Computerized method for determination of the likelihood of malignancy for pulmonary nodules on low-dose CT

• US 6,891,964 B2
• Filed: 11/23/2001
• Issued: 05/10/2005
• Est. Priority Date: 11/23/2001
• Status: Expired due to Fees

• Current Assignee: University of Chicago The (University of Chicago)
• Original Assignee: University of Chicago
• Inventors: Aoyama, Masahito, Li, Qiang, Doi, Kunio
• Primary Examiner(s): Dastouri, Mehrdad
• Application Number:
  US09/990,310
• Publication Number:
  US 20030099387A1
• Time in Patent Office: 1,264 Days
• Field of Search: 382/130, 382/131, 382/132, 382/190, 382/195, 382/170, 382/199, 382/263, 600/407
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 7/0012 Biomedical image inspection

Abstract

An automated computerized scheme for determination of the likelihood of malignancy in pulmonary nodules. The present invention includes steps of obtaining at least one computed tomography medical image of a pulmonary nodule in determining if the pulmonary nodule is malignant based on the examination of seven patient or image features. The method can be implemented when instructions are loaded into a computer to program the computer. The significance of employing seven patient or image features is that statistically, seven features are the most practical based on the unique implementation of statistical analysis. Out of the seven features that are now analyzed to determine if a pulmonary nodule is malignant, these features are selected to optimize the accuracy of the diagnosis of a pulmonary nodule. Through a unique sampling scheme, different embodiments of the present invention utilize different combinations of features to optimize the accuracy of the method of the present invention.

Claims

1. A method for determining if a pulmonary nodule is malignant, comprising the steps of:

obtaining at least one computed tomography medical image of the pulmonary nodule;

obtaining at least one patient feature of a patient having the pulmonary nodule;

extracting image features of the pulmonary nodule from the at least one computed tomography medical image; and

evaluating whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features, wherein

the step of obtaining the patient features consists of obtaining the sex of the patient; and

the step of extracting image features consists of extracting effective diameter of the pulmonary nodule, contrast of the pulmonary nodule, overlap measure of two gray-level histograms for the inside and outside regions of a segmented nodule of the medical image, overlap measure of two gray-level histograms for the inside and outside region of a segmented nodule of an edge gradient of the medical image, radial gradient index for an inside region of a segmented nodule of the medical image, and peak value of a histogram for an inside regions of a segmented nodule of an edge gradient of the medical image.

2. The method of claim 1, wherein the step of obtaining at least one computed tomography medical image comprises obtaining a medical X-ray image.

3. The method of claim 2, wherein the step of obtaining at least one computed tomography medical image comprises obtaining at least one low-dose computed tomography image.

4. The method of claim 1, wherein the step of obtaining at least one patient feature comprises obtaining at least one of sex and age of the patient having the pulmonary nodule.

5. The method of claim 1, wherein the step of extracting image features comprises at least one of the steps of:

identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

identifying image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule.

6. The method of claim 5, wherein step of identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule comprises at least one of the following steps:

identifying the effective diameter of the outline;

identifying the degree of circularity of the outline;

identifying the degree of ellipticity of the outline;

identifying the degree of irregularity of the outline;

identifying the root-mean-square variation of the power spectrum of the outline; and

identifying the power spectrum of the outline.

7. The method of claim 5, wherein the step of identifying image features based on the linear patterns of the at least one computed tomography medical image of the pulmonary nodule comprises:

identifying the magnitude of line patterns for inside region of a segmented nodule of the medical image; and

identifying the magnitude of line patterns for outside region of a segmented nodule of the medical image.

8. The method of claim 5, wherein the step of identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule comprises:

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the inside of a segmented nodule of the medical image;

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the outside of a segmented nodule of the medical image;

identifying tangential gradient index computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the inside of a segmented nodule of the medical image; and

identifying tangential gradient computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the outside of a segmented nodule of the image.

9. The method of claim 1, wherein the step of extracting image features consists of extracting the effective diameter of the pulmonary nodule and the contrast of the pulmonary nodule.

10. The method of claim 1, wherein the step of obtaining at least one medical image comprises:

obtaining three medical images; and

the step of evaluating if the pulmonary nodules are malignant is based on the examination of the three medical images.

11. A computer readable medium storing computer program instructions for determining if a pulmonary nodule is malignant, which when used to program a computer to cause the computer to perform the steps of:

obtaining at least one computed tomography medical image of the pulmonary nodule;

obtaining at least one patient feature of a patient having the pulmonary nodule;

extracting image features of the pulmonary nodule from the at least one computed tomography medical image; and

evaluating whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features, wherein

the step of obtaining the patient features consists of obtaining the sex of the patient; and

the step of extracting image features consists of extracting effective diameter of the pulmonary nodule, contrast of the pulmonary nodule, overlap measure of two gray-level histograms for the inside and outside regions of a segmented nodule of the medical image, overlap measure of two gray-level histograms for the inside and outside region of a segmented nodule of an edge gradient of the medical image, radial gradient index for an inside region of a segmented nodule of the medical image, and peak value of a histogram for an inside regions of a segmented nodule of an edge gradient of the medical image.

