Automated Gestational Age Estimation For Monitoring Fetal Growth

Abstract' Gestation is defined as a period of time between conception and birth. Estimation of gestational age is necessary in order to predict the early date of delivery and monitor the growth of fetus throughout the three trimester of pregnancy. Assessment of gestational age is based on measurement of various fetal biometric parameters like gestational sac, biparietal diameter, femur length, abdominal circumference, head circumference during the gestation period.
In medical image processing, ultrasound technique plays an important role for imaging organs for an obstetrician and gynecologist. Monitoring of these parameters is done with human interaction. These methods are responsible for multiple subjective decisions which increase the inter-observer error. The main objective of this work is to measure fetal biometric parameter for accurate estimation of gestational age. An automated computer based algorithm has been used to apply morphological operation in order to recognize the desired parameter contour in the ultrasound image, refine its shape and compensate for distinct irregularities, then correctly measure its length, attaining optimum accuracy and reproducibility of measurements. Automation algorithm utilizes morphological operation, Hough transform and tracing methods. It has been found that, the proposed scheme, is able to estimate the gestational age of the fetus with a prediction accuracy of ??2 days.
Keywords-Hough transform analysis, morphological operation, Gestational Sac, Biparietal diameter, Head Circumference, Femur length, Filtration
The obstetric care has been enhanced with the development in the field of computer technology in the recent past. Accurate estimation of gestational age is desirable for monitoring and accessing the fetal growth. It also confirms well being of the pregnancy particularly in patient with the history of bleeding/pain, particularly in high risk pregnancies [1]. These estimations have also been expected to provide valuable information to take decisions in three trimesters of pregnancy. Fetal growth assessment by ultrasound evaluation depends on accurate estimation of Yolk Sac (YS), Gestational SAC (GS), Crown Rump Length (CRL), Femur Length (FL), Head Circumference (HC), Abdominal Circumference (AC) and Biparietal Diameter (BPD). In first trimester assessment of Gestational sac, Yolk sac, Crown rump length plays an important role in predicting the gestational age [1]. In second and third trimester extraction of Femur length, Abdominal Circumference, Biparietal diameter, Head circumference of fetus is done to predict the gestation period accurately. In implemented scheme extracted parameters include GS, FL, BPD, HC which can be used for the development of Automated Clinical Decision Support System (ACDSS) in obstetrics and gynecology.

