Press Releases > HUAWEI CLOUD EI Wins First Place in the International Ultrasonic Image Segmentation Competition

HUAWEI CLOUD EI Wins First Place in the International Ultrasonic Image Segmentation Competition

Dec 25, 2018 GMT+08:00

Recently, the HUAWEI CLOUD enterprise intelligence (EI) medical imaging team made a technical breakthrough in ultrasonic image segmentation and measurement. They won the first place in the HC18 challenge hosted by Grand-Challenge with an average absolute value of 1.89 mm head circumference (HC).

What is Grand-Challenge?

Grand-Challenge is an international platform that holds medical image analysis competitions. It is dedicated to providing uniform data and standards for fair comparison among state-of-the-art medical image algorithms, promoting technical development. Over the years, thousands of world-class research teams have participated, and their data and results are often published as papers and presented at top international medical imaging conferences, such as MICCAI. Over 100 universities and research institutes worldwide took part in HC18, including Queen's University, The Chinese University of Hong Kong (CUHK), and the Chinese Academy of Sciences. The following figure shows the rankings of the Grand-Challenge HC18 challenge so far. For details, visit

Huawei Strives to Improve Medical Imaging

The fetal HC refers to the maximum circumference of the fetal head. It is commonly used to evaluate the size of a fetal head, predicting fetal development. If a pregnant woman does not remember the date of her last menstrual period, she can go to the hospital for B-scan ultrasonography to help predict the gestational age and her due date. Gestational age estimation predicts fetus development, including issues such as slow development. Therefore, the fetal HC measurement is vital.

Huawei proposes a deep neural network image segmentation model developed based on the general image segmentation neural network model in the industry. It integrates various advanced technologies such as a GAN, a multi-scale dilated convolution, and new Loss function to solve diverse issues such as small sample learning, low contrast of ultrasonic images, and blurred fetal head edges. It beats the previous world record achieved in fetal HC measurement and lays a solid technical foundation in the field of ultrasonic image segmentation.

Image segmentation plays an important role in quantitative and qualitative analysis of medical ultrasound images. The result directly affects subsequent analysis and processing. Correct segmentation guarantees accurate extraction of diagnostic information from ultrasound image as clinical applications, and is critical for quantitative analysis in clinics.

HUAWEI CLOUD EI medical imaging team's win in the challenge demonstrates that the medical image recognition technology of HUAWEI CLOUD is at the forefront of the industry. The team will continue to explore possible applications of AI in various medical image fields, such as ultrasound, pathology, CT, MRI, and endoscopy.

At HUAWEI CONNECT 2018, Huawei released its AI strategy and full-stack solution. As part of its full-stack AI solution, HUAWEI CLOUD focuses on making AI more convenient for enterprises and developers, for example, by developing the Ascend series chips and ModelArts. HUAWEI CLOUD EI offers multiple solutions to enable industrial upgrade and reconstruction by leveraging AI, big data, and cloud computing.