CN111081380B - Method for optimizing microwave breast image based on image quality index and simulated annealing - Google Patents

Abstract

The invention relates to a method for optimizing a microwave breast image based on image quality indexes and simulated annealing, which comprises the following steps of: stretching each breast MRI source image, discretizing each breast tissue, adding a layer of skin by taking the boundary contour of the breast as a reference, and establishing a three-dimensional breast model by interpolation; setting a tumor position and a tumor radius in a three-dimensional breast model, arranging antenna arrays on the surface of skin to replace point sources, sequentially transmitting signals by each antenna, receiving the signals by other antennas, and carrying out imaging processing on all received signals by using a confocal algorithm; constructing an improved focus quality metric index; determining an objective function, namely an image signal-to-clutter ratio SCR; solving the maximum value of the microwave breast image SCR by using a simulated annealing algorithm to obtain the corresponding average dielectric constant of the breast; determining an optimal value of the model average dielectric constant; and (6) imaging.

Description

Method for optimizing microwave breast image based on image quality index and simulated annealing

Technical Field

The invention belongs to the technical field of biomedical detection, and relates to a method for optimizing a microwave breast image based on an image quality metric index and a simulated annealing algorithm.

Background

Mammary gland tumor is the malignant tumor disease with the highest incidence rate in women, and the mortality rate is the first of the mortality rates of the malignant tumors in women. The diagnosis of early breast tumors is of decisive significance for improving the treatment rate of breast diseases and the long-term survival rate of patients. The conventional detection methods for early breast cancer include mammography, ultrasonic imaging technology, computed tomography, magnetic resonance imaging technology, thermal imaging detection and the like, but all of the methods have certain disadvantages, such as radiation damage to human body, low imaging contrast, high cost and the like. The principle of detecting breast cancer by using ultra-wideband electromagnetic waves lies in that different biological tissues have different absorption, reflection and transmission characteristics of the electromagnetic waves, so that the electromagnetic field generated when pulse signals transmitted by an antenna are transmitted in breast tissues can reflect rich information of malignant tissues. Confocal microwave imaging for breast cancer detection relies on accurate knowledge of the average dielectric properties of a particular breast of a patient. After the average dielectric property is accurately estimated, the microwave signals are subjected to coherent superposition at the tumor part to generate a clear microwave image. Conversely, if the average dielectric property estimate is inaccurate, a blurred, unfocused image may be reconstructed, possibly masking cancerous lesions.

Core indexes for evaluating the quality of a microwave image in a photoelectric imaging system include signal-to-noise ratio, contrast and the like. The signal-to-noise ratio is provided on the basis of radar detection, and is calculated by integrating a series of factors in the process of reaching a detector from the background characteristic of a target through a transmission environment and the like. A Focus Quality Metric (FQMs) is an image quality metric for estimating the degree of focus of an entire image. The simulated annealing method is also called a simulated cooling method, a statistical cooling method, a probability hill climbing method and the like. The simulated annealing method is a statistical optimization method, and the algorithm is widely applied to the fields of design of a giant computer system, optimization of a large-scale integrated circuit, image processing, biology, molecular physics and chemistry, numerical analysis, complex layout and the like.

The invention utilizes an improved focus quality metric and an image Signal-to-clutter ratio (SCR) to accurately estimate the average dielectric characteristic of microwave breast imaging, and uses a simulated annealing algorithm to accelerate the estimation speed and reduce the imaging time.

Disclosure of Invention

The invention aims to provide a method for optimizing a microwave breast image based on an image quality metric index and a simulated annealing algorithm. The technical scheme of the invention is as follows:

a method for optimizing a microwave breast image based on image quality metrics and simulated annealing algorithms includes the following steps:

(1) Stretching each breast MRI source image, discretizing each breast tissue, adding a layer of skin by taking the boundary contour of the breast as a reference, and establishing a three-dimensional breast model by interpolation;

(2) Setting a tumor position and a tumor radius in a three-dimensional breast model, arranging antenna arrays on the surface of the skin to replace point sources, sequentially transmitting signals by each antenna, receiving the signals by other antennas, and performing imaging processing on all received signals by using a confocal algorithm;

(3) Constructing an improved focus quality metric phi _MF ：

I'＝I*M (1)

Where I' is the image after linear convolution, M is the linear convolution kernel, X, Y are the image size, F _x,y (m, n) is the value at (m, n) of an 8 × 8 Discrete Cosine Transform (DCT) sub-block centered at (x, y).

(4) Determining an objective function, namely an image signal-to-clutter ratio SCR:

wherein alpha is _T Is the maximum energy of the tumor region, alpha _B Is the maximum of energy except in the tumor area.