Assessment of gestational age is based on patient historical information and the physical examination, maternal sensation of fetal movement [5]. With the advent of ultrasound, obstetrics examination has been made easier and hence non invasive technique has been used for extraction of fetal biometric parameter. For assessment of various parameters, gynecologist first freeze the ultrasound image of desired biometric parameter, consequently, selecting two points on the boundaries of parameter by using joy sticks or light pens to measure its length. As a result output in terms of length of parameter is displayed. This method involves multiple subjective decisions increasing the inter-observer error. Because of tedious and time consuming nature of manual measurement an automated method is necessary which aims to locate the contour segment of desired parameter accurately.
Ultrasound images are the result of reflection, refraction and deflection of ultrasound beams from various types of tissue with distinct acoustic impendence. Hence these images are characterized by several types of perturbations: removal of real structural echoes, displacement and distortion of echoes [8]. Moreover,
echography consists of strong presence of speckle and additive noise. It also include presence of other highly echogenic adjacent to the head contour and non uniform bone texture. In addition, the acoustic beam deflections at the bone surface causes a change in the wave propagation direction, as a result weak echoes are detected by transducer at the assumed angle of reflection. These entire factors make the analysis of ultrasound images more difficult. Hence automatic segmentation of these images is necessary.
The process flow chart implemented to extract desired parameter for assessment of gestational age is detailed in figure 1. In the implemented scheme, contrast of an image is enhanced to modify the intensity values of pixels. The fetal images are processed by an improved Gaussian filter followed by wiener filter to remove additive and speckle noise present in ultrasound image and the desired parameter are segmented using global thresholding technique. Results obtained have shown that this segmentation scheme is capable of segmenting the features effectively and locate and preserve the edges for further processing As a result, due to high intensity, image having large number of gaps is formed. Morphological reconstruction is used to minimize the false region. Desired parameter boundary is detected with morphological operation in case of BPD and GS, while detection of femur length is done based on Hough transform. Extraction of desired fetal biometric parameter estimates the gestation age with optimum accuracy.
Figure 1: Process Flow Chart for Gestational Age Estimation
A. Preprocessing
As ultrasound image is result of reflection, refraction and deflection of ultrasound beam from tissue interface of human body, the contrast of ultrasound image is usually low. Hence in preprocessing stage, contrast enhancement technique is used. Enhancement is the process of manipulating an image so that the result is more suitable than the original for specific application [9]. Contrast of an image is enhanced to improve the interpretability and to modify attributes of an image. The contrast can be limited in order to avoid amplifying noise present in the image. The contrast enhanced image is further filtered to suppress the speckle noise. An improved Gaussian filter reported by the authors in their earlier work has been used [4]. This despeckling removes the speckle and preserves the details at the edges. Further wiener filter is applied to remove the additive noise from the image. Combination of Gaussian filter followed by Wiener filter gives more uniform results in further steps. The despeckled images are segmented using global thresholding method, presented in [3].
B. Segmentation
The global thresholding method reported in the literature, segments the features that lie close to each other based on intensity histogram of an image. This method is applicable over the entire range when intensity and background pixels are sufficiently distinct. The objective of thresholding is to minimize the average error incurred in assigning pixels to two or more classes. The threshold giving the best separation in the classes in terms of their intensity values considered as a optimum threshold. The method is optimum in the sense that it maximizes between class-variance, a well known measure used in statistical discriminant analysis.
Binary Image obtained after thresholding shows desired fetal biometric parameter and also large number of false regions. Morphological operation is aimed at locating the contour segment of parameter region while discarding the other foreground structures. It is described as a tool for extracting image components that are useful in the representation and description of region shape such as boundaries, skeleton and the convex hull [3]. The boundary of set A, denoted by B(A) can be obtained by first eroding A by B and then performing the set difference between A and its erosion, Where, B is structuring element, set A is original image.
C. Extraction of Gestational Sac and Biparietal Diameter
Number of false edges is located in segmented image. Removal of false edges is based on knowledge based filtering[7]. Gestational sac and biparietal diameter are circular in shape. Generally ultrasound image is freezed
when above mentioned parameter is at central part. Hence false edges that lies near the boundary of an image and which are not circular are need to be removed [6].
Circularity of an object is found by using eccentricity of an object. Eccentricity of circular or elliptical object lies in the range of 0 to 1. Gestational sac and biparietal diameter are almost circular in shape, however, eccentricity value of these object is almost 1. Hence algorithm is developed to generate the eccentricity value of each and every object present in image. Object whose eccentric value found out below 0.7 are eliminated. In this way GS and BPD boundaries estimated accurately.
D. Extraction of fetal femur length using Hough Transform
Segmented image of femur length consist of large number of false edges. Hence knowledge based filtering is done to locate the boundaries or edges of desired parameter. Prior knowledge based filtering proves invalid in case of femur length. Hence in this case Hough transform is used. In Hough transform all pixels are candidate for linking and thus have to be accepted or eliminated based on predefined global properties. An approach has been developed based on whether set of pixels lie on curves of predefined shape. Once detected these curves form the edges or region boundaries of interest. The linear Hough transform algorithm uses a two-dimensional array, called as an accumulator, to detect the existence of a line described by (1)
The dimension of the accumulator equals the number of unknown parameters, i.e., two, considering quantized values of r and ?? in the pair (r, ??). For each pixel at (x, y) and its neighborhood, the Hough transform algorithm determines if there is enough evidence of a straight line at that pixel. If so, it will calculate the parameters (r, ??) of that line, and then look for the accumulator's bin that the parameters fall into, and increment the value of that bin. By finding the bins with the highest values, typically by looking for local maxima in the accumulator space, the most likely lines can be extracted. The simplest way of finding these peaks is by applying some form of threshold, but other techniques may yield better results in different circumstances - determining which lines are found as well as how many. Since the lines returned do not contain any length information, it is often necessary, in the next step, to find which parts of the image match up with which lines. Moreover, due to imperfection errors in the edge detection step, there will usually be errors in the accumulator space, which may make it non-trivial to find the appropriate peaks, and thus the appropriate lines. The final result of the linear Hough transform is a two-dimensional array (matrix) similar to the accumulator one dimension of this matrix is the quantized angle ?? and the other dimension is the quantized distance r. Each element of the matrix has a value equal to the number of points or pixels that are positioned on the line represented by quantized parameters (r, ??). So the element with the highest value indicates the straight line that is most represented in the input image. Extraction of femur length is based on the concept of strongest line in the emerged image. Hough transform matrix indicates the two or more Hough-peaks because of noise present in medical ultrasound image. These hough-peaks represent the end points of most likely straight lines which appeared in output image. Norm of each and every line is calculated in order to detect the edges of femur length. Line whose norm is calculated as a positive integer value is considered to be the strongest line.
A. Data Acquisition
The test images are obtained from scanning system, namely THI Siemens machine with curvilinear probe with transducer frequency of 3-5 MHz. The fetal images are obtained from 5 weeks of gestation. Necessary care has been given to preserve the shape, size and gray level distribution as it affect the sonographic content of information. Manual examination of fetal parameters involves the use of joystick or pen for its measurement. Care should be taken to position the joysticks as any variation in the initial placement of joystick will cause the error in decision making and may lead to false prediction. It is often difficult to place the joysticks at the right position due to the inferior nature of clinical ultrasound.
In the automated gestational age estimation system implemented in this paper, the medical ultrasound images to be analyzed are preprocessed to increase the brightness of an image using an adaptive histogram equalization technique. Further the speckling artifacts are removed using an improved Gaussian filter; further wiener filter is applied to suppress the additive noise present in an image. These include the assessment of gestational sac in the first trimester and Femur Length, head circumference, biparietal diameter in the later trimesters. Figure 3 shows original image of desired fetal biometric parameters to be extracted. Figure 4 shows contrast enhanced image by application of adaptive histogram equalization algorithm. Figure 5 shows output images of GS, BPD, and FL after application of wiener filter.

