Part Of The Female Reproductive System Engineering Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Ultrasound imaging is the most widely used method of imaging human abdominal organs like kidney, spleen and liver. Ultrasonic echoes from human tissues displayed as a B-scan image form a texture pattern. One application of diagnostic ultrasound is uterus imaging. Since uterus is a vital organ in women, uterus abnormalities are identified. Because these abnormalities are concentrated in a small area of the uterus tissue, while the rest being normal, finding the abnormality sometimes becomes tough. The ultrasound images however are degraded by speckle noises and other artifacts. So, it is difficult for the automated extraction of tumors.

The uterus is the part of the female reproductive system. It plays a major role. There are a few abnormalities present in the uterus such as tumors. These tumors can be benign or malignant in nature. Example of benign tumors is fibroids. Fibroids are nothing but lumps of muscle tissue that grows in the wall of the uterus. Presence of this can lead to problem to women like heavy menstrual bleeding, painful periods, pain in the lower abdomen etc. The aim of this project is to detect this abnormality and classify them into their respective category.


The main objective of the project is to compare the various classification techniques for identifying uterus abnormalities. The main concept of texture analysis involves feature extraction. The feature vectors are obtained from the Gabor Filter. From the Gabor Features, Statistical Parameters are calculated. The Neural Network is applied to classify. They are compared to evaluate the best performance of classification.








1.Prostate cancer diagnosis based on Gabor filter texture segmentation of ultrasound image

Gabor filter texture segmentation.Using multi channel filtering technique.


the ultrasound images promising results were obtained by using

different Gabor filters at different frequencies.

2. It produced distinct segments inside the prostate.

Care has to

be taken not to over smooth the outputs so that correct

localization of the texture edges can he achieved.

1. It produced distinct segments inside the prostate.

2. Smoothing

resulted in suppressing the variations within the same

texture region in the output image, thus enhancing the

segmentation process significantly

3. Excellent tool that can be used for prostate texture segmentation.

2.Gabor Filter Visualization

Response of Gabor filter to an image.

1.Summarizes the response of the filter in projected dimensions.

2.Visual interpretation of Gabor filters.

1.Designing a 2D Gabor filter bank is difficult filter.

Extraction of spatially localized spectral features.

3.Breast Tissue Classification using Statistical Feature Extraction of Mammogram.

The statistical

features extracted are the mean,standard deviation, smoothness, third moment, uniformity and entropy which signify the

important texture features of breast tissue.Based on the difference in values they classify it.

1.Classification of normal and cancerours breast.

2.Early detection of breast cancer.

1. The selection of ROI for the application of the algorithm.This process is very difficult in this technique.


results show a visual clue to any radiologist about the abnormality

in the growth of breast.

4.Improvement of Ultrasound Image Based on Wavelet Transform:

Speckle Reduction and Edge Enhancement

Here the directional filtering and noise reducing procedures from the coarse

to fine resolution images that are obtained from the wavelet-transformed data.

Speckle reduction is obtained in ultrasound images.

The algorithm reduces speckle well but the image looks artificial due to blurring. And

some edges are over enhanced along the unwanted direction and some are lost.

Edges are enhanced in

terms of both continuity and contrast.

5.Efficient Image Texture Analysis and Classification for

Prostate Ultrasound


Minimum Squared Error

(MSE) clustering algorithm

MSE Classifier Results - While normalization can

reduce any error that may result from an excessively

large parameter biasing the results

1. Normalization can

also cause problems by reducing the separation among

dispersed clusters

2. The mean used is highly dependent on the ROI chosen by the investigator and

could skew the data set leading to misleading or

inaccurate results

Significant reduction in textural analysis processing time

6. Segmenting Tumors in Ultrasound Images

Application of simple Gaussian filter

1.The breast tumor is

segmented from rest of the image using Matlab


2. After segmenting the tumor, the shape

is analyzed to classify whether it is benign or


1. Chances of mis classification are there.

1.Precise early detection of


2.Separated into malignant and benign

7.Use of the Laplacian of Gaussian operator in prostrate Ultrasound Image processing

