Advanced computer-aided diagnosis of lung nodules

Patent metadata

Field	Value
Publication number	US-10121243-B2
Application number	US-44195007-A
Country	US
Kind code	B2
Filing date	Sep 18, 2007
Priority date	Sep 22, 2006
Publication date	Nov 6, 2018
Grant date	Nov 6, 2018

Abstract

Methods are herein provided for decision support in diagnosis of a disease in a subject, and for extracting features from a multi-slice data set. Systems for computer-aided diagnosis are provided. The systems take as input a plurality of medical data and produces as output a diagnosis based upon this data. The inputs may consist of a combination of image data and clinical data. Diagnosis is performed through feature selection and the use of one or more classifier algorithms.

First claim

What is claimed is:

1. A method of providing decision support in diagnosis of a disease in a subject, the method comprising:
receiving a multi-slice computed tomography dataset;
extracting, with a hardware processor, a volume of interest from the multi-slice computed tomography dataset;
segmenting, with the hardware processor, the multi-slice computed tomography dataset to delineate at least one lesion within the volume of interest from a background of the multi-slice computed tomography dataset;
extracting, with the hardware processor, image-based features from the at least one lesion in two and a half or pseudo-three dimensions over a set of two or more slices of the multi-slice computed tomography dataset, wherein the image-based features include a numerical feature;
converting, with the hardware processor, clinical information into multiple categories and from a categorical format into a numerical format of the image-based features and;

combining the converted clinical information with the image-based features, wherein the combined converted clinical information and the image-based features form a feature pool; and
selecting, with the hardware processor, an optimal feature subset from the feature pool, wherein the selecting including selecting both image-based features and converted clinical information based on a training dataset;

creating, with the hardware processor, a classifier or a committee of classifiers using the selected optimal feature subset and training dataset; and
employing the classifier or committee to provide a prediction of a diagnosis of the at least one lesion.

2. The method according to claim 1 , wherein the numerical feature represents one from a group comprising a coefficient, a moment, a gray scale value, a derivative of an image intensity, and an image intensity.

3. The method according to claim 2 , wherein the clinical information includes at least two categories, male and female, and each of these categories is converted into the numerical format of the image-based features.

4. The method according to claim 1 , wherein selecting an optimal feature subset further comprises using at least one of a correlation filter, a recursive feature elimination, and a random feature selection.

5. The method according to claim 4 , wherein selecting an optimal feature subset comprises executing a genetic algorithm-based feature selection for a plurality of iterations with a different randomly selected set of training data and testing data to obtain resulting plurality of feature subsets.

6. The method according to claim 1 , wherein the image-based features include the fractal dimension of the shape of the at least one lesion, an intensity and a size of dark internal pixel clusters in the at least one lesion, an intensity and a size of bright internal pixel clusters in the at least one lesion, chain code measurements of the at least one lesion, and at least one neighborhood gray-tone difference matrix of the least one lesion.

7. The method according to claim 1 , wherein the classifier or committee of classifiers is a linear discriminant analysis.

8. The method according to claim 1 , further comprising the classifier or committee of classifiers determining whether the at least one lesion is malignant or benign.

9. The method according to claim 1 ., further comprising the classifier or committee of classifiers determining a likelihood of malignancy of the at least one lesion.

10. The method according to claim 8 , where the determination by the classifier or committee of classifiers is determined by at least one calculation selected from the group consisting of: a simple mean value, a simple vote, a weighted mean value, and a weighted vote.

11. The method according to claim 1 , wherein the clinical information comprises a size of lymph nodes.

12. The method according to claim 1 , wherein the clinical information comprises a number of lesions around the at least one lesion.

13. The method according to claim 1 , wherein the clinical information comprises a location of the at least one lesion.

14. The method according to claim 1 , wherein the clinical information comprises a presence of nodes.

15. The method according to claim 1 , wherein the clinical information comprises a gender.

16. The method according to claim 1 , wherein the clinical information comprises a past pulmonary disease.

17. The method according to claim 1 , wherein the clinical information comprises a present pulmonary disease.

18. The method according to clan 1 , wherein the clinical information comprises a prior chest surgery.

19. The method according to claim 1 , wherein the clinical information comprises an occupational exposure.

20. The method according to claim 1 , wherein the clinical information comprises a recreational exposure.

21. A computer-aided diagnosis system, the system comprising:
a processor configured to:
receive a multi-slice computed tomography dataset;
extract an image-based region of interest from the multi-slice computed tomography dataset;
delineate at least one lesion in the region of interest;
extract image-based features from the at least one lesion in two and a half or pseudo-three dimensions over a set of two or more slices of the multi-slice computed tomography dataset from the region of interest, wherein the image-based features include a numerical feature;
accept and convert clinical information from a categorical format into multiple categories and into a numerical format of the numerical feature and form a feature pool in which the clinical information and the image-based features constitute the feature pool;
select an optimal feature subset from the feature pool and containing both image-based features and converted clinical information based on a training dataset;

create a classifier or a committee of classifiers using the selected optimal feature subset and training dataset; and
employ the classifier or committee to provide a prediction of a diagnosis of the at least one lesion.

22. The computer-aided diagnosis system according to claim 21 , wherein the processor further selects, during training of the system, relevant image-based features and clinical features from a feature pool.

23. The computer-aided diagnosis system according to claim 22 , wherein the processor utilizes an optimization technique selected from the group consisting of at least one of a correlation filter, a recursive feature elimination, and a random feature selection; and a classifier or committee of classifiers selected from the group consisting of at least one of a support vector machine, a decision tree, linear discriminant analysis, and a neural network.

24. The computer-aided diagnosis system according to claim 21 , wherein the image-based region of interest is constructed by performing at least one morphological operation.

25. The computer-aided diagnosis system according to claim 24 , wherein the at least one morphological operation is selected from the group of erosion, and dilation.

26. The computer-aided diagnosis system according to claim 21 , wherein the image-based region of interest is constructed by selecting the largest contiguous object resulting from the segmentation.

27. The computer-aided diagnosis system according to claim 21 , wherein the image-based region of interest further comprises an internal region, an external region, and a boundary region wherei

Assignees

• Boroczky Lilla
• Agnihotri Lalitha
• Zhao Luyin
• LEE Michael Chun Chieh
• Powell Charles Andrew
• Borczuk Alain C
• Kawut Steven
• Koninklijke Philips Nv

Inventors

• Boroczky Lilla
• Agnihotri Lalitha
• Zhao Luyin
• Lee Michael Chun-Chieh
• Powell Charles Andrew
• Borczuk Alain C
• Kawut Steven

Classifications

• A61B6/5217

  extracting a diagnostic or physiological parameter from medical diagnostic data · CPC title

• A61B6/5223

  generating planar views from image data, e.g. extracting a coronal view from a 3D image · CPC title

• G06T7/48

  using fractals · CPC title

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title