Image analysis – Applications – Biomedical applications
Reexamination Certificate
2000-02-15
2003-04-15
Chang, Jon (Department: 2623)
Image analysis
Applications
Biomedical applications
C382S157000, C382S199000, C382S203000, C382S205000, C382S257000, C382S274000
Reexamination Certificate
active
06549646
ABSTRACT:
TECHNICAL FIELD
The present invention relates to an automated method and system for processing digital radiological images, and more specifically, to a Divide and Conquer (DAC) method and system for the detection of abnormalities, like lung nodules, in radiological chest images using zone-based digital image processing and artificial neural network techniques.
BACKGROUND OF THE INVENTION
Lung cancer, next to heart disease, is the second highest leading cause of death in the United States. Successful detection of early-stage cancer tumors is able to increase the cure rate. Detection and diagnosis of cancerous lung nodules in chest radiographs are among the most important and difficult tasks performed by radiologists. To date, diagnosis in x-ray chest radiographs is the most important diagnostic procedure for the detection of early-stage, clinically occult lung cancer. However, the radiographic miss rate for the detection of lung nodules is quite high. Observer error, which causes these lesions to be missed, may be due to the camouflaging effect of the surrounding anatomic background on the nodule of interest, or to the subjective and varying decision criteria used by radiologists. Under-reading of a radiograph may be due to many other reasons, such as lack of clinical data, focusing of attention on another abnormality by virtue of a specific clinical question, etc. However, most peripheral lung cancers are visible in retrospect on previous films. Thus, a need remains for an automated method and system for digital image processing of radiographic images to alert radiologists to the location of highly suspected abnormality areas (SAAs).
Early radiological detection of lung nodules can significantly improve the chance of survival of lung cancer patients. A system capable of locating the presence of nodules commonly obscured by overlying ribs, bronchi, blood vessels, and other normal anatomic structures on radiographs would greatly improve the detection process. The automated system and method of the present invention allow the reduction of false negative diagnoses, and hence lead to earlier detection of lung cancers with high accuracy.
PRIOR ART
Several computer-aided diagnosis or detection (CAD) techniques using digital image processing and artificial neural networks have been described in the open literature and in patents. Of particular relevance to the present invention are the following:
Michael F. McNitt-Gray, H. K. Huang, and James W. Sayre, “Feature Selection in the Pattern Classification Problem of Digital Chest Radiograph Segmentation”,
IEEE Transactions on Medical Imaging
, Vol. 14, No. 3, September 1995, describes a method for the segmentation of digital chest radiographs using feature selection in different anatomic class definitions. McNitt-Gray et al. apply stepwise discriminant analysis and neural network techniques to segment the chest image into five anatomic zones. The five anatomic classes are: (1) heart/subdiaphram/upper mediastinum; (2) lung; (3) axilla (shoulder); (4) base of head
eck; and (5) background (area outside the patient but within the radiation field). This method was developed for use in exposure equalization. Note that the segmentation method of McNitt-Gray et al. is based on well-known knowledge of the anatomic structure of the lung region. Additionally, the zone boundaries in the McNitt-Gray et al. paper are crisply delineated and do not overlap.
Ewa Pietka, “Lung Segmentation in Digital Radiographs”,
Journal of Digital Imaging
, Vol. 7, No. 2 (May), 1994, uses a three-step algorithm involving histogram-dependent thresholding, gradient analysis, and smoothing to identify lung and non-lung regions. The method is developed for use in exposure equalization.
Jeff Duryea and John M. Boone, “A fully automated algorithm for the segmentation of lung zones on Digital Chest Radiographic Images”,
Medical Physics,
22 (2), February, 1995, describes a multi-step edge-tracing algorithm to find the lung
on-lung borders and hence to identify the lung and non-lung regions. The method is not developed for CAD purposes. The method is developed for use in exposure equalization.
