System and method for sequential image analysis of an in vivo image stream

1. A computer-implemented method for partitioning an image stream into a plurality of segments, the method comprising:

receiving an image stream captured by an in vivo device, the image stream comprising image frames, each frame comprising a plurality of pixels;

calculating a set of pixel-based properties for each frame of at least a selected subset of frames from the image stream; and

detecting segments of constant mean values in the sets of pixel-based properties, by:

selecting windows of sequential frames in the image stream, and for each window;

determining possible partition points for the window, each partition point splitting the window of sequential frames into two corresponding sub-windows of sequential frames,

calculating, for each of the possible partition points of the window, a difference between mean values of the sets of pixel-based properties of the corresponding sub-windows of frames, and

identifying points of change in the calculated differences between the mean values of the sets of pixel-based properties;

wherein the detection of segments of constant mean values is based on the points of change; and

displaying a summarized representation of the image stream on a visual display unit, wherein the summarized representation includes a plurality of segments, each segment corresponding to a segment of the segments of constant mean values in the sets of pixel-based properties.

2. The method of claim 1 wherein displaying a summarized representation of the image stream comprises assigning a color value to each of the detected segments, and displaying the assigned colors sequentially in a color bar.

3. The method of claim 1 , wherein identifying points of change in the calculated differences between the mean values comprises:

selecting a value p for computing a p-norm value for each difference of the differences; and

calculating the p-norm value of said each difference.

4. The method of claim 1 , wherein identifying points of change in the calculated difference between the mean values further comprises comparing the calculated difference to a segmentation threshold.

5. The method of claim 4 , wherein the segmentation threshold is calculated based on a confidence value δ, a value p for computing a p-norm value, and the size of the window of frames.

6. The method of claim 1 , comprising determining a confidence value δ, wherein confidence value δ is correlated to the probability that a point of change is incorrectly identified in the calculated differences between the mean values of the sets of pixel-based properties.

7. The method of claim 1 , wherein at least one property in the set of pixel-based properties indicates a level of intestinal content depicted in the image frame, and wherein a predetermined value in the summarized representation of the image stream indicates a segment of the image stream which includes image frames with an above threshold level of intestinal content.

8. The method of claim 2 , wherein at least a subset of properties in the set of pixel-based properties indicates a mean color of the image frame, and wherein colors in the color bar are assigned based on the mean color of the corresponding detected segment.

9. The method of claim 1 , wherein the set of pixel-based properties indicates a diameter of an intestinal lumen hole detected in the image frame, and wherein segments of the image stream which include image frames with an intestinal lumen diameter within a predetermined range are mapped to predetermined colors in a color bar.

10. The method of claim 1 , wherein the set of pixel-based properties indicates a contraction density detected in the image frame, and wherein segments of the image stream which include image frames with a contraction density within a predetermined range are mapped to predetermined colors in a color bar.

11. The method of claim 1 , comprising calculating difference measures which correlate to the differences between the mean values, and identifying points of change in the difference measures.

12. A computer-implemented method for segmenting an image stream to a plurality of segments, the method comprising:

receiving an image stream captured by an in vivo device, the image stream comprising image frames, each frame comprising a plurality of pixels;

based on the frame pixels, calculating a feature vector for each frame of at least a selected subset of frames from the image stream;

determining possible splits of at least a portion of the image stream into two sequential series of frames;

calculating mean values for the feature vectors of each sequential series of frames of a possible split of the image stream;

calculating difference measures based on differences between the mean values of the feature vectors of each sequential series of frames of a possible split of the image stream to identify points of change in the mean values; and

segmenting the image stream according to the identified points of change.

13. The method of claim 12 comprising displaying a summarized representation of the image stream on an image monitor, wherein the summarized representation includes a plurality of segments corresponding to the segmenting.

14. The method of claim 13 , wherein displaying a summarized representation comprises assigning a color value to segments and displaying the assigned colors sequentially in a color bar.

15. The method of claim 12 , wherein the feature vectors comprise pixel-based properties.

16. A system for segmenting an image stream to a plurality of segments, the system comprising:

a memory; and

a computer processor configured to:

receive image frames captured by an in vivo imaging device, each frame comprising a plurality of pixels;

calculate a set of pixel-based properties for each frame of at least a selected subset of frames from the image stream;

detect segments of constant mean values in the sets of pixel-based properties, by:

selecting windows of sequential frames in the received image frames, and for each window;

determining possible partition points for the window, each partition point splitting the window of sequential frames into two corresponding sub-windows of sequential frames,

calculating, for each of the possible partition points of the window, a difference between mean values of the sets of pixel-based properties of the corresponding sub-windows of frames, and

identifying points of change in the calculated differences between the mean values of the sets of pixel-based properties;

wherein the detection of segments of constant mean values is based on the points of change; and

a display unit to display a summarized representation of the image stream, wherein the summarized representation includes a plurality of segments, each segment corresponding to a segment of the segments of constant mean values in the sets of pixel-based properties.

17. The system of claim 16 , wherein identifying points of change in the calculated differences between the mean values further comprises comparing the calculated differences to a segmentation threshold.

18. The system of claim 16 , wherein the computer processor is further configured to determine a confidence value δ, wherein confidence value δ is correlated to the probability that a point of change is incorrectly identified.

19. The system of claim 16 , wherein displaying a summarized representation of the image stream comprises assigning a color value to each of the detected segments, and displaying the assigned colors sequentially in a color bar.

20. The system of claim 16 , wherein identifying points of change in the calculated differences between the mean values comprises:

selecting a value p for computing a p-norm value for each difference of the differences; and

calculating the p-norm value of said each difference.