© Getty Images Bank

Today, we’ll meet with Alethio, a developer of a prenatal ultrasound image conversion solution. Let’s hear from the president of the company, Kim Da-woon(김다운). 

For the past two years our company has been a developer of an artificial intelligence or AI-based image conversion platform. 

I read a lot of fantasy tales to come up with an appropriate name for my company. “The Golden Compass” was one of them. In the book, the Golden Compass is referred to as the alethiometer, or ‘truth measure’. Inspired, I named the company ‘Alethio’. The name contains our hope that we’ll show something people didn’t know before – the face of their unborn child while it is still growing in the womb.

Our service converts a three-dimensional ultrasound image of a seven-month old unborn baby into an image of what he or she will look like when born.

Before setting up his company in 2016, Mr. Kim worked at the big data research institute Samsung SDS, an information technology arm of Samsung Group. There, he began to take an interest in “deep learning,” a form of artificial intelligence or AI. It refers to computer algorithms that can learn to perform certain tasks on their own, just like humans. 

Kim later entered the Korea Advanced Institute of Science and Technology. While studying and preparing to start his own business at the school, he happened to see ultrasound images of growing babies posted by his colleagues on social media. He thought that it would be possible to predict what the babies would look like after they are born by combining fetal imaging with deep learning technology. After hard work and experimentation, Kim was confident in his business idea.

© ALETHIO

Our technology is based on highly-regarded Generative Adversarial Networks or GAN algorithms. It is considered quite a new AI technique, as it was first introduced in 2014.

Parents are naturally curious about their new child, but it can be difficult to see what he or she will look like. Our service gives them that opportunity.

These days, more and more deep learning algorithms are applied to facial recognition in cameras. But with more advanced technology, we create images themselