System and method for computer-aided detection and characterization of diffuse lung disease

1. A method for detecting the presence of an abnormality in at least one medical image, comprising:

obtaining image data including pixels of an organ;

segmenting the image data into organ image data and non-organ image data;

extracting predetermined features from said organ image data to produce a set of image features;

comparing said set of image features with a reference set of organ image features derived from known abnormal image data and known normal image data; and

producing a comparison result, wherein said comparing said set of image features comprises comparing with reference organ features selected from the set consisting of:

a normal opacity;

a ground-glass opacity;

a reticular/linear opacity;

a nodular opacity;

a honeycombing pattern;

an emphysematous change pattern; and

a consolidation pattern.

2. The method of claim 1 , wherein said obtaining image data comprises:

obtaining lung image data.

3. The method of claim 1 , wherein said obtaining image data comprises:

obtaining a high-resolution (HR) computed tomography (CT) image.

4. The method of claim 1 , wherein said comparing step comprises:

comparing with an artificial neural network (ANN).

5. The method of claim 1 , wherein said comparing step comprises:

comparing with a Bayesian classifier.

6. The method of claim 1 , wherein said comparing said set of image features comprises comparing with reference organ features selected from the set consisting of:

a non-specific diffuse opacity; and

an abnormality including each of

atelectasis,

pleural thickening,

bronchectasis,

pleural effusion,

bulla,

a focal lung lesion, and

an artifact.

7. The method of claim 1 , wherein said extracting step comprises:

determining at least one measure of gray-level distribution of pixel values in one of a 2D region of interest (ROI) and a 3D volume of interest (VOI); and

determining at least one geometric feature.

8. The method of claim 7 , wherein said determining at least one measure of gray-level distribution comprises:

determining at least one of a mean, a standard deviation, and a fraction of an area with air density components.

9. The method of claim 7 , wherein said determining at least one measure of gray-level distribution comprises:

determining each of a mean, a standard deviation, and a fraction of an area with air density components.

10. The method of claim 8 , wherein said determining a fraction of an area with air density components comprises:

determining an area having CT values between −910 and −1000 HU, inclusively.

11. The method of claim 7 , wherein said determining at least one geometric feature comprises:

determining at least one of a nodular geometric feature, a line geometric feature, and a multi-locular geometric feature.

12. The method of claim 7 , wherein said determining at least one geometric feature comprises:

determining each of a nodular geometric feature, a line geometric feature, and a multi-locular geometric feature.

13. The method of claim 11 , wherein said determining a nodular geometric feature comprises:

applying a morphological white top-hat transform to the one of a 2D ROI and a 3D VOI to produce a nodule candidate; and

calculating one of a degree of circularity and a degree of sphericity of said nodule candidate to produce a nodule candidate.

14. The method of claim 11 , wherein said determining a line geometric feature comprises:

applying a morphological white top-hat transform to said one of a 2D ROI and a 3D VOI to produce a nodule candidate;

calculating one of a degree of circularity and a degree of sphericity to said nodule candidate to produce a nodule estimate; and

applying a gray-level threshold to said nodule estimate to produce a thresholded nodule estimate.

15. The method of claim 11 , wherein said determining a multi-locular geometric feature comprises:

applying a morphological black top-hat transform to the one of a 2D ROI and a 3D VOI to produce a nodule candidate; and

calculating a standard deviation of said nodule candidate.

16. The method of claim 7 , wherein said determining at least one measure comprises:

determining a measure of gray-level distribution for a first 2D ROI; and

determining a measure of gray-level distribution for a second 2D ROI, said second 2D ROI selectively set to be larger than said first 2D ROI.

17. The method of claim 7 , wherein said determining at least one measure comprises:

determining a measure of gray-level distribution for a first 3D VOI; and

determining a measure of gray-level distribution for a second 3D VOI, said second 3D VOI selectively set to be larger than said first 3D VOI.

18. The method of claim 7 , wherein said 2D region of interest (ROI) comprises:

determining a measure of gray-level distribution for a 16×16 mm ROI; and

determining a measure of gray-level distribution for a 48×48 mm ROI.

19. The method of claim 4 , wherein said comparing with an ANN comprises:

comparing with an ANN having

12 input units,

10 hidden units, and

7 output units.

20. The method of claim 7 , wherein said pixel values comprise CT values.

21. The method of claim 1 , wherein said extracting step comprises:

determining at least one feature from a gray-level distribution of pixel values one of in a 2D region of interest (ROI) and a 3D volume of interest (VOI).

22. The method of claim 21 , wherein said determining step comprises:

determining a measure of gray-level distribution for a first 2D ROI; and

determining a measure of gray-level distribution for a second 2D ROI, said second 2D ROI selectively set to be larger than said first 2D ROI.

23. The method of claim 21 , wherein said determining step comprises:

determining a measure of gray-level distribution for a first 3D VOI; and

determining a measure of gray-level distribution for a second 3D VOI, said second 3D VOI selectively set to be larger than said first 3D VOL.

24. The method of claim 1 , wherein said obtaining step comprises:

obtaining 2D data including data derived from one of an axial plane, a sagittal plane, and a coronal plane.

25. The method of claim 1 , wherein said obtaining step comprises:

obtaining 3D data including data derived from an axial plane, a sagittal plane, and a coronal plane.

26. A method for detecting the presence of an abnormality in at least one medical image, comprising:

obtaining image data including pixels of an organ;

segmenting the image data into organ image data and non-organ image data;

extracting predetermined features from said organ image data to produce a set of image features;

comparing said set of image features with a reference set of organ image features derived from known abnormal image data and known normal image data; and

producing a comparison result, wherein said comparing said set of image features comprises comparing with each of:

a normal opacity;

a ground-glass opacity;

a reticular/linear opacity;

a nodular opacity;

a honeycombing pattern;

an emphysematous change pattern; and

a consolidation pattern.

27. The method of claim 26 , wherein said comparing said set of image features comprises comparing with each of:

a non-specific diffuse opacity; and

atelectasis;

pleural thickening;

bronchectasis;

pleural effusion;

bulla;

a focal lung lesion; and

an artifact.

28. A system for implementing the method recited in any one of claims 1 - 27 .

29. A computer readable medium storing instructions for execution on a computer system, which when executed by the computer system, causes performance of the method recited in any one of claims 1 - 27 .