Fetal ultrasound images are usually noisy and have blurred borders. Ultrasonic images at this stage are mostly measured and evaluated by manual operation and rely deeply on the experience of professional clinical staff, which can lead to observer invariant or errors due to human operation. Moreover, the Nuchal translucency thickness needs to be measured on the middle sagittal plane (MSP) of the fetus; finding this correct observation plane within the ultrasound image is a very time-consuming and difficult task.  Our invention is to provide a method for automatically finding the MSP. This system uses artificial intelligence (AI) deep learning generative adversarial network to complete the task of detecting the MSP of the 3D ultrasound fetal image, which is more effective and stable than the general convolutional neural network and traditional image processing methods. The experiments show that the system has very high accuracy as well as very fast detection time.

The Nuchal translucency thickness needs to be measured on the middle sagittal plan (MSP) of the fetus. Traditional image processing or general convolutional neural network methods must find the fetal head or other feature parts before obtaining the MSP. This invention utilizes deep learning generative adversarial network, which can directly obtain the position of the MSP. Experiments show that this system has very high accuracy and much faster speed than the previous methods, which is a major breakthrough in automatic detection and practical application.

Ultrasonic scanning is non-invasive and has the advantages of low cost and real-time observation, so it is widely used in antenatal examination. At present, most of the ultrasonic image measurements and evaluation rely upon the experiences of professional clinical staff. To find the correct observation surface within the ultrasound image is a time-consuming and difficult task. This invention uses deep learning generative adversarial network, which automatically detects the position of the middle sagittal plane used to measure the Nuchal translucency thickness. Experiments show that this system can provide a fast and accurate solution for this task.