Figure 2: (a) Original Image of Biparietal diameter (b) Original Image of Femur length (c) Original Image of Gestational sac
Figure 3: (a) Contrast enhancement of BPD (b) Contrast enhancement of FL (c) Contrast enhancement of GS
Desired features are segmented using global thresholding technique. Morphological approach is used for boundary detection and false edge removal of GS, BPD. Figure 5 and figure 6 shows binarised image, edge detection and extraction of GS and BPD respectively.
Figure 4: (a) Wiener filtered image of BPD (b) Wiener filtered image of FL
(c) Wiener filtered image of GS
Figure 5: (a) Binarised image obtained after thresholding (b) Edge detection using morphological operation (c) Image obtained after false edge removal (d) Extraction of gestational sac
Figure 6: (a) Binarised image obtained after thresholding (b) Edge detection using morphological operation (c) Image obtained after false edge removal (d) Extraction of biparietal diameter
Hough transform is used to detect the femur length with optimum accuracy and extraction of femur length is based on concept of strongest straight line whose end co-ordinate pixel value is maximum. Straight line whose norm is found to be positive integer value is considered to be the strongest line and represented in the red color. Figure 7 shows edge detection of femur length by morphological approach, Hough transform matrix to detect the straight line present in the image.
Figure 7:(a) Edge detection using morphological operation (b) Hough transform matrix of Femur Length (c) Extraction of Femur Length
Desired fetal biometric parameters i.e. GS, BPD, FL size measured on ultrasound machine by radiologist is considered and compared with results obtained from automated method.
In case of GS, largest diameter is found because sac is not exact circle. This largest diameter is found out by finding the maximum distance between two points on the
boundary. This boundary is expressed as array of (x, y) coordinates of points on boundary. For this, one point on boundary and it's distance with all other points is considered. Maximum distance between any two points is returned.
Out of 7 images, in 6 images gestational sac is accurately detected. Table 1 indicates comparison of automated and manual method of gestational sac at various level of trimester and gestational age is predicted. Error measurement of size can be found out by following equation
error% = 100 | Sman ' Sauto / Sman | (2)
Where, Sman - size obtained by manual method
Sauto - size obtained by automated method
Gestational age = [ 28 + (diameter of gsac* 10) ] (3)
Table 1:Comparison of automated and manual gestational sac parameter in
First trimester
Automated Measurement in cm
Manual measurement in cm
Error in
Gestational Age in weeks
6 wk 1day
6 wk
6 wk 2 day
7 wk 5 day
6 wk 3 day
8 wk 2 day
9 wk 5 day
Fetal growth estimation in the later trimesters requires the estimation of biparietal diameter, femur length. In automated method, the BPD is measured on a transverse axial section of the fetal head. The BPD is measured from the outer edge of the nearer parietal bone to the inner edge of the more distant parietal bone. bone to the inner edge of the more distant parietal bone. In assessment of femur length, Euclidian distance between the end (x, y) coordinate is measured. Then this distance is converted to cm. Out of 7 images, in 6 images biparietal diameter is accurately detected. Set of 5 images is obtained for femur length measurement. Comparison of estimated and manual measurements of BPD and FL is detailed in Table 2 and 3. Error measurement can be calculated using eq 2.

