Computer aided diagnosis from multiple energy images

• US 20030215119A1
• Filed: 05/15/2002
• Published: 11/20/2003
• Est. Priority Date: 05/15/2002
• Status: Active Grant

• Current Assignee: GE Medical Systems Global Technology Company LLC (GE Healthcare Technologies, Inc.)
• Original Assignee: GE Medical Systems Global Technology Company LLC (GE Healthcare Technologies, Inc.)
• Inventors: Jabri, Kadri Nizar, Sabol, John Michael, Avinash, Gopal B., Uppaluri, Renuka, Rader, Amber Elaine, Thomas, Carson Hale
• Granted Patent: US 7,263,214 B2
• Time in Patent Office: Days
• US Class Current: 382/128
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G16H 30/20 for handling medical images, e.g. DICOM, HL7 or PACS

Abstract

A method, system, and storage medium for computer aided processing of dual or multiple energy images includes employing a data source, the data source including a dual or multiple energy image set, defining a region of interest within one or more images from the dual or multiple energy image set, extracting a set of feature measures from the region of interest, and employing a feature extraction algorithm on the feature measures for identifying an optimal set of features. The method may be employed for identifying bone fractures, disease, obstruction, or any other medical condition.

Claims

1. A method for computer aided processing of dual or multiple energy images, the method comprising:

employing a data source, the data source including a dual or multiple energy image set;

defining a region of interest within an image from the dual or multiple energy image set;

extracting a set of feature measures from the region of interest; and

,

reporting the feature measures on the region of interest.

2. The method of claim 1 further comprising employing a feature selection algorithm on the region of interest and classifying the region of interest.

3. The method of claim 2 further comprising incorporating prior knowledge from training for classifying the region of interest.

4. The method of claim 3 wherein incorporating prior knowledge from training includes computing features on known samples of different normal and pathological medical conditions.

5. The method of claim 4 wherein the feature selection algorithm sorts through the features of known samples, selects useful features of known samples, and discards features of known samples which do not provide useful information.

6. The method of claim 2 wherein classifying the region of interest using the optimal set of features comprises classifying one or more medical conditions.

7. The method of claim 1 wherein processing dual or multiple energy images comprises detecting and diagnosing one or more medical conditions.

8. The method of claim 1 wherein defining a region of interest comprises manually selecting a region of interest.

9. The method of claim 1 wherein defining a region of interest comprises utilizing an automated algorithm with or without user specifications input.

10. The method of claim 1 wherein the data source further includes at least one of image acquisition system information and demographic information, symptoms, and history of patient, wherein the image acquisition system information, demographic information, symptoms, and history of patient serve as feature measures in the feature extraction algorithm.

11. The method of claim 4 wherein the feature selection algorithm comprises determining a feature measure“

s ability to separate regions into different classification groups.

12. The method of claim 11 wherein the feature selection algorithm further comprises ranking each feature measure in the set of feature measures based on each feature measure“

s ability to separate regions into a classification group.

13. The method of claim 12 wherein the feature selection algorithm further comprises reducing quantity of feature measures by eliminating correlated features.

14. The method of claim 12 wherein the feature selection algorithm further comprises selecting highest ranked feature measure and adding additional feature measures in descending order.

15. The method of claim 1 wherein the dual or multiple energy image set is a dual energy image set comprising a high energy image, a low energy image, a bone image, and a soft tissue image, the method comprising defining regions of interest and incorporating features from all regions of interest on all images.

16. The method of claim 1 wherein the dual or multiple energy image set is a dual energy image set comprising a high energy image, a low energy image, a bone image, and a soft tissue image, the method comprising defining at least one region of interest, employing a feature extraction algorithm, and classifying a candidate region of interest on each image and subsequently combining results of all operations.

17. The method of claim 1 further comprising indicating at least one classified region using a marker on a display of each image within the dual or multiple energy image set where the at least one classified region is located.

18. The method of claim 17 further comprising displaying a single image which incorporates all markers from each image within the dual or multiple energy image set.

19. A system for computer aided processing of dual energy images, the system comprising:

a detector generating a first image representative of photons at a first energy level passing through a structure and a second image representative of photons at a second energy level passing through the structure;

a memory coupled to the detector, the memory storing the first image and the second image;

a processing circuit coupled to the memory, the processing circuit processing a dual energy image set including a bone image, a soft tissue image, a high energy image, and a low energy image from the first image and the second image;

storing the dual energy image set in the memory as a data source;

defining a region of interest within an image from the dual energy image set;

extracting a set of feature measures from the region of interest; and

,

a reporting device coupled to the processing circuit, the reporting device reporting at least one feature.

