Ensemble-based advancements in maternal fetal plane and brain plane classification for enhanced prenatal diagnosis

In the realm of maternal healthcare, accurate fetal plane detection is of paramount importance. This paper introduces a novel approach that leverages ensemble techniques to enhance the precision and dependability of fetal plane classification. We address two pivotal classification tasks: first, the categorization of fetal planes into six distinct classes, encompassing critical regions such as the fetal abdomen, brain, femur, thorax, maternal cervix, and other areas. The second task focuses on the nuanced classification of fetal brain planes, further refined into trans-thalamic, trans-cerebellum, and trans-ventricular subtypes. To address these challenges, we harnessed the power of six distinct pre-trained models, rigorously training each for 50 epochs. Our results underscore the consistent superiority of InceptionResNetV2, DenseNet121, and Xception. Through a pioneering ensemble approach, we synergistically harnessed their capabilities, leading to enhanced classification performance. This research promises to augment the precision of maternal-fetal plane classification, with the potential to revolutionize clinical practice and healthcare research.