(5) Setting the dielectric constant epsilon of the breast _r Obtaining an initial solution of the target function SCR;

(6) Method for solving microwave breast image SCR maximum value by using simulated annealing algorithm _max To obtain the maximum value SCR _max Corresponding mean dielectric constant of breast

(7) Determining a dielectric constant interval of

Calculate phi _MF To obtain phi _MF -ε _r A curve;

(8) Selecting phi _MF -ε _r The curve is minimized and

if the minimum value is the minimum value, the minimum value is the optimal value of the mean dielectric constant of the breast model;

(9) And (3) simulating and imaging the tumor by using a confocal imaging algorithm by using the optimal value of the mean dielectric constant of the breast model.

Drawings

FIG. 1 algorithm flow chart

FIG. 2 breast model diagram

FIG. 3 model antenna position diagram

FIG. 4 simulated annealing iteration curve

In the interval of FIG. 5 _r -φ _MF Curve

Figure 6 tumor imaging

Detailed Description

The invention is described below with reference to the figures and examples.

(1) FIG. 1 is a flow chart of a method of the present invention, which is combined with the flow chart to perform the following steps;

(2) Fig. 2 is a three-dimensional breast model, first, an MRI source image is stretched once to 600 × 600 pixels to make each pixel point in the image correspond to an FDTD grid, the step is performed in MATLAB, first, a source image only including the image itself is read in, converted into a gray value matrix, discretized for each tissue of the breast in the source image, and a layer of skin is added and interpolated with the breast boundary contour as a reference to obtain a three-dimensional breast simulation model;

(3) Setting tumor positions (46.5 mm,56.5mm and 100mm) and tumor radius of 3mm in the model, arranging antenna arrays on the surface of skin, replacing the antenna arrays with point sources (as shown in figure 3), sequentially transmitting signals by each antenna, receiving the signals by other antennas, and imaging all the received signals by using a confocal imaging algorithm;

(4) Set epsilon _r Is 5,6,8, 10, 13, 15, 17, 19, 20, 23;

(5) Solving a microwave breast image SCR maximum (SCR) according to the flowchart of FIG. 1 _max ) To obtain an SCR _max Mean dielectric constant of breast corresponding to maximum value

(6) The dielectric constant interval is determined to be [5.3,6.3]Calculating phi of microwave breast image in interval _MF To obtain epsilon _r -φ _MF Curves and normalization processing (as shown in fig. 5);

(7) Observe FIG. 5, phi _MF Is corresponding to the minimum value of _r =5.9, the mean dielectric constant of this breast model is 5.9.

(8) Selecting epsilon _r =5.9, the tumor was simulated by confocal imaging, the imaging position of the tumor was (45mm, 53mm, 98mm), which is similar to the preset tumor position, so the tumor imaging result of the breast model is shown in fig. 6.

Claims (1)

1. A method for optimizing a microwave breast image based on image quality metrics and simulated annealing, comprising the steps of:

(1) Stretching each breast MRI source image, discretizing each tissue of the breast, adding a layer of skin by taking the boundary contour of the breast as a reference, and establishing a three-dimensional breast model by interpolation processing;

(2) Setting a tumor position and a tumor radius in a three-dimensional breast model, arranging antenna arrays on the surface of skin to replace point sources, sequentially transmitting signals by each antenna, receiving the signals by other antennas, and carrying out imaging processing on all received signals by using a confocal algorithm;

(3) Constructing an improved focus quality metric phi _MF ：

I'＝I*M (1)

Where I' is the image after linear convolution, M is the linear convolution kernel, X, Y are the image size, F _x,y (m, n) is the value at (m, n) of an 8 × 8 Discrete Cosine Transform (DCT) sub-block centered at (x, y);

(4) Determining an objective function, namely an image signal-to-clutter ratio SCR:

wherein, the first and the second end of the pipe are connected with each other,α _T is the maximum energy of the tumor region, alpha _B Is the maximum of energy except for the tumor region;

(5) Setting the dielectric constant epsilon of the breast _r Obtaining an initial solution of the target function SCR;

(6) Method for solving microwave breast image SCR maximum value by using simulated annealing algorithm _max To obtain the maximum value SCR _max Corresponding mean dielectric constant of breast

(7) Determining a dielectric constant interval of

Calculate phi _MF To obtain phi _MF -ε _r A curve;

(8) Selecting phi _MF -ε _r The curve is minimized and

if the minimum value is the minimum value, the minimum value is the optimal value of the model average dielectric constant;

(9) And (3) simulating and imaging the tumor by using a confocal imaging algorithm by using the optimal value of the mean dielectric constant of the breast model.

Images