Surgery – Diagnostic testing – Detecting nuclear – electromagnetic – or ultrasonic radiation
Reexamination Certificate
2002-05-20
2003-08-19
Lateef, Marvin M (Department: 3737)
Surgery
Diagnostic testing
Detecting nuclear, electromagnetic, or ultrasonic radiation
C382S131000, C378S004000
Reexamination Certificate
active
06609021
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to computer-aided diagnosis (CAD) and, in particular, to a CAD method using cartwheel projection analysis to detect lung nodules.
BACKGROUND OF THE INVENTION
Lung cancer has been reported as the second most commonly diagnosed cancer for both men and women, as well as the leading cause of cancer death in the United States. Although the overall cure rate for lung cancer remains quite low, the five-year survival rate for lung cancer detected and treated at an early stage is promising. Clearly, then, it would be highly desirable to detect lung cancer at an early stage.
Unfortunately, routine chest X-ray is often unsuccessful in detecting lung cancer at an early stage. However, recent advances in computerized tomography (CT) have made it feasible to detect lung nodules that otherwise would not be detected using routine chest X-ray. In CT scanning, X-ray beams are directed at an object of interest from various angles via a rotating device to obtain cross-sectional images (image slices) of lung tissue.
Multi-slice high resolution computerized tomography (MSHRCT) scanning provides a way in which nodules from 2 to 30mm or so in diameter can be imaged. However, the large amount of data presents formidable challenges to radiologists. A typical multi-slice high-resolution scan with slice thickness of 1 to 1.5 mm may have 300 or more image slices. If MSHRCT for lung cancer screening becomes widespread, there will be a tremendous demand for such examinations. It is both time-consuming and impractical for radiologists to study every single slice image.
Automatic nodule detection has attracted tremendous efforts recently. But automatic nodule detection methods often fail to detect nodules attached to blood vessels and have the further drawback of having a high false positive rate. Usually, nodules appear in slice images as nearly circular-shaped opacities, which are similar to cross-sections of vessels. Conventional automatic nodule detection methods have had great difficulty dealing with the subtlety of nodules and the camouflaging effects of normal structures.
Accordingly, it would be desirable and highly advantageous to have a computer aided diagnosis technique for detecting lung nodules that avoids the problems associated with conventional methods.
SUMMARY OF THE INVENTION
A technique is disclosed for automated detection of lung nodules, so that radiologists can be freed from the heavy burden of reading through hundreds of image slices and also so that lung nodule detection can be more accurate and less time consuming.
According to various embodiments of the present invention, cartwheel projection of image slices is performed to obtain a series of slices at different angles centered at the structure of interest. When two-dimensional images are rotated in cartwheel fashion, it is generally much easier to discover the three-dimensional shape of an object and whether it has any connecting blood vessels.
Computer analysis can be performed on the cartwheel projection slices to determine whether they show the characteristics of a lung nodule. This analysis may include shape analysis on certain cartwheel projection slices automatically selected. For each of these cartwheel projection slices, the principle axis of the object of interest can be computed by eigen-vector analysis, and then curves of the sizes/areas of the object of interest along the principle axes are created. Shapes of these curves are analyzed to determine if the object of interest is a nodule.
According to a first aspect of the invention, there is provided a method for detecting lung nodules using cartwheel projection analysis of an object of interest in a set of volumetric image data. The method includes the step of creating a set of cartwheel projection image slices by applying cartwheel projection centered at the object of interest in the set of volumetric image data. A subset of the cartwheel projection image slices is analyzed to determine whether the characteristics of a lung nodule are indicated. If so, the object of interest is identified as a lung nodule.
According to a second aspect of the invention, the method also includes the steps of extracting the object of interest for each of the cartwheel projection image slices. The circularity values for each of the extracted object of interest are then determined. The subset of cartwheel image slices is defined to include the cartwheel projection image slices with the M lowest circularity values.
According to a third aspect of the invention, the step of analyzing a subset of the cartwheel projection image slices includes creating weighted area curves for the cartwheel projection image slices in the subset of image slices with the M lowest circularity values. The shapes of the weighted area curves are analyzed to determine whether they indicate the characteristics of a lung nodule.
According to a fourth aspect of the invention, a lung nodule mask is created using cartwheel projection image slices having the N highest circularity values. The shapes of the weighted area curves are examined along the position estimated by the lung nodule mask. If the shapes of the weighted area curves along the position estimated by the lung nodule mask are Gaussian, the object of interest is considered to be a lung nodule.
According to a fifth aspect of the invention, creating the weighted area curves includes first determining the principle axis of the object of interest on each of the cartwheel projection image slices with the M lowest circularity values. Once the principle axes are determined, the sizes/areas of the object of interest along the principle axes are measured and the weighted area curves can be generated.
According to a sixth aspect of the invention, the object of interest is considered to be a lung nodule if the cartwheel projection slices with the M lowest circularity values have circularity values above a predefined threshold value.
According to a seventh aspect of the invention, the object of interest is considered not to be a lung nodule if the cartwheel projection slices with the N highest circularity values have circularity values below a predefined threshold value.
According to an eighth aspect of the invention, the volumetric image data is obtained from a multi-slice high resolution CT (MSHRCT) scan.
According to a ninth aspect of the invention, the rotation angles for the cartwheel projection are preset.
REFERENCES:
patent: 5598453 (1997-01-01), Baba et al.
patent: 5987094 (1999-11-01), Clarke et al.
Fan Li
Qian Jianzhong
Wei Guo-Qing
Lateef Marvin M
Pass Barry
LandOfFree
Pulmonary nodule detection using cartwheel projection analysis does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Pulmonary nodule detection using cartwheel projection analysis, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Pulmonary nodule detection using cartwheel projection analysis will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3103210