CASCADE ANALYSIS FOR INTESTINAL CONTRACTION DETECTION

• US 20090284589A1
• Filed: 03/13/2007
• Published: 11/19/2009
• Est. Priority Date: 03/13/2006
• Status: Active Grant

• Current Assignee: Given Imaging Limited (Medtronic Plc)
• Original Assignee: Given Imaging Limited (Medtronic Plc)
• Inventors: Accarino, Anna, De Iorio, Fosca, Spyridonos, Panagiota, Malagelada, Juan, Vilarino, Fernando, Azpiroz, Fernando, Radeva, Petia, Vitria, Jordi
• Granted Patent: US 8,441,530 B2
• Time in Patent Office: Days
• US Class Current: 348/77
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10068 Endoscopic image
• CPC Class Codes: G06T 2207/30028 Colon; Small intestine
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/136 involving thresholding

Abstract

A method and system cascade analysis for intestinal contraction detection is provided by extracting from image frames captured in-vivo. The method and system also relate to the detection of turbid liquids in intestinal tracts, to automatic detection of video image frames taken in the gastrointestinal tract including a field of view obstructed by turbid media, and more particularly to extraction of image data obstructed by turbid media.

Claims

1-2. -2. (canceled)

3. A method to extract portions of an image stream comprising:

detecting an obstructed portion of the image frame;

setting a threshold for obstructed image frames, to determine invalid frames; and

removing the invalid frames from the image stream.

4. The method of claim 3 wherein the obstructed frame is a frame with turbid content or a wall frame or a tunnel frame.

5. The method of claim 3 wherein the invalid frames are automatically removed from the image stream.

6. The method of claim 3 wherein detecting an obstructed portion of the image frame is performed by detecting textural descriptors.

7. The method of claim 6 wherein the textural descriptors include a texture pattern consisting of bubbles.

8. The method of claim 3 wherein Gabor filters are used for detecting an obstructed portion of the image frame.

9. The method of claim 3 wherein detecting an obstructed portion of the image frame is based on color information of a given image stream.

10. A method for automatic classification of frames in a series of video frames comprising:

(i) receiving a series of video frames from an in-vivo device;

(ii) constructing an initial training set of the video based on a predefined dynamic pattern;

(iii) training a classifier with the initial training set comprising labelled data;

(iv) using the classifier to classify unlabeled data;

(v) adding to the initial training set a most confident unlabeled point and its predictive label, to obtain an enhanced training set;

(vi) re-training the classifier with the enhanced training set; and

(vii) classifying frames having the predefined dynamic pattern from the series of video frames.

11. The method of claim 10, wherein the predefined dynamic pattern is selected from a group consisting of:

color characterization, global intensity, turbidity, presence of blood and presence of bubbles.

12. The method of claim 10, wherein classifying is based on a support vector machine.

13. A system for detection of obstructed in vivo image data comprising:

an imaging device for capturing in vivo images of a body lumen; and

a turbid media detector to detect obstructed frames in the images.

14. The system of claim 13 wherein the turbid media detector is self-trained to detect intestinal juices, bubbles and content.

15. The system of claim 14 wherein the turbid media detector is trained based on color information of a given stream of images.

16. The system of claim 13 further comprising a dynamic event detector to detect a pattern of global intensity of image frames in an image frame sequence of a predefined length.

17. The system of claim 13 further comprising a wall detector to automatically detect frames in which the in-vivo imaging device obtains images of a lumen wall.

18. The system of claim 13 further comprising a tunnel detector to automatically detect frames in which the lumen appears static.

19. The system of claim 13 further comprising a contraction detector to automatically detect image frames captured during contractile activity.

20. The system of claim 13 wherein the in vivo imaging device is a swallowable capsule.

21. A machine-learning system to automatically learn and classify intestinal contractile activity comprising:

a contraction detector to select, from a stream of in-vivo images, image frames that follow a predefined dynamic pattern; and

a classifier to classify contractile activity in the selected image frames, based on duration, number, geometrical pattern, frequency or distribution of contractions along the intestinal tract.

22. The system of claim 21, wherein the classifier is a support vector machine classifier trained with a labeled data set.

23. The system of claim 21, further comprising a display unit to display an indication of contractile activity in an analyzed image stream.