Feature Selection and Classification to Detect Fetal Abnormalities

The process of feature selection involves selecting a specific group of features necessary for image analysis. This process is a part of medical image processing, which includes pre-processing, segmentation, feature extraction, and feature selection. Medical imaging has advanced significantly in the recent past, which has aided physicians in diagnosing diseases using various medical modalities. Fetal biometrics, such as Amniotic Fluid Volume, Parietal Diameter, Head and Abdominal Girth, Femur Length, and Gestational Age, are identified and extracted using feature extraction. We can obtain a subset of features that enhance accuracy and computation time by using feature extraction and selection. This paper presents feature selection and classification methods that improve the accuracy of detecting fetal abnormalities in ultrasound images. We have used SVM-based LDA and PNN-based MLE methods for classification and compared the performance of both methods The PNN-based MLE technique outperforms the SVM-based LDA method, with an accuracy of 95.10% compared to 90.45%.