12. The computer readable medium of claim 11, wherein the step of obtaining at least one computed tomography medical image comprises obtaining a medical X-ray image.

13. The computer readable medium of claim 12, wherein the step of obtaining at least one computed tomography medical image comprises obtaining at least one low-dose computed tomography image.

14. The computer readable medium of claim 11, wherein the step of obtaining at least one patient feature comprises obtaining at least one of sex and age of the patient having the pulmonary nodule.

15. The computer readable medium of claim 11, wherein the step of extracting image features comprises at least one of the steps of:

identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

identifying image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule.

16. The computer readable medium of claim 15, wherein step of identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule comprises at least one of the following steps:

identifying the effective diameter of the outline;

identifying the degree of circularity of the outline;

identifying the degree of ellipticity of the outline;

identifying the degree of irregularity of the outline;

identifying the root-mean-square variation of the power spectrum of the outline; and

identifying the power spectrum of the outline.

17. The computer readable medium of claim 15, wherein the step of identifying image features based on the linear patterns of the at least one computed tomography medical image of the pulmonary nodule comprises:

identifying the magnitude of line patterns for inside region of a segmented nodule of the medical image; and

identifying the magnitude of line patterns for outside region of a segmented nodule of the medical image.

18. The computer readable medium of claim 15, wherein the step of identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule comprises:

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the inside of a segmented nodule of the medical image;

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the outside of a segmented nodule of the medical image;

identifying tangential gradient index computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the inside of a segmented nodule of the medical image; and

identifying tangential gradient computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the outside of a segmented nodule of the image.

19. The computer readable medium of claim 11, wherein the step of extracting image features consists of extracting the effective diameter of the pulmonary nodule and the contrast of the pulmonary nodule.

20. The computer readable medium of claim 11, wherein the step of obtaining at least one medical image comprises:

obtaining three medical images; and

the step of evaluating if the pulmonary nodules are malignant is based on the examination of the three medical images.

21. A system for determining if a pulmonary nodule is malignant, comprising:

a mechanism configured to obtain at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to obtain at least one patient feature of a patient having the pulmonary nodule;

a mechanism configured to extract image features of the pulmonary nodule from the at least one computed tomography medical image; and

a mechanism configured to evaluate whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features, wherein

the mechanism configured to obtain the patient features consists of a mechanism for obtaining the sex of the patient; and

the mechanism configured to extract image features consists of a mechanism configured to extract effective diameter of the pulmonary nodule, contrast of the pulmonary nodule, overlap measure of two gray-level histograms for the inside and outside regions of a segmented nodule of the medical image, overlap measure of two gray-level histograms for the inside and outside region of a segmented nodule of an edge gradient of the medical image, radial gradient index for an inside region of a segmented nodule of the medical image, and peak value of a histogram for an inside regions of a segmented nodule of an edge gradient of the medical image.

22. The system of claim 21, wherein the mechanism configured to obtain at least one computed tomography medical image comprises a mechanism configured to obtain a medical X-ray image.

23. The system of claim 22, wherein the mechanism configured to obtain at least one computed tomography medical image comprises a mechanism configured to obtain at least one low-dose computed tomography image.

24. The system of claim 21, wherein the mechanism configured to obtain at least one patient feature comprises a mechanism configured to obtain at least one of sex and age of the patient having the pulmonary nodule.

25. The system of claim 21, wherein the mechanism for extracting image features comprises at least one of:

a mechanism configured to identify image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

a mechanism configured to identify image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

a mechanism configured to identify image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule.

26. The system of claim 25, wherein the mechanism configured to identify image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule comprises at least one of the following:

a mechanism configured to identify the effective diameter of the outline;

a mechanism configured to identify the degree of circularity of the outline;

a mechanism configured to identify the degree of ellipticity of the outline;

a mechanism configured to identify the degree of irregularity of the outline;

a mechanism configured to identify the root-mean-square variation of the power spectrum of the outline; and

a mechanism configured to identify the power spectrum of the outline.

27. The system of claim 25, wherein the mechanism configured to identify image features based on the linear patterns of the at least one computed tomography medical image of the pulmonary nodule comprises:

a mechanism configured to identify the magnitude of line patterns for inside region of a segmented nodule of the medical image; and

a mechanism configured to identify the magnitude of line patterns for outside region of a segmented nodule of the medical image.

28. The system of claim 25, wherein the mechanism configured to identify image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule comprises:

a mechanism configured to identify radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the inside of a segmented nodule of the medical image;

a mechanism configured to identify radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the outside of a segmented nodule of the medical image;

a mechanism configured to identify tangential gradient index computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the inside of a segmented nodule of the medical image; and

a mechanism configured to identify tangential gradient computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the outside of a segmented nodule of the image.

29. The system of claim 21, wherein the mechanism configured to extract image features consists of a mechanism configured to extract the effective diameter of the pulmonary nodule and the contrast of the pulmonary nodule.

30. The system of claim 21, wherein the mechanism configured to obtain at least one medical image comprises:

a mechanism configured to obtain three medical images; and

the mechanism configured to evaluate if the pulmonary nodules are malignant is based on the examination of the three medical images.