20. A system for computer aided processing of dual energy images, the system comprising:

detection means for generating a first image representative of photons at a first energy level passing through a structure and a second image representative of photons at a second energy level passing through the structure;

storage means for storing the first image and the second image;

processing means for;

processing a dual energy image set including a bone image, a soft tissue image, a high energy image, and a low energy image from the first image and the second image;

storing the dual energy image set in the memory as a data source;

defining a region of interest within an image from the dual energy image set;

extracting a set of feature measures from the region of interest;

employing a feature selection algorithm on the set of feature measures and identifying an optimal set of features;

classifying the optimal set of features; and

,

incorporating prior knowledge from training into classifying the optimal set of features; and

,

display means for displaying at least one classified region of interest.

21. A storage medium encoded with a machine readable computer program code, said code including instructions for causing a computer to implement a method for aiding in processing of dual or multiple energy images, the method comprising:

employing a data source, the data source including a dual or multiple energy image set;

defining a region of interest within an image from the dual or multiple energy image set;

extracting a set of feature measures from the region of interest; and

,

employing a feature extraction algorithm on the feature measures for identifying an optimal set of features.

22. A method for detecting bone fractures, calcifications and metastases, the method comprising:

utilizing a bone image from a dual or multiple energy image set;

selecting a region of interest within the bone image to search for a calcification, fracture or metastatic bone lesion;

segmenting bone from a background of the bone image; and

,

identifying a candidate region within the bone as a candidate for a calcification, fracture or metastatic bone lesion.

23. The method of claim 22 further comprising classifying an identified candidate region.

24. The method of claim 22 wherein segmenting bone comprises utilizing a region growing algorithm.

25. The method of claim 24 wherein the region growing algorithm is manually initialized by having a user select a seed point.

26. The method of claim 24 wherein the region growing algorithm is automatically initialized by utilizing bone attributes to select a seed point.

27. The method of claim 22 wherein segmenting bone comprises multi-level intensity thresholding.

28. The method of claim 22 wherein identifying a candidate region comprises utilizing an edge detection algorithm.

29. The method of claim 28 wherein image processing using morphological erosion is used for eliminating noise and false edges.

30. The method of claim 28 wherein rib edges are eliminated using a connectivity algorithm.

31. The method of claim 22 wherein classifying identified candidate regions comprises using a computer aided rule based approach, wherein different rules apply for calcifications, metastases and fractures, and for different types of fractures and different properties of metastases.

32. The method of claim 31 wherein rules are based on size measurements of line edges of the identified candidate regions.

33. The method of claim 22 further comprising indicating candidate regions on a display.

34. The method of claim 33 wherein indicating candidate regions comprises placing a marker on the bone image indicative of a classification of the candidate region.

35. A method for detecting lung disease, the method comprising:

utilizing a soft-tissue image from a dual or multiple energy image set;

selecting a region of interest within the soft-tissue image to search for an indication of disease;

segmenting the region of interest from a background of the soft-tissue image;

employing a feature selection algorithm on feature measures for identifying an optimal set of features;

identifying a candidate region within the bone image which correlates to the region of interest in the soft-tissue image;

extracting features from the candidate region in the bone image; and

,

classifying the region of interest in the soft-tissue image as a candidate for soft-tissue disease utilizing the features extracted from the bone image.

36. The method of claim 35, wherein identifying a disease in the soft-tissue image comprises identifying a solitary pulmonary nodule or lesion, and wherein the features extracted from the bone-image are indicative of calcification of the nodule, the method further comprising utilizing the bone-image calcification features to classify the region of interest in the soft-tissue image as probably benign.

37. The method of claim 35 wherein classifying comprises using a computer aided rule based approach, wherein different rules apply for different medical conditions, and different rules are used for the soft-tissue and bone-images.

38. The method of claim 35 further comprising displaying a single image which incorporates all markers from each image within the dual or multiple energy image set.

39. The method of claim 38 further comprising displaying a single image which incorporates markers uniquely indicative of results from the soft-tissue image that have been further classified based on results from the bone-image.