Laplacian of Gaussian (LoG) operator edge detection scheme

1 The result of applying the LOG operator based edge detection

scheme to an ultrasound image of the prostate is the original image and

the result of applying the simple edge detection scheme,

with the detected edges superimposed on the original image. The detected contours closely match

the boundaries of the regions discernible by the eye, including

not only most parts of the prostate itself, but also

some features inside and outside it

1. There are also some spurious edges, such as those little

irregular circles, which are difficult to justify.

2. Tracing, interpolation and linking

is still required to yield a meaningful outline of the


1. After the edge segments of the prostate are found to be very clear which can give rise to the clear distinction tumor and normal image.

8. Prostate Cancer Detection using Texture and

Clinical Features in Ultrasound Image

Histogram equalization followed by Support

Vector Machine(SVM) classifications.

After initial prostate segmentation, multi-resolution

autocorrelation is extracted to be used as a feature


the clinical features work with texture features, it shows 96% sensitivity and 91.9 specificity. At 92% sensitivity, it

achieves 95.9% specificity.

2. It shows

that if texture features are used with clinical features such

as location and shape of hypo echoic region, it can maintain

high sensitivity with high specificity.

1. Presence of many false positive regions.

1. This method achieved promising results.

2. In spite of some amount of false positives, the

method has good ability to detect cancer region.




Ultrasound has become increasingly important in medicine and has taken its place along with X-ray and nuclear medicine as a diagnostic tool. Its main attraction as an imaging modality lies in its non-invasive character and ability to distinguish interfaces between soft tissues. Diagnostic ultrasound is applied for obtaining images of almost the entire range of internal organs in the abdomen. These include the uterus, kidney, liver, spleen, pancreas, bladder, major blood vessels and other internal organs.


An ultrasonic picture is obtained by a method known as the pulse-echo technique. This employs a high frequency sound generated by the piezo electric transducer. The probe produces a short pulse of sound which will travel into the body, along a pencil beam, in the direction in which the probe is pointing. Each time the pulse , on its passage into the body, encounters a small reflection of the sound is reflected back to the transducer. Here the arrival of the echo causes a tiny voltage to be produced across the piezoelectric element as the echo gives a minute squeeze to the transducer. The voltage is detected electronically and amplified ready for display. The majority of energy in the sound pulse passes through each interface and gives rise to further echoes from deeper in the body. The size, or amplitude, of the voltage produced on the probe as each echo arrives is proportional to the strength of the echo.


  • The echoes are displayed that do not correspond to real structures.
  • Sometimes the genuine echoes are omitted from the display.
  • There is displacement of echoes.
  • The echo characteristics are subjected to severe distortion.


  • Ultrasound scanning is noninvasive (no needles or injections, in most cases) and is usually painless.
  • Ultrasound is widely available and easy to use.
  • Ultrasound imaging uses no ionizing radiation, and is the preferred image modality for diagnosis and monitoring of pregnant women and their unborn infants.
  • Ultrasound provides realtime imaging, making it a good tool for guiding minimally invasive procedures such as needle biopsies.
  • Ultrasound images can visualize structure, movement and live function in the body's organs and blood vessels.



This scan can be done abdominally, transvaginally, or both. The abdominal scan tends to give a larger field of view, but less detail, particularly for structures deep in the pelvis and partially hidden by the pubic symphysis. If scanning abdominally, a full bladder is helpful as sound transmits well through water. In this case, the full bladder serves as an acoustic "window" into the pelvis. The full bladder also helps raise pelvic structures up from behind the symphysis and into view. If scanning transvaginally, a full bladder makes the scan more difficult because it pushes the uterus, tubes and ovaries further away from the vaginal transducer.