31. A method for determining if a pulmonary nodule is malignant, comprising:

obtaining at least one computed tomography medical image of the pulmonary nodule;

obtaining at least one patient feature of a patient having the pulmonary nodule;

extracting image features of the pulmonary nodule from the at least one computed tomography medical image; and

evaluating whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features,

wherein the step of extracting image features comprises at least one of the steps of

identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

identifying image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule, and

wherein the step of identifying image features based on the linear patterns of the at least one computed tomography medical image of the pulmonary nodule comprises

identifying the magnitude of line patterns for inside region of a segmented nodule of the medical image; and

identifying the magnitude of line patterns for outside region of a segmented nodule of the medical image.

32. A method for determining if a pulmonary nodule is malignant, comprising:

obtaining at least one computed tomography medical image of the pulmonary nodule;

obtaining at least one patient feature of a patient having the pulmonary nodule;

extracting image features of the pulmonary nodule from the at least one computed tomography medical image; and

evaluating whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features,

wherein the step of extracting image features comprises at least one of the steps of

identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

identifying image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule, and

the step of identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule comprises

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the inside of a segmented nodule of the medical image;

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the outside of a segmented nodule of the medical image;

identifying tangential gradient index computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the inside of a segmented nodule of the medical image; and

identifying tangential gradient computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the outside of a segmented nodule of the image.

33. A computer readable medium storing computer program instructions for determining if a pulmonary nodule is malignant, which when used to program a computer to cause the computer to perform the steps of:

obtaining at least one computed tomography medical image of the pulmonary nodule;

obtaining at least one patient feature of a patient having the pulmonary nodule;

extracting image features of the pulmonary nodule from the at least one computed tomography medical image; and

evaluating whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features,

wherein the step of extracting image features comprises at least one of the steps of

identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

identifying image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule, and

the step of identifying image features based on the linear patterns of the at least one computed tomography medical image of the pulmonary nodule comprises

identifying the magnitude of line patterns for inside region of a segmented nodule of the medical image; and

identifying the magnitude of line patterns for outside region of a segmented nodule of the medical image.

34. A computer readable medium storing computer program instructions for determining if a pulmonary nodule is malignant, which when used to program a computer to cause the computer to perform the steps of:

obtaining at least one computed tomography medical image of the pulmonary nodule;

obtaining at least one patient feature of a patient having the pulmonary nodule;

extracting image features of the pulmonary nodule from the at least one computed tomography medical image; and

evaluating whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features,

wherein the step of extracting image features comprises at least one of the steps of

identifying image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

identifying image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

identifying image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule, and

the step of identifying image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule comprises

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the inside of a segmented nodule of the medical image;

identifying radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the outside of a segmented nodule of the medical image;

identifying tangential gradient index computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the inside of a segmented nodule of the medical image; and

identifying tangential gradient computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the outside of a segmented nodule of the image.

35. A system for determining if a pulmonary nodule is malignant, comprising:

a mechanism configured to obtain at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to obtain at least one patient feature of a patient having the pulmonary nodule;

a mechanism configured to extract image features of the pulmonary nodule from the at least one computed tomography medical image; and

a mechanism configured to evaluate whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features,

wherein the mechanism configured to evaluate image features comprises at least one of

a mechanism configured to identify image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

a mechanism configured to identify image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

a mechanism configured identify image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule,

wherein the mechanism configured to identify image features based on the linear patterns of the at least one computed tomography medical image of the pulmonary nodule comprises

a mechanism configured to identify the magnitude of line patterns for inside region of a segmented nodule of the medical image; and

a mechanism configured to identify the magnitude of line patterns for outside region of a segmented nodule of the medical image.

36. A system for determining if a pulmonary nodule is malignant, comprising:

a mechanism configured to obtain at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to obtain at least one patient feature of a patient having the pulmonary nodule;

a mechanism configured to extract image features of the pulmonary nodule from the at least one computed tomography medical image; and

a mechanism configured to evaluate whether the pulmonary nodule is malignant based on an examination of a total of seven of the patient or image features,

wherein the mechanism configured to identify image features comprises at least one of

a mechanism configured to identify image features based on an outline of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on linear patterns of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on gray-level distribution of the at least one computed tomography medical image of the pulmonary nodules;

a mechanism configured to identify image features based on the gray level distribution of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule;

a mechanism configured to identify image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of the at least one computed tomography medical image of the pulmonary nodule; and

a mechanism configured to identify image features based on the relationship between two histograms in the inside and outside regions of the segmented nodule of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule,

wherein the mechanism configured to identify image features based on edge orientation of an edge gradient of the at least one computed tomography medical image of the pulmonary nodule comprises

a mechanism configured to identify radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the inside of a segmented nodule of the medical image;

a mechanism configured to identify radial gradient index computed by the mean absolute value of a radial edge gradient projected along a radial direction for the outside of a segmented nodule of the medical image;

a mechanism configure to identify tangential gradient index computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the inside of a segmented nodule of the medical image; and

a mechanism configured to identify tangential gradient computed by the mean absolute value of a tangential edge gradient projected along a tangential direction for the outside of a segmented nodule of the image.