Performance Comparison of Various Machine Learning Algorithms for Ultrasonic Fetal Image Classification Problem
The machine learning uses statistical algorithms to provide the ability to learn with and without data programming. Image classifier is used to categorize a subject or an object present in an image into predefined classes. A statistical algorithm used in
- PDF / 507,733 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 59 Downloads / 228 Views
Abstract The machine learning uses statistical algorithms to provide the ability to learn with and without data programming. Image classifier is used to categorize a subject or an object present in an image into predefined classes. A statistical algorithm used in numerous fields like pattern classification, regression, control systems, identification, and prediction. This research paper presents a novel approach performance comparison of machine learning classifiers used for the classification of ultrasonic fetal images. It presents Gabor feature extraction and various classification techniques to classify three different classes of ultrasound images. The proposed method is presented as follows. Initially, the Gabor features are obtained from raw images. To remove redundancy in features and dimensionality reduction principal component analysis (PCA) is applied. Finally, the features obtained from PCA are fed to the various machine learning classifiers and its performances are evaluated. The simulations results are carried out using MATLAB image processing toolbox. From the results, it is observed that decision tree (DT) algorithm and multi-layer perceptron (MLP) perform to be closer and this classifier outperforms all other classifiers. Keywords Machine learning · Principal component analysis (PCA) · Decision tree classifier (DT) · Classifier · Multi-layer perceptron (MLP)
N. Sathish Kumar (B) · M. Kasiselvanathan · S. P. Vimal Department of Electronics and Communication Engineering, Sri Ramakrishna Engineering College, Coimbatore, Tamilnadu 641022, India e-mail: [email protected] M. Kasiselvanathan e-mail: [email protected] S. P. Vimal e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2021 P. Suresh et al. (eds.), Advances in Smart System Technologies, Advances in Intelligent Systems and Computing 1163, https://doi.org/10.1007/978-981-15-5029-4_6
61
62
N. Sathish Kumar et al.
1 Introduction Machine learning algorithm is for identifying patterns that can be applied to realworld applications. It is very useful tool that can help in interpretation of medical images [1]. Machine learning uses machine learning algorithm which used to compute the image features that are supposed to be of importance in making the prediction or final conclusion regarding the diagnosis of disease. Multi-modal biometric recognition systems have greater advantages that provide good antispoofing abilities. Gaikwad proposed a system that uses contourlet transform for analyzing the features present in palm print and palm vein images [2]. Feature detection may be used to find the image information and provide a local decision at every image point to check whether there is an image feature of the given type presented in that point. This feature detection technique provides robust to image transformations [3, 4]. A near-infrared (NIR) imaging of Palm Dorsa Subcutaneous Vein Pattern(PDSVP) system has been introduced for data acquisition, PDSVP extraction [5]. The machine learning algorithm identifies the most excellen
Data Loading...
