Oct 11, 2018,  by Allianz Partners Business Insights

Artificial intelligence: A Japanese system can detect problems in foetal hearts

The aim of RIKEN is to allow for the early diagnosis of congenital heart defects during pregnancy. To do this, the Japanese institute has developed a tool based on artificial intelligence and, in particular, the “object detection” technique. This promising advance could significantly reduce the number of newborn deaths, and should also help specialists to hone their expertise.

Japanese scientists from the RIKEN institute have developed a device that can detect abnormalities in foetal hearts in real time. This system, which is based on artificial intelligence (AI) and was designed within the framework of the AIP (Advanced Intelligence Project), could help doctors to diagnose heart problems both earlier and more accurately. This would be a valuable new tool for perinatal and neonatal medicine, reports News Medical.


20% of all newborn deaths


Congenital heart defects – which include abnormalities in the auricles, ventricles or blood vessel connections – are responsible for 20% of neonatal deaths. Diagnosing these abnormalities in utero could considerably improve the foetus’ prognosis by allowing specialists to perform prompt treatment within a week of birth. But it is still tricky to diagnose these defects early, due to the ultrasound scanning techniques currently used.

This is why AI technologies (like deep learning) are so useful in the medical industry. For example, machine learning can detect pathologies more quickly and more accurately. In neonatal cardiology, a lack of data prevented the effective use of AI. This problem was overcome by the RIKEN institute’s researchers: in collaboration with a Japanese university, they managed to develop a machine learning system capable of drawing precise conclusions from incomplete and small sets of data.


“Object detection” technology


Heart problems are usually diagnosed by comparing images of normal and abnormal hearts, an approach that is similar to the “object detection” technique. AI can learn how to spot and classify a number of objects using a set of basic data as a reference model. The scientists from the RIKEN institute provided their AI with a set of images of healthy hearts including 18 parts of the heart and its neighbouring organs. 

This enables them to develop a “Foetal Heart Screening System” that is capable of quickly distinguishing – in real time, if necessary – any significant differences between a scanned image and the reference images. This system might now be used to standardise differing diagnoses from several hospitals that do not use the same techniques or equipment. 

The Japanese institute is now planning on leading clinical tests in several hospitals across the archipelago, in order to broaden the learning database of its AI and further enhance its speed and accuracy. This innovative solution could, one day, help to overcome any medical disparities between regions by honing the specialists’ training and allowing them to make remote diagnoses using cloud-based systems.


Allianz Partners

Linked Topics
About Allianz Partners Business Insights