Amniotic Fluid Assessment in Ultrasound Images Using Convolutional Neural Networks and YOLOv2

Amniotic fluid assessment is a crucial task in prenatal care, aiding in the early detection of potential fetal health risks. This study proposes an automated classification and detection system for amniotic fluid analysis in ultrasound images using Convolutional Neural Networks (CNN) and YOLOv2. The approach involves preprocessing ultrasound images by resizing them to a standard resolution, followed by classification into six categories: Oligohydramnion Clear, Oligohydramnion Echogenic, Polyhydramnion Clear, Polyhydramnion Echogenic, Normal Clear, and Normal Echogenic. A CNN-based architecture is employed for feature extraction and classification. In cases where the CNN identifies an abnormal condition, YOLOv2 is utilized to detect and localize the amniotic fluid region within the ultrasound image. Due to dataset limitations, the model is trained and tested on a restricted dataset, ensuring optimal generalization through augmentation techniques. The integration of CNN for classification and YOLOv2 for detection enhances diagnostic accuracy and efficiency. This automated approach can assist radiologists in making quicker and more accurate decisions, ultimately improving prenatal care and reducing the risks associated with amniotic fluid abnormalities. The proposed method demonstrates the potential of deep learning techniques in medical imaging and highlights the importance of AI-driven diagnostic tools in obstetric ultrasound analysis. 

Keywords: Dataset, Image Processing Techniques, Deep Learning, YoloV2 Detection and Convolution Neural Network.