Segmentasi Kepala Janin pada Citra Ultrasound Menggunakan Arsitektur Jaringan U-Net

Abstract

Digital image processing has been utilized in various fields, including the medical field. One example is used to detect the location of vital organs in the human body. This research aims to produce a cut shape of the fetal head area on an ultrasound image (USG) by applying a deep-learning segmentation method. This research stage was carried out with ultrasound image acquisition, followed by an image preprocessing process to improve image quality so that the results of the segmentation process were better, followed by applying the segmentation method. This research focuses on using the U-Net method to segment the fetal head in ultrasound images. Using 995 ultrasound images of the fetal head in the training process, the best accuracy was obtained at 90.55%. The performance of the segmentation results for 335 ultrasound images of the fetal head at the testing stage using the Jaccard coefficient measurement was obtained on average at 87%. The results of this segmentation can be used for further purposes, such as fetal biometric measurements or 3D visualization.

Keywords – Image Segmentation; USG Image; Fetal Head; U-Net Architecture; Image Preprocessing