Typical diagnostic sonographic scanners operate in the frequency range of 2 to 18 megahertz, hundreds of times greater than the limit of human hearing. The choice of frequency is a trade-off between spatial resolution of the image and imaging depth: lower frequencies produce less resolution but image deeper into the body.


When performing this type of scan, adjusting various settings for the equipment can have a significant effect on improving the images and clarifying detail.

  • Increasing to higher ultrasound frequency will give better resolution, but poorer depth of penetration. In the obese patient, depth of penetration is very important and resolution may need to be sacrificed somewhat in order to see all of the structures.
  • Increasing the gain (amplification) will bring out more echoes on the screen, particularly at the lower end of the image, but increasing the gain results in more artifact. Decreasing the gain will clear up some of the artifact (particularly in cystic masses), but with some loss of signal, particularly deep in the tissues.
  • Focal distances can be varied. Set the focus just below the deepest structure you wish to see clearly.
  • Field of view can be widened or narrowed. The narrower the field of view, generally the better the image quality within the field


The Radiologist or Sonographer examines the patient in two ways

  • Abdominal Ultrasound
  • Transvaginal Ultrasound

Procedure for obtaining the Abdominal Uterus Ultrasound:

If an ultrasound is ordered by the clinician, it is instructed to have a full bladder for the procedure. Air interferes with sound waves, so if the bladder is distended, the air-filledbowelis pushed out of the way by the bladder and an image of theuterus is obtained.

The patient is positioned on an examination table, and a clear gel is applied to the abdomen to help the transducer make secure contact with the skin. The sound waves produced by the transducer cannot penetrate air, so the gel helps to eliminate air pockets between the transducer and the skin. The sonographer or radiologist then presses the transducer firmly against the skin and sweeps it back and forth to image the area of interest.

The captured image is viewed on the monitor and the abnormality if any is marked on the image and is forwarded to the radiologist for confirmation.

Procedure for obtaining the Transvaginal Ultrasound:

An ultrasound technologist will use a transducer, or camera, to take high-resolution pictures of the pelvis. The transducer looks much like a wand and is inserted into the vagina. Ultrasonic, or high- frequency, sound waves are sent out and return to the transducer, much like sonar. Images are created from these "echoes."

The most common reasons for a transvaginal ultrasound is pelvic pain, pain from fibroid tumors, ovarian cysts, endometriosis, pelvic infections, and a myriad of other reasons.




The main aim of the project is to detect the type of abnormalities present in the ultrasound uterus image. The various processes involved are

  • Getting the input image.
  • Feature extraction using Gabor Filter.
  • Processing of Gabor Feature (Statistical parameters).
  • Classification is done using these features.


The images acquired from the scanning center are preprocessed. First, the image is converted to gray scale. Then, for processing of the image it is resized to a standard size of 256 X 256.


The Feature Extraction is done using Gabor Filter. The input image is filtered using Gabor Filter. From each set of Gabor filtered image, the Gabor features are calculated that included mean, varience, standard deviation, kurtosis and other statistical parameters called feature vectors. They are given as an input to the next step in the process.


The various Classification techniques can be used such as Neural Network, Fuzzy Reasoning and Pattern Recognition Methods. For this application we use artificial neural network algorithm, MLP (Multi Layer Perceptron) and SMV (Support vector Machine) for classification.




If the input data to an algorithm is too large to be processed and not much information in the data then the input data will be transformed into a reduced representation set of features (also named features vector). Transforming the input data into the set of features is calledfeature extraction. Feature extraction involves simplifying the amount of resources required to describe a large set of data accurately.

Feature extraction makes the analysis of the particular application easier. There are different types of algorithms based on the application. Features can be extracted based on the

  • Statistical parameters
  • Textural parameters
  • Morphological parameters

Statistical parameters are the features where the statistical features are considered such as the mean, variance, standard deviation, skewness and kurtosis. The textural parameters take the texture of the image into account and the features are extracted. The morphological parameters take the shape, size and position of the required feature and it is extracted.


