Histogram Equalization Methodologies Comparison for Deep Learning model in Fetal Heart Image Semantic Segmentation

This study aims to evaluate the impact of image quality enhancement on the accuracy of fetal heart image segmentation using fetal echocardiography and deep learning approaches. Fetal echocardiography is used to assess the structure and function of the fetal heart and detect heart abnormalities from the early stages of fetal development. In this study, deep learning methods, specifically the U-Net algorithm, are employed to identify fetal heart structures through echocardiographic images. The research findings indicate that enhancing the quality of fetal heart images through preprocessing methods such as Histogram Equalization (HE), Adaptive Histogram Equalization (AHE), and Contrast Limited Adaptive Histogram Equalization (CLAHE) affects the results of image segmentation. From the experimental results, it was observed that images processed with AHE yielded the best outcomes with an accuracy rate of 74.69%, IoU of 0.2428%, and DSC of 0.4842%.