A Deep Learning Approach for Masking Fetal Gender in Ultrasound Images

  • 2021-09-14 16:00:29
  • Amit Borundiya, Arshak Navruzyan, Dennis Igoschev, Feras C. Oughali, Hemanth Pasupuleti, Mike Fuller, Vinay Kanigicherla, T S Aniruddha Kashyap, Rishabh Chaurasia, Sonali Vinod Jain
  • 0


Ultrasound (US) imaging is highly effective with regards to both cost andversatility in real-time diagnosis; however, determination of fetal gender byUS scan in the early stages of pregnancy is also a cause of sex-selectiveabortion. This work proposes a deep learning object detection approach toaccurately mask fetal gender in US images in order to increase theaccessibility of the technology. We demonstrate how the YOLOv5L architectureexhibits superior performance relative to other object detection models on thistask. Our model achieves 45.8% AP[0.5:0.95], 92% F1-score and 0.006 FalsePositive Per Image rate on our test set. Furthermore, we introduce a boundingbox delay rule based on frame-to-frame structural similarity to reduce thefalse negative rate by 85%, further improving masking reliability.


Quick Read (beta)

loading the full paper ...