The region of interest is user defined. The area over which the image has to be further processed is selected by viewing the image. That area is cropped and is convolved with the feature extraction algorithm. For different algorithms the same region is selected manually and the results are tabulated. From the results obtained the images are classified into normal and abnormal category.



AGabor filteris alinear filterused inimage processingfor edge detection. It has been found to be particularly appropriate for texture representation and discrimination. The Gabor Filter, whose kernels are similar to the 2-D receptive field profiles of the mammalian cortical simple cells, has been proven to be able to derive desirable features of spatial frequency, spatial locality and orientation selectivity. The spatial / frequency analysis has played a central role in feature extraction as it combines two fundamental domains and allows the simultaneous representation of signal in both domains. The Gabor filters effectively model the receptive field profiles (or) cortical simple cells in the primary visual cortex.


AGabor filteris a linear function whose impulse response is defined by a harmonic function multiplied by a Gaussian function. [1]


Θ is the special frequency and,

Φ is the orientation

The response of a Gabor filter to an image is obtained by a 2D convolution operation. Let I(x, y) denote the image and G(x, y, θ, φ) denote the response of a Gabor filter with frequency θ and orientation φ to an image at point (x, y) on the image plane. G(.) is obtained as [1]


  • Get the input image I(x, y) of size (P x Q).
  • Find the Gabor function G(x, y).
  • The Gabor filtered image is obtained by convolving I(x, y) and G(x, y).
  • The magnitude is obtained from the set of Gabor filtered coefficients.
  • The statistical parameters are computed.


Given an image I(x,y) of size (P x Q) , its Gabor filter is defined as Gmn (x,y) = I(x,y) * gmn(x,y), where * denotes convolution.

Assume that the local texture regions are spatially homogeneous. Gabor filters are a group of filters. A set of filtered images is obtained by convolving the given image with Gabor filters. Each image represents image information at certain scale and orientation. From each filtered output, the Gabor features are calculated and images are retrieved. Depending on the scale and orientation, the number of Gabor features differs. Irrespective of the size of the image, Gabor features of size (S x N) are obtained.

The input image (P x Q) is passed through a set of Gabor filters to obtain the Gabor filtered output. After applying Gabor filter, different Gabor features are obtained. The magnitude is calculated from the set of Gabor coefficients, which represents the energy content at different scale and orientation of the image.

From the Gabor filtered image various statistical parameters are obtained. For this particular application statistical parameters like Mean, Standard Variation, Variance, Skewness, Kurtosis, Eigen Values, GLCM contrast and Energy is calculated.

The Statistical parameters calculated as a feature of the Gabor filtered output is:

  • Mean: This is defined as the average or mean of the matrix elements. It is given by the formula, [2]
  • Standard deviation: this is the standard deviation of the matrix values. It is given by,[2]
  • Variance: This is ameasureof theaveragedistance between each of a setofdatapointsand their meanvalue. It is also equal to the sum of the squares of thedeviationfrom the mean value.
  • Skewness: Skewness is a measure of the asymmetry of the data around the sample mean. If skewness is negative, the data are spread out more to the left of the mean than to the right. If skewness is positive, the data are spread out more to the right. The skewness of the normal distribution (or any perfectly symmetric distribution) is zero.

The skewness of a distribution is defined as,

  • Kurtosis: Kurtosis is a measure of how outlier-prone a distribution is. The kurtosis of the normal distribution is 3. Distributions that are more outlier-prone than the normal distribution have kurtosis greater than 3; distributions that are less outlier-prone have kurtosis less than 3.

The kurtosis of a distribution is defined as

  • Eigen values: [V,D] = eig(A) produces matrices of eigenvalues (D) and eigenvectors (V) of matrix A, so that A*V = V*D. Matrix D is the canonical form of A — a diagonal matrix with A's eigenvalues on the main diagonal. Matrix V is the modal matrix — its columns are the eigenvectors of A.
  • GLCM Contrast: Measures the local variations in the gray-level co-occurrence matrix.
  • GLCM Energy: Provides the sum of squared elements in the GLCM. Also known as uniformity or the angular second moment.

