Image feature annotation in diagnostic imaging
The present invention relates to an image processing device ( 10 ) comprising a data input ( 11 ) for receiving volumetric image data comprising a plurality of registered volumetric images of an imaged object, a noise modeler ( 12 ) for generating a noise model indicative of a spatial distribution of noise in each of the plurality of registered volumetric images, a feature detector ( 13 ) for detecting a plurality of image features taking the volumetric image data into account, and a marker generator ( 14 ) for generating a plurality of references indicating feature positions of a subset of the plurality of detected image features, in which said subset corresponds to the detected image features that are classified as difficult to discern on a reference volumetric image in the plurality of registered volumetric images based on a classification and/or a visibility criterium, wherein the classification and/or the visibility criterium takes the or each noise model into account.
1. An image processing device, comprising:
a data input for receiving volumetric image data organized in voxels, the volumetric image data comprising a plurality of registered volumetric images of an imaged object;
a noise modeler for generating a noise model indicative of a spatial distribution of noise in each of the plurality of registered volumetric images;
a feature detector for detecting a plurality of image features taking said volumetric image data into account; and
a marker generator for generating a plurality of references indicating feature positions of a subset of the plurality of detected image features, wherein said subset corresponds to the detected image features that are classified as difficult to discern on a reference volumetric image in the plurality of registered volumetric images based on a classification and/or a visibility criterium, wherein the classification and/or the visibility criterium takes the noise model into account.
2. The image processing device of claim 1 , wherein the data input is configured to receive the volumetric image data in the form of the plurality of registered volumetric images obtained and/or derived from a single spectral computed tomography acquisition sequence.
3. The image processing device of claim 2 , wherein the reference volumetric image is representative of a conventional computed tomography image as obtained by or derived from the spectral computed tomography acquisition sequence.
4. The image processing device of claim 1 , wherein the feature detector is configured to identify a location and a scale of each detected image feature, and wherein the marker generator is configured to determine a plurality of visibility criteria for each detected feature taking the identified location and the identified scale of each detected image feature into account.
5. The image processing device of claim 4 , wherein the marker generator is configured to classify a detected image feature as difficult to discern on the reference volumetric image if a predetermined number of the visibility criteria is met.
6. The image processing device of claim 1 , wherein the marker generator is configured to estimate a contrast to noise ratio for each image feature in at least the image in which the image feature was detected and/or to estimate a contrast to noise ratio for each image feature in the reference image.
7. The image processing device of claim 1 , wherein the marker generator is configured to, for each image feature, calculate a standard deviation of the voxel values in a region around the location of the image feature in the image in which the image feature was detected.
8. The image processing device of claim 1 , wherein the marker generator is configured to, for each image feature, calculate a normalized cross-correlation and/or a mutual information between regions around the location of the image feature in respectively the reference image and at least the image in which the image feature was detected.
9. The image processing device of claim 1 , comprising a segmentation unit for performing a segmentation of the structure corresponding to a detected feature in at least the image in which the feature was detected.
10. The image processing device of claim 1 , wherein the feature detector is configured to calculate, for each of the plurality of volumetric images or each of a subset thereof, a scale-normalized Laplacian of Gaussian.
11. The image processing device of claim 1 , wherein the noise modeler is configured to estimate the noise model using a Monte-Carlo estimation method, an analytical method, and/or a direct extraction technique.
12. A method for generating a plurality of references indicating image features that are classified as difficult to discern on a reference volumetric image, the method comprising:
receiving volumetric image data organized in voxels, the volumetric image data comprising a plurality of registered volumetric images of an imaged object;
generating a noise model indicative of a spatial distribution of noise in each of the plurality of registered volumetric images;
detecting a plurality of image features taking the volumetric image data into account; and
generating the plurality of references indicating feature positions of a subset of the plurality of detected image features, wherein the subset corresponds to the detected image features that are classified as difficult to discern on a reference volumetric image in the plurality of registered volumetric images based on a numerical classification and/or a visibility criterium, wherein the classification and/or the visibility criterium takes the noise model into account.
13. A non-transitory computer readable storage medium encoded with one or more computer executable instructions, which, when executed by a processor of a computing system, causes the computing system to perform a method in accordance with claim claim 12 .