advantages because the image rendered by the elastogram is also two-dimensional. Using two-dimensional models also enables us to run analyses quickly and efficiently.

Algor's automatic meshing capabilities provide a finite element mesh that can be quickly generated. We found a very fine mesh to be unnecessary for this application. Because the elastogram renders all areas of a sample with the same resolution, there is no need to refine the generated mesh in areas of interest.

With a standard mechanical hydraulic testing apparatus, we determined the material properties of real breast tissues including muscle, fat, glandular tissues and various types of lesions. Data from our real-life tests of the various breast tissues is entered into Algor's data input screens. Once entered, material data is available for use by the linear stress processor.

Typically, the model of the tissue is compressed about one percent. Fixed boundary conditions and boundary elements are applied to simulate pressure. Algor's linear stress analysis software determines the stress, deflections and strains that result from the simulated application of pressure.

Comparing Algor Results to Physical Tests

Whereas a design engineer is most interested in the numerical values calculated by finite element analysis software to determine whether failure will occur, we look at the visual displays of displacement and strain results to predict what we will see in an elastogram. Because elastograms of physical models include noise, the displays of Algor strain results represent the best case image we can achieve. Image processing can then be optimized to the Algor strain display.

From the analysis displays, we can also determine if a particular tissue arrangement may be difficult to detect. If that is the case, we perform a real-life test using gelatin-based tissue models that imitate various lesions and breast tissues.


A gelatin test object contains a 3/4" diameter circular inclusion that has the same ultrasonic properties as the surrounding medium, but is 3-times harder. The sonogram (left) does not detect the presence of the inclusion, while the elastogram (center) demonstrates it well. The bright region centered on the inclusion in the elastogram is a stress-concentration artifact predicted from the Algor simulation of the sample at a 45 degree angle (right).
(The test object was created by Dr. T. Hall from the University of Kansas Medical Center.)

Using these gelatinous materials has several advantages over using human test subjects. First, human tissues are more complex. By constructing gelatin models, we can learn to recognize true tumors from other kinds of tissues. Second, inconvenience to both researcher and subject is eliminated.

We have learned much about what we should expect from real tissues by working with finite element models and gelatin test objects. Comparing visual results of the finite element analyses with elastograms of test objects has enabled us to optimize the procedure and develop new elastography software algorithms.

Clinical Testing Begins

An experimental clinical elastography system is currently being tested by Dr. Brian Garra at the Department of Radiology at Georgetown University Medical Center. The protocol being tested relates to improving the specificity of ultrasound follow-up examinations of mammographically detected breast lesions. Currently, the most accepted and the most sensitive means for detecting breast lesions is with x-ray mammography. While the method is very sensitive for detecting lesions, only about 20% of those lesions identified by mammography are found to be cancers when they are biopsied.

Reducing the number of unnecessary biopsies is an important goal in breast cancer management. The average biopsy costs between $2,000 and $3,000 and causes considerable stress to patients. Given both the cost and trauma associated with performing biopsies in all cases where patients had mammographically detected lesions, there is a strong incentive to develop additional non-invasive methods such as elastography to accurately determine if a lesion is benign or malignant.

To date, the initial results of this clinical work are promising. Dr. Garra has identified several indicators using this technique, which suggest means for distinguishing between benign and cancerous lesions. However, while the results are encouraging, the clinical work is still in an early stage.

Taking Elastography in New Directions

In the future, we plan to experiment with using the elastography to detect and evaluate other kinds of cancer, particularly prostate cancer. Currently, two diagnostic methods are used to detect prostate cancer: digital rectal examination and traditional sonography. Even with these two detection options, a large number of prostate cancer cases go unrecognized.

Successful cancer treatment depends on early detection and evaluation. As we venture forward in our research, we plan to continue to use Algor software to analyze the mechanical behavior of the tissues. This information will enable us to develop and refine elastography as a tool that physicians can use to detect and diagnose cancer as early as possible.

Dr. Jonathan Ophir is a Professor of Radiology at University of Texas Medical School (UTMS) at Houston and is the Program Director of National Cancer Institute Program Project # P01-CA64597. Dr. Thomas Krouskop, a Professor at Baylor College of Medicine in Houston, Dr. Faouzi Kallel, a post-doctoral fellow at UTMS, and Dr. Michael Insana, Professor of Radiology at the University of Kansas Medical Center also contributed to this article.


Researchers at UTMC Houston modeled a cross-section of semi-tendinous bovine muscle structure, idealizing the muscle bundles as circular cylinders. The Algor strain simulation (left) suggests the kind of image that will be produced by an elastogram of an actual muscle sample (right).


Copyright 2011 Autodesk, Inc. All rights reserved.   Privacy Policy Legal Notices & Trademarks Report Piracy