5.3.5 RESULTS:




Std. Deviation


Eigen Values

GLCM Contrast

GLCM Energy

Normal Images









Fibroid Images









Table 1 Gabor features Range

After applying gabor filter and calculating the statistical parameters, we see that there is a distinct variation between the normal and Fibroid cases.



Wavelet packet decomposition(WPD) is awavelet transform where the image is passed through more filters than thediscrete wavelet transform. DWT is anywavelet transformfor which the waveletsare discretely sampled. It captures both frequencyandlocation information.


The image is decomposed into several resolution levels. First, the original image is decomposed by two complementary half-band filters (high-pass and low-pass filters) that divide a spectrum into high-frequency (detail coefficients; D1) and low-frequency (approximation coefficients; A1) components (bands). For example, the low-pass filter will remove all half-band highest frequencies. Information from only the low frequency band (A1), with a half number of points, will be filtered in the second decomposition level. The A2outcome will be filtered again for further decomposition.

Here when the image is passed to a filter at stage one, it is passed through a high and a low pass filter. The high and the low pass components of the image are separated and they are further introduced to another set of filters at each stage. At each stage we get the horizontal, vertical and diagonal coefficients. This whole process is called decomposition.

Any object present in an image is a high frequency component. For this application, the presence of a fibroid is an object and this is a high frequency component. Therefore we consider only the high frequency components i.e. the detailed coefficients of the image. The detailed horizontal and vertical coefficients are numerical values and the values for normal and abnormal images are tabulated.




Normal Images

-14 to -5

-5 to +0.4

Fibroid Images

-0.5 to 3

-0.4 to 1.5

Table 2 Wavelet coefficients Range

After applying Wavelet Decomposition Technique and calculating the coefficients values, we see that there is a distinct variation between the Normal and Fibroid cases.



A shade of gray assigned to a pixel. The shades are usually positive integer values taken from the gray scale. A gray level image can have a value from 0 to 255. For different medical imaging techniques the image can be a gray image or a color image. For this particular application we use ultrasound uterus images. This is a gray scale image. Each pixel has a particular range between the ranges 0 to 255.


For this particular application, the gray level intensities are measured. The gray levels are different for different type of tissue. This is done for two sizes (256 X 256 and 128 X 128). Based on the values of the gray level it can be identified whether it is normal or abnormal.

The region of interest is the user defined area. From the selected area the gray level intensity is acquired for both normal and fibroid images.


· The ultrasound uterus image is loaded.

· A part of the image is considered. The size of the ROI is either 256 x 256 or 128 X 128.

· For this particular size, the gray level intensity is obtained as a matrix.

· The mean of these matrix values are noted.

· For all different images the intensity values are tabulated and the variation is observed.


From the results obtained, it was observed that there is distinct variation in gray level intensities for both the tissue abnormalities and the window size. From the observation we learn that, there is a definite textural and intensity levels.

5.5.5 RESULTS:

Mean of intensity for 256 by 256

Mean of intensity for 128 by 128







Table 3 Gray level Intensity range



Identification of objects in an image is morphological feature extraction. Sometimes it becomes very difficult to identify this object. The object might be a tumor (benign or malignant), cysts, extra muscle growth etc. The objects present don't have any particular shape, size of position. For different types of medical images different objects are present. Morphological operations are usually used to understand the structure or form of an image. This is usually used to identify objects or boundaries within an image. Morphological operations play a key role in medical imaging.


The ultrasound image is converted to black and white.The eccentricity and circularity of the fibroid are calculated. Then a particular threshold is set for circularity and if it is within the threshold the fibroid is detected and segmented.


The following results are observed. This is a method of visual interpretation. Only fibroids are segmented in this method.