Samuel G. Armato III, Maryellen L. Giger, and Heber MacMahon, “Computerized Detection of Abnormal Asymmetry in Digital Chest Radiographs”,
Medical Physics,
21 (2), November, 1994, describes an algorithm to detect abnormal asymmetry in digital chest radiographs using multi-stage gray-level thresholding. The purpose is to identify the left and right lungs and to detect large-scale abnormalities, like the asymmetry of the two lungs. The method is not developed for CAD of lung nodules.
Maria J. Carreira, Diego Cabello, and Antonio Mosquera, “Automatic Segmentation of Lung zones on Chest Radiographic Images”,
Computers and Biomedical Research
32, 1999, describes a method for automatic segmentation of lung zones in chest radiographic images. The purpose of the method is to use the lung zones as a first estimate of the area to search for lung nodules.
Neal F. Vittitoe, Rene Vargas-Voracek and Carey E. Floyd, Jr, “Identification of Lung regions in Chest Radiographs Using Markov Random Field Modeling”,
Medical Physics,
25 (6), June, 1998, presents an algorithm utilizing Markov Random Field modeling for identifying lung regions in a digitized chest radiograph. The purpose of the algorithm is to identify lung zone so that specific computer-aided diagnosis algorithms can be used to detect lung abnormalities including interstitial lung disease, lung nodules, and cardiomegaly. Note that the CAD method of Vittitoe et al. is limited to the identified lung zone and ignores the obscured lung regions, such as mediastinum, cardiac, and subdiaphragmatic areas.
Akira Hasegawa, Shih-Chung B. Lo, Jyh-Shyan Lin, Matthew T. Freedman, and Seong K. Mun, “A Shift-Invariant Neural Network for the Lung Field Segmentation in Chest Radiography”,
Journal of VLSI Signal Processing
18, 1998, describes a method of using a shift-invariant neural network to segment the chest image into lung and non-lung zones. A set of algorithms is used to refine the detected edge of the lung field. Hasegawa et al. do not further segment the lung zone into different zones. Though mentioning the potential usage of their result for the CAD applications, they discard all the pixels in the obscured areas in the lung zone. The paper suggests that CAD be applied to the non-obscured areas of the lung, while the obscured areas, such as heart, spine, and diaphragm, are excluded.
Osamu Tsujii, Matthew T. Freedman, and Seong K. Mun, “Automated Segmentation of Anatomic Regions in Chest Radiographs Using an Adaptive-sized Hybrid Neural Network”,
Medical Physics,
25 (6), June 1998 (the article also appears in SPIE,
Image Processing
, Vol. 3034, 1997), describes a method of using image features to train an adaptive-sized hybrid neural network to segment the chest image into lung and non-lung zones.
U.S. Pat. No. 4,907,156 to Doi et al. describes a method for detecting and displaying abnormal anatomic regions existing in a digital X-ray image. A single projection digital X-ray image is processed to obtain signal-enhanced image data with a maximum signal-to-noise ratio (SNR) and is also processed to obtain signal-suppressed image data with a suppressed SNR. Then, difference image data are formed by subtraction of the signal-suppressed image data from the signal-enhanced image data to remove low-frequency structured anatomic background, which is basically the same in both the signal-suppressed and signal-enhanced image data. Once the structured background is removed, feature extraction is performed. For the detection of lung nodules, pixel thresholding is performed, followed by circularity and/or size testing of contiguous pixels surviving thresholding. Threshold levels are varied, and the effect of varying the threshold on circularity and size is used to detect nodules. Pixel thresholding and contiguous pixel area thresholding are performed for the detection of lung nodules. Clusters of suspected abnormality areas are then detected.
J. S. Lin,
Li Ruiping
Lin Jyh-Shyan
Lure Yuan-Ming Fleming
Xu Xin-Wei
Yeh Hwa-Young Michael
Chang Jon
Deus Technologies LLC
Gluck Jeffrey W.
Kim Charles
Kinberg Robert
LandOfFree
Divide-and-conquer method and system for the detection of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Divide-and-conquer method and system for the detection of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Divide-and-conquer method and system for the detection of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3045589