AI enables non-invasive screening for Down syndrome

The most common chromosomal abnormality causing intellectual disability and developmental delay is Down syndrome, also known as trisomy 21, which can be detected during pregnancy. Many pregnant women want to know if their fetus has this disease.

Viewing the response region to display class-specific information for tiered functionality. Image credits: Chinese Academy of Sciences Institute of Automation (CASIA)

To perform a non-invasive screening for Down syndrome using ultrasound images, researchers from the Chinese Academy of Sciences Institute of Automation (CASIA) have now built an intelligent prediction model.

June 21stst2022, the study appeared in Open JAMA Network.

Due to the technique’s reliability, ease of use, and low cost, ultrasounds have been routinely used for decades to screen fetuses for Down syndrome. But in actual ultrasonic testing, detection accuracy is less than 80% using standard ultrasonic indications.

Detection of Down syndrome frequently involves invasive techniques such as villus biopsy, amniocentesis, and fetal umbilical venipuncture.

In this study, researchers built a deep learning (DL) model using a convolutional neural network (CNN) to learn representative features from ultrasound images to detect fetuses with Down syndrome.

A CNN is a deep learning system that can take an input image, prioritize the distinct elements and objects it contains (i.e. apply learnable biases and weights), and distinguish between them.

Several hidden layers could be found in a CNN. The first layer captures edge detection, while the last layer captures more complex shape detection. There were eleven secret layers in this study.

The researchers also used a class activation map (CAM) to clarify what the model focused on and how it directly allowed CNN to acquire discriminating features for risk ratings, further interpreting the DL model of a human-readable manner.

Between 11 and 14 weeks of gestation, the midsagittal plane of the fetal face was imaged using two-dimensional ultrasound technology. Each image was divided into sections using a bounding box to show only the head of the fetus. The study included 272 people in the validation set and 550 participants in the training set, for a total of 822 cases and controls.

The first five levels of feature maps produced by CAM, according to the researchers, vividly described the learning process of representative features. The last layer CAM revealed the visually represented response zones for model decision making.

This non-invasive screening model designed for Down syndrome in early pregnancy is significantly superior to existing commonly used manual labeling markers, improving prediction accuracy by more than 15%. It is also superior to the current conventional invasive screening method for Down syndrome based on maternal serum.

Jie Tian, ​​Corresponding Author of the Study, Chinese Academy of Sciences Institute of Automation (CASIA)

The suggested concept should develop into a non-invasive, affordable and practical early pregnancy screening tool for Down syndrome.

The National Natural Science Foundation of China and the key R&D program of the Ministry of Science and Technology funded the study.

Journal reference:

Zhang, L. et al. (2022) Development and validation of a deep learning model to screen for trisomy 21 in the first trimester from nuchal ultrasound images. JAMA network open. doi:10.1001/jamanetworkopen.2022.17854


Comments are closed.