Automatic eardrum registration from light field data
A three-dimensional image of the eardrum is automatically registered. In one aspect, a three-dimensional image (e.g., depth map) of an eardrum is produced from four-dimensional light field data captured by a light field otoscope. The three-dimensional image is registered according to a predefined standard form. The eardrum is then classified based on the registered three-dimensional image. Registration may include compensation for out-of-plane rotation (tilt), for in-plane rotation, for center location, for translation, and/or for scaling.
1. A method for automatic classification of eardrums from light field data, the method performed by a computer system having a processor and memory, the method comprising:
producing a three-dimensional image of an eardrum from light field data captured for the eardrum;
registering the three-dimensional image of the eardrum as defined by a predefined standard form; wherein the predefined standard form includes a location of a center of the eardrum, and registering the three-dimensional image of the eardrum comprises determining the location of the center of the eardrum in the three-dimensional image by:
selecting a subset of the three-dimensional image of the eardrum that have a depth below a predefined threshold; and
analyzing the selected subset to determine the location of the center of the eardrum; and
classifying the eardrum based on the registered three-dimensional image of the eardrum.
2. The method of claim 1 wherein the center of the eardrum is defined as a location of the umbo.
3. The method of claim 1 wherein the center of the eardrum is defined as a location with a most negative depth.
4. The method of claim 1 wherein determining the location of the center of the eardrum in the three-dimensional image comprises:
identifying depressed areas of the eardrum defined by closed isocontours of negative depth; and
determining the location of the center of the eardrum based on the identified depressed areas.
5. The method of claim 1 wherein the predefined standard form includes a predefined rotational orientation of the three-dimensional image of the eardrum, and registering the three-dimensional image of the eardrum further comprises rotating the three-dimensional image of the eardrum to the predefined rotational orientation.
6. The method of claim 5 wherein the predefined rotational orientation is defined relative to an anatomical landmark on the eardrum.
7. The method of claim 5 wherein the predefined rotational orientation is defined by a tilt of the eardrum in an ear canal.
8. The method of claim 1 wherein the predefined standard form includes a predefined size of the three-dimensional image of the eardrum, and registering the three-dimensional image of the eardrum further comprises scaling the three-dimensional image of the eardrum to match the predefined size.
9. The method of claim 1 wherein the predefined standard form identifies left ear or right ear as standard, and registering the three-dimensional image of the eardrum further comprises producing a mirror image of the three-dimensional image of the eardrum if the three-dimensional image is not for the standard ear.
10. The method of claim 1 wherein the three-dimensional image of the eardrum is segmented to include only the eardrum and excludes structures beyond the eardrum.
11. A method for automatic classification of eardrums from light field data, the method performed by a computer system having a processor and memory, the method comprising:
producing a three-dimensional image of an eardrum from light field data captured for the eardrum;
registering the three-dimensional image of the eardrum as defined by a predefined standard form; wherein the predefined standard form includes a location of a center of the eardrum, and registering the three-dimensional image of the eardrum comprises determining the location of the center of the eardrum by:
identifying depressed areas of the eardrum defined by closed isocontours of negative depth;
selecting the largest of the depressed areas; and
determining the location of the center of the eardrum within the largest depressed area; and
classifying the eardrum based on the registered three-dimensional image of the eardrum.
12. The method of claim 11 wherein determining the location of the center of the eardrum in the three-dimensional image further comprises:
applying graph-based clustering to identify the depressed areas of the eardrum; and
determining the location of the center of the eardrum based on the identified depressed areas.
13. The method of claim 11 wherein determining the location of the center of the eardrum in the three-dimensional image comprises:
attempting to identify first depressed areas defined by closed isocontours of negative depth;
if the first depressed areas are identified, then determining the location of the center of the eardrum based on the identified first depressed areas;
if no first depressed areas are identified, then applying graph-based clustering to identify second depressed areas and determining the location of the center of the eardrum based on the identified second depressed areas.
14. A method for automatic classification of eardrums from light field data, the method performed by a computer system having a processor and memory, the method comprising:
producing a three-dimensional image of an eardrum from light field data captured for the eardrum;
registering the three-dimensional image of the eardrum as defined by a predefined standard form; wherein the predefined standard form includes an untilted three-dimensional image of the eardrum, and registering the three-dimensional image of the eardrum comprises removing tilt from the three-dimensional image of the eardrum by:
determining a plane that is a best fit to the three-dimensional image of the eardrum;
calculating a tilt of the best fit plane; and
removing the calculated tilt from the three-dimensional image of the eardrum; and
classifying the eardrum based on the registered three-dimensional image of the eardrum.
15. The method of claim 14 wherein the predefined standard form includes a predefined rotational orientation of the three-dimensional image of the eardrum, and registering the three-dimensional image of the eardrum further comprises rotating the three-dimensional image of the eardrum to the predefined rotational orientation.
16. The method of claim 15 wherein the predefined rotational orientation is defined relative to an anatomical landmark on the eardrum.
17. The method of claim 15 wherein the predefined rotational orientation is defined by the tilt of the eardrum in an ear canal.
18. The method of claim 14 wherein the predefined standard form includes a predefined size of the three-dimensional image of the eardrum, and registering the three-dimensional image of the eardrum further comprises scaling the three-dimensional image of the eardrum to match the predefined size.
19. The method of claim 14 wherein the predefined standard form identifies left ear or right ear as standard, and registering the three-dimensional image of the eardrum further comprises producing a mirror image of the three-dimensional image of the eardrum if the three-dimensional image is not for the standard ear.
20. The method of claim 14 wherein the three-dimensional image of the eardrum is segmented to include only the eardrum and excludes structures beyond the eardrum.