Table 2:Comparison of automated and biparietal parameter at various level of gestation
In cm
In cm
Gestational age in weeks
19 weeks 5 days
11 week 6 days
14 weeks 3days
13 weeks 5days
16 weeks 3 days
20 weeks
21 weeks 5 days
*dates in the table correspond to the date in which a subject underwent ultrasound scanning for estimating the GA
Table 3: Comparison of automated and manual femur length parameter at various level of gestation
In cm
In cm
Gestational age in weeks
13weeks 1 days
17 weeks
13 weeks 3 days
16 weeks 2 days
13 weeks 6 days
*dates in the table correspond to the date in which a subject underwent ultrasound scanning for estimating the GA
The BPD measures 1,82 cm at the end of 11th week. It grows approximately by 2 mm per day and reaches 5.3 cm at the end of 21 weeks whereas the femur bone which has been measured at 1.0cm at the start of 13th week reaches 1.71 cm at the end of 17th weeks.
The experimental results obtained were compared with the Ultra sound examination report of a subject with clear
menstrual history index, obtained from Dr Mahajan polyclinic. Parameters estimated using the automated segmentation algorithm give an overall accuracy of 87%
This paper has presented an improved preprocessing and segmentation scheme for fetal biometric features segmentation from ultrasound images. The morphological operation based segmentation scheme has produced stable and reproducible results of GS and BPD with enhanced edges. Hough transform detect the femur length with maximum accuracy. The accuracy of the gestational age prediction largely depends upon the estimated values and it has been shown that the gestational age can be predicted with an accuracy of ??2 days which is much less compared to other methods. Hence this proposed work is expected to assist health care professionals in making decisions effectively and provide enhanced health care.
[1] Narendra Malhotra, Pratap Kumar, S. Dasgupta, R Rajan , 'Ultrasound in Obstetrics and Gynecology,3rd Edition ', 2012 Jaypee Federation of Obstetrics and Gynecological Societies of India pp. 71-72, 78.
[2] Sayan D. Pathak, Vikram Chalana, David R. Haynor, and Yongmin Kim,'Edge Guided Boundary Delineation
in Prostate Ultrasound Images', IEEE, transactions on medical imaging, vol.19, No.12 Dec 2000, pp. 1211-1119.
[3] Gonzalez, Richard E. Woods 'Digital Image Processing,', 2nd edi., pp. 528,534,575,635,613
[4] Jiankang Wang and Xiaobo Li, 'A System for Segmenting Ultrasound Images,' Pattern Recognition proceedings 14th international conference,vol. 1, pp. 456-461, 1998.
[5] Christine W. Hanna, Prof. Abou Bakr M. Youssef 'Automated Measurements in Obstetric Ultrasound Images' 1997, IEEE International Conference On Image Processing, volume 3, pp. 504-507
[6] Anthony Krivanek, Milan Sonka, 'Ovarian Ultrasound Image Analysis: Follicle Segmentation,' IEEE Trans. Med. Imag. Vol. 17, No. 6, Dec 1998, pp. 935-944.
[7] Richard N. Czerwinski, Douglas L. Jones, and William D. O'Brien Jr., 'Line and Boundary Detection in Speckle Images', IEEE Trans. on Image Processing, Vol. 7, No. 12, Dec 1998, pp. 1700-1713
[8] V. Dutt and J. F. Greenleaf, 'Adaptive speckle reduction filter for logcompressed B-scan images,' IEEE Trans. Med. Image vol. 15, no. 6,pp. 802'813, Dec. 1996.

Source: Essay UK -

About this resource

This Medicine essay was submitted to us by a student in order to help you with your studies.

Search our content:

  • Download this page
  • Print this page
  • Search again

  • Word count:

    This page has approximately words.



    If you use part of this page in your own work, you need to provide a citation, as follows:

    Essay UK, Automated Gestational Age Estimation For Monitoring Fetal Growth. Available from: <> [24-08-19].

    More information:

    If you are the original author of this content and no longer wish to have it published on our website then please click on the link below to request removal: