Description:The block diagram shown in figure 1 explains the various steps involved in diagnosing the PVSD. 101 describes the loading of 4CV ultrasound images. A pre-processing stage is needed in order to filter the speckle noise present in the medical images. Anisotropic diffusion filtering is identified as it preserves very fine details like edges, contour and other important features which are necessary for an image 102. In order to segment the sonographic biomarkers of the fetal heart, K-means clustering technique is proposed 103. To classify the abnormal image from the normal image deep learning methodology is identified wherein here we propose Multilayer perceptron which is an efficient classifier 104. The prognosis stage is one where one can come up with a
conclusion whether the loaded image is a normal image or an anomaly image 105 and 106. The diagnostic result from the doctor is given in the last stage 107. Finally, both the diagnostic data and the output from the classifier are compared to find the efficacy.
Figure 2 shows the original four chamber view of the fetal heart with regions marked. It also shows the zoomed version of the heart 201. The fetal heart has thin wall chambers, which is one of the key reasons for developing a smart system to identify the PVSD from ultrasound images. Thus, the process of designing and implementing a smart system for ultrasound image analysis is still a difficult task.
Thus, there exists a demand to develop a computerized image analysis approach for detecting fetal cardiac structures from 2D 4CV ultrasound images, which would help the gynecologists to enhance the rate of prenatal CHD diagnosis. This proposed automatic detection model is the first system implemented to predict ventricular septal of the heart prenatally.
This proposed system is the first attempt designed to help the radiologist, sonographologists and gynecologists to interpret the diagnostic details from the ultrasound images and to take fine
decisions about the absence or presence of PVSD. Preprocessing the image with ADF based filtering is the first step in the diagnosis procedure.
When compared with the existing filters, the robust ADF despeckling approach outperforms well. Image segmentation module combines the robust pre-processing methodology and segmentation technique based on K-means clustering approach to highlight the sonographic biomarker of the PVSD CHD. The simulated results proved that the proposed segmentation algorithm is predominantly suitable in delineating the ultrasound fetal RVLV chambers and to interpret the pathological markers of the US images. Classifying the US images as normal or abnormal is an important step in this research study. The segmented output is taken as input for the classifier. The proposed MP classifier worked extensively in classifying the disease that is PVSD. This is the first pioneering method to quantify the features of the image automatically and to diagnose the fetal heart asymmetric appearance. This diagnostic classification accuracy level is quite high for predicting the images which contains anomaly. , C , C , Claims:I/We Claim,
1. This smart system comprising of anisotropic diffusion filtering module, K-means segmentation module and MP module for diagnosing whether the US image is detected with ventricular septal defects; this method will act as a secondary tool for the sonographers, radiologist and gynecologist for detecting either structural or functional disease of the heart prenatally.
2. The method claimed in 1 make use of advance deep learning classifiers to automatically detect the heart features to increase the accuracy rate.