After processing the images, all the values and results need to be classified. When there is large amount of data present, checking if the given image is normal or abnormal individually is a tedious task. In order to make this easier classification algorithms are used.

For this particular application, Artificial Neural Network Algorithms are used. Two types of algorithms are used, namely, MLP (Multi Layer Perceptron) and SVM (Support Vector Machine).



Amultilayer perceptronis afeedforwardartificial neural networkmodel that maps sets of input data onto a set of appropriate output. It is a modification of the standard linearperceptron in that it uses three or more layers of neurons (nodes) with nonlinearactivation functions, and is more powerful than theperceptronin that it can distinguish data that is notlinearly separable. They are supervised networks so they require a desired response to be trained. They learn how to transform input data into a desired response, so they are widely used for pattern classification. With one or two hidden layers, they can approximate virtually any input-output map. They have been shown to approximate the performance of optimal statistical classifiers in difficult problems. Most neural network applications involve MLPs.

Training procedure:

The first step in developing a neural network is to create a database of examples. The examples will contain all the types of data which has to be classified. For this application we are using a multi feature classification. Here, many features are considered to classify the given data into the right category. The neural network used here is fully interconnected architecture. All this data is now fed to the system. This procedure is called training the neural network. By training the system, the system will familiar with the type of data given during the testing process and categorize them properly. The number of data used for this application is 50 (normal and abnormal data together). Features used are mean, variance, standard variation, skewness, kurtosis, eigen value, GLCM contrast and GLCM energy- gabor features, grey level intensity variation, and wavelet decomposition- horizontal and vertical coefficients.

75% correct classification 90% correct classification

Testing procedure:

For testing purpose, a set of data is set aside. After the training procedure, the data is given to the artificial neural network system. Based on the training efficiency, the data given for testing will be classified. After all the classification is done the efficiency of that system is known. For this application, Artificial Neural Network Tool box is used. This makes the implementation of the process easier.


For running the Multilayer perceptron Artificial Neural Network we found that there is 95% efficiency (multiple features) in the system to classify Ultrasound Uterus Images into normal and abnormal categories.



Support vector machines (SVMs)

are a set of related supervised learning methods used for classification and regression. In simple words, given a set of training examples, each marked as belonging to one of two categories, an SVM training algorithm builds a model that predicts whether a new example falls into one category or the other. Intuitively, an SVM model is a representation of the examples as points in space, mapped so that the examples of the separate categories are divided by a clear gap that is as wide as possible.

Training procedure:

The first step in developing a neural network is to create a database of examples. The examples will contain all the types of data which has to be classified. For this application we are using a multi feature classification. Here, many features are considered to classify the given data into the right category. The neural network used here is fully interconnected architecture. All this data is now fed to the system. This procedure is called training the neural network. By training the system, the system will familiar with the type of data given during the testing process and categorize them properly. The number of data used for this application is 50 (normal and abnormal data together). Features used are mean Vs Standard variation and Skewness Vs Kurtosis.

Testing procedure:

For testing purpose, a set of data is set aside. After the training procedure, the data is given to the artificial neural network system. Based on the training efficiency, the data given for testing will be classified. After all the classification is done the efficiency of that system is known. For this application, Artificial Neural Network Tool box is used. This makes the implementation of the process easier.


Percentage Correct classification

Percentage Incorrect classification

Skewness & Kurtosis



Mean & Std. Deviation



Table4 MLP Classification





Skewness & Kurtosis




Mean & Std. Deviation




Table5 SVM Classification


From the above analysis, we can conclude that from Statistical parameters, Wavelet coefficients (horizontal and vertical) and Gray level Intensity variations, we can identify whether the given image is normal or abnormal. The classification efficiency for MLP (skewness Vs kurtosis) was 75% and MLP (mean Vs Standard deviation) 90%. The same features were classified using SVM and better accuracy of 91.67% and 95.83% respectively were obtained.