Image analysis – Applications – Biomedical applications
Reexamination Certificate
2000-11-24
2004-01-06
Johns, Andrew W. (Department: 2621)
Image analysis
Applications
Biomedical applications
C382S260000
Reexamination Certificate
active
06674880
ABSTRACT:
TECHNICAL FIELD
This invention is in the field of modifying a visual display of an object and more particularly, to a convolution filter for a visual display.
BACKGROUND OF THE INVENTION
It is frequently desired to provide to a user an image of an object on a visual display. The object whose image is to be displayed may be a common physical object, such as a building, a landscape at a certain time of day, tree, or other living objects. In the medical field, it is frequently desired to provide images of all or different parts of an animal or the human body, such as the spinal cord, brain, bone, muscle, or other tissue. For medical diagnostic purposes, it is especially important to correctly identify abnormal tissue such as a tumor, cancerous growth, or other tissue of interest. These diagnostic purposes are often accomplished with medical images.
The images of the object may be collected by a variety of different modalities and stored using different techniques. In the medical technology field, it is common to collect images using a variety of techniques including magnetic resonance imaging (MRI). Various types of information are collected using such medical devices. The data are collected, stored, and then formed in an image to be viewed by a user, most often a medical diagnostic team which may include a radiologist and the patient's primary caregiver. As a part of the process of generating the image data and the resulting image, a convolution filter may be applied to the initial data from the imager to smooth the data and reduce the noise.
Within the medical imaging field, having an accurate and precise image of the true nature of an object has particular importance. Frequently, MRI is used to locate abnormal tissue in a human body. The abnormal tissue is, in many instances, a tumor which may be either malignant or benign. It is especially critical to determine whether the tumor is malignant or benign early in the medical diagnostic process. In the event the tumor is determined to be malignant, it becomes of particular importance to understand the extent the tumor has spread to other parts of the body, known medically as metastasizing. In the event the tumor has not yet metastasized, a certain treatment regimen is preferred and frequently is beneficial in ridding the body of the cancer. On the other hand, in the event the cancer has metastasized, a different treatment regimen and medical procedure is called for, often to more aggressively treat the cancer to increase the likelihood of removing all the cancerous tissue from the patient's body completely so that health may be restored to the patient. Such images become particularly important when the cancer is in such areas as the brain, the lungs, the lymph system, or other parts of the body which are not easily examined by or treated with surgery.
In the case of medical imaging, it is also important that false positives be minimized so as to not over-treat the patient. In the event the image appears to show the cancer in numerous places in the body, it is especially critical to ensure that healthy non-cancerous tissue has not been indicated as cancerous. A false positive causes an improper medical diagnosis and results in a different or more aggressive treatment than may otherwise be needed. Accordingly, within the medical field it is particularly important to have an image in which the abnormal growth within the human body is properly and correctly identified, including all the locations within the body, while ensuring that no healthy tissue is mistakenly identified as an abnormal growth.
SUMMARY OF THE INVENTION
According to principles of the present invention, a method is provided for modifying the value of data elements which have been stored as representation of an object whose image has been stored in memory. Using the appropriate sensor(s), data are collected from the object. A plurality of discrete data elements that represent corresponding locations in the object under investigation are stored in memory. A first value is assigned to each of these sensor-derived data elements, the first value being representative of a property of the object under examination. If multiple data sets are generated from the imaging of the object, multiple first values will exist for each corresponding location in the object. The first values are associated with the discrete data elements, which are organized in a pattern such that the data elements adjacent to each other in the pattern represent adjacent locations within the object. At this point additional modification of the first data values may occur via such filters as convolution filters or high pass or low pass filters which achieve smoothing and denoising of the resulting image. The result of this process will be second values for each discrete data element.
Next, a new value, called the similarity value, is assigned to each of the data elements, this value being representative of the similarity of one data element to a standard or reference set of data elements that represent physical properties of the object under examination. The similarity values are associated with the discrete data elements, which are organized in a pattern such that the data elements adjacent to each other in the pattern represent adjacent locations within the object. The similarity value of a particular data element is then modified based on a weighted similarity value of the data element itself and a group of adjacent data elements. This process is known as convolution filtering and yields a modified similarity value. The modified similarity value is stored for each data element as it is modified. An image is then formed which is viewable by a user that is composed of an array of discrete visual pixels whose display properties are based on the modified similarity values of the data elements.
According to one embodiment of the present invention, a convolution filter is applied to the sensor-derived data as stored in a memory. The convolution filter is preferably applied as a final step in any multi-step processing of the data. In a first sequence of steps, the sensor-derived data are collected and stored in a memory. The sensor-derived data are then organized according to the various properties of the tissues under examination in the case of medical images. A region of interest, containing at least the reference tissue or standard tissue, is designated by the user. A training set of data elements is created within the selected region of interest. One or more of the training set data elements are selected by the user to accurately reflect the properties of the tissue for which further searching is to be performed. After the selection of the training set, the remaining data elements representing the object are treated as test samples which are compared to the selected training set. The Euclidean distance between each of the test samples and each member of the selected training set is calculated and a similarity value is obtained, representative of the shortest distance between any member of the training set and the test sample. This similarity value provides a numerical measure of the relative similarity between the test sample and the training data set. The similarity value is stored in a memory location so that it may be used to create an image.
After the similarity data are created, the convolution filter of the present invention is applied. The convolution filter compares the similarity value of one data element with the similarity values of adjacent data elements and, applying one or more weighting factors, modifies the similarity value of the data element under examination. This convolution filtering is performed on an element-by-element basis for the entire array. An image is then displayed, based upon the modified similarity values, that provides a more accurate representation of the properties of the object.
According to one embodiment of the present invention, the convolution filter applied is a typical 5×5 square array weighted according to a gaussian formula in which the standard
Stork Christopher L.
Wyman Bradley T.
Confirma, Inc.
Davis , Wright, Tremaine, LLP
Donohue Michael J.
Johns Andrew W.
Nakhjavan Shervin
LandOfFree
Convolution filtering of similarity data for visual display... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Convolution filtering of similarity data for visual display..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Convolution filtering of similarity data for visual display... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3207511