Optimized stochastic resonance signal detection method

• US 8,214,177 B2
• Filed: 02/22/2010
• Issued: 07/03/2012
• Est. Priority Date: 10/20/2005
• Status: Active Grant

• Current Assignee: Syracuse University
• Original Assignee: Syracuse University
• Inventors: Peng, Renbin, Chen, Hao, Varshney, Pramod K.
• Primary Examiner(s): Huynh, Phuong
• Application Number:
  US12/710,143
• Publication Number:
  US 20100169051A1
• Time in Patent Office: 862 Days
• Field of Search: 702 19- 22, 702179-189, 702/191, 704/226, 375/227
• US Class Current: 702/191
• CPC Class Codes: A61B 6/502 for diagnosis of breast, i.e. mammography
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/30068 Mammography; Breast
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: H04B 17/26 using historical data, averaging values or statistics

Abstract

Apparatus and method for detecting micro-calcifications in mammograms using novel algorithms and stochastic resonance noise is provided, where a suitable dose of noise is added to the abnormal mammograms such that the performance of a suboptimal lesion detector is improved without altering the detector'"'"'s parameters. A stochastic resonance noise-based detection approach is presented to improve suboptimal detectors which suffer from model mismatch due to the Gaussian assumption. Furthermore, a stochastic resonance noise-based detection enhancement framework is presented to deal with more general model mismatch cases.

Claims

1. A method of increasing the probability of detecting at least one micro-calcification lesion in a mammogram, said method comprising the steps of:

a. obtaining digital mammogram data by using an X-ray digital mammography system;

b. calculating stochastic resonance noise with a processor by determining the stochastic resonance noise probability density function that does not increase the probability of false alarm and calculating the stochastic resonance noise data probability density function from a known probability density function for said digital mammogram data, wherein the stochastic resonance noise probability density function equals p_n^opt(n)=λ

δ

(n−

n₁)+(1−

λ

)δ

(n−

n₂), with values n₁ and n₂ equal to two delta function locations having probabilities of λ and

1−

λ

respectively, where n₁ and n₂ denote two discrete vectors, λ

denotes the probability of occurrence of said two vectors, and n is a randomization of said two discrete vectors added with the probabilities of λ and

1−

λ

, respectively; and

c. adding said stochastic resonance noise to said digital mammogram data by using said processor, thereby improving the detection of said at least one micro-calcification lesion in the mammogram.

2. The method of claim 1, wherein the step of obtaining further comprises the step of performing coarse detection of said at least one micro-calcification lesion in the mammogram using a local maxima filter.

3. The method of claim 2, further comprising the step of calculating the detection threshold for a Gaussian background assumption-based detector.

4. The method of claim 3, further comprising the step of estimating mean μ

_b and variance σ

_b² of the background based on detected negative pixels.

5. The method of claim 4, updating the detection threshold for the Gaussian background assumption-based detector.

6. The method of claim 5, wherein the step of updating the detection threshold is based on the desired P_F, mean μ

_b and variance σ

_b² using the following equation;

7. The method of claim 5, further comprising the step of detecting positive and negative pixels with said updated detection threshold.

8. The method of claim 7, further comprising the step of estimating probability densities using kernel density estimation method based on the resulting detected positive and negative pixels under the following two hypotheses:

H₀;

y[m]=w[m]

H₁;

y[m]=s[m]+w[m]

where m is a pixel index corresponding to a pixel observation under consideration, y[m] is intensity of said pixel observation under consideration, larger than or equal to zero, s[m] is lesion signal, and w[m] is background noise that is assumed to obey Gaussian distribution with mean μ

_b and variance σ

_b².

9. The method of claim 8, wherein the step of calculating stochastic resonance noise further comprises the step of using the updated detection threshold and the estimated probability densities to calculate the stochastic resonance noise.

10. The method of claim 9, further comprising the step of performing the step of detection on the stochastic resonance noise modified digital mammogram data.

11. The method of claim 10, further comprising the step of determining the difference between the course detection results and the results obtained from the detection performed on the stochastic resonance noise modified digital mammogram data.

12. The method of claim 11, repeating the following steps in an iterative manner:

estimating mean μ

_b and variance σ

_b² the background based on detected negative pixels;

updating the detection threshold for the Gaussian background assumption-based detector;

detecting positive and negative pixels with said updated detection threshold;

estimating probability densities;

calculating stochastic resonance noise;

modifying said digital mammogram data by adding said stochastic resonance noise to said digital mammogram data; and

performing the step of detection on the stochastic resonance noise modified digital mammogram data.

13. The method of claim 1, further comprising the step of introducing memory into said method by applying multiple two-peak stochastic resonance noises multiple times.

14. The method of claim 13, wherein the step of introducing memory into said method by applying multiple two-peak stochastic resonance noises multiple times further comprises the step of using the following multiple two-peak stochastic resonance noise scheme:

z= y+ n*,

where n* represents multiple-peak stochastic resonance noises added to the original mammogram data, where z denotes the original mammogram data modified through adding the stochastic resonance noise to the original mammogram data and y denotes the original mammogram data, and