Tube detection in diagnostic images
View Patent ↗A method for detecting tubing in a radiographic image of a patient, executed at least in part by a control logic processor, obtains radiographic image data for a patient and identifies a region of interest in the radiographic image. A gradient magnitude image of the region of interest is formed and analyzed to identify one or more linear features by defining a band lying substantially within the region of interest and having a center point and repeating a sequence with two or more iterations of assigning a rotation angle for the rotatable band about the center point and computing the ensemble average of gradient magnitude values along each of a plurality of lines extending within the rotatable band at the defined rotation angle, then computing relative magnitudes for the lines. The one or more identified linear features are evaluated according to the results of the ensemble average computing.
1. A method for detecting tubing in a radiographic image of a patient, comprising:
obtaining radiographic image data for a patient;
identifying a region of interest in the radiographic image;
forming a gradient magnitude image of the region of interest;
analyzing the gradient magnitude image to identify one or more linear features by defining a band lying substantially within the region of interest and having a center point and repeating a sequence comprising two or more iterations of:
(i) assigning a rotation angle for the rotatable band about the center point; and
(ii) computing the ensemble average of gradient magnitude values along each of a plurality of lines extending within the rotatable band at the defined rotation angle and computing relative magnitudes for the lines; and
evaluating the one or more identified linear features according to the results of the ensemble average computing.
2. The method of claim 1 further comprising computing a ratio of magnitudes for one or more lines within a rotatable band.
3. The method of claim 1 further comprising analyzing the one or more identified linear features to remove one or more false-positive tubing candidates.
4. The method of claim 3 wherein analyzing the one or more identified linear features to remove one or more false-positive tubing candidates comprises analyzing one or more of curvature, length, position relative to patient anatomy, detected tip position, and tube width standard deviation using quadratic discriminant analysis.
5. The method of claim 1 further comprising combining two or more identified linear features to form a tube segment.
6. The method of claim 5 further comprising extending the tube segment by:
using first-order or third-order polynomial curve fitting of the at least one detected tube segment; and
iteratively extending from one or both ends of the tube segment by identifying a group of consecutive pixels along a fitted curve.
7. The method of claim 6 further comprising displaying the identified tube segment.
8. The method of claim 7 wherein displaying the identified tube segment comprises highlighting the identified tube segment on a display screen.
9. The method of claim 5 wherein combining two or more identified linear features is done according to the distance between them.
10. The method of claim 1 further comprising performing contrast enhancement on the radiographic image.
11. The method of claim 1 wherein identifying a region of interest in the radiographic image further comprises:
identifying one or more linear features in each of a plurality of portions of the image in a sequence that comprises, for each portion:
forming a gradient magnitude image within the portion of the image;
analyzing the gradient magnitude image to identify one or more linear features by defining a band lying substantially within the portion of the image and having a center point and repeating a sequence comprising two or more iterations of:
(i) assigning a rotation angle for the rotatable band; and
(ii) computing the ensemble average of gradient magnitude values along each of a plurality of lines extending within the rotatable band at the defined rotation angle and computing relative magnitudes for the lines;
evaluating the one or more identified linear features within the portion; and
comparing the identified linear features from neighboring individual portions to identify the region of interest.
12. A method for identifying a region of interest in a radiographic image comprising:
identifying one or more linear features in each of a plurality of portions of the image in a sequence that comprises, for each portion;
forming a gradient magnitude image within the portion of the image;
analyzing the gradient magnitude image to identify one or more linear features by defining a band lying substantially within the portion of the image and having a center point and repeating a sequence comprising two or more iterations of:
(i) assigning a rotation angle for the rotatable band; and
(ii) computing the ensemble average of gradient magnitude values along each of a plurality of lines extending within the rotatable band at the defined rotation angle and computing relative magnitudes for the lines to detect the one or more identified linear features within the portion; and
comparing the identified linear features from neighboring individual portions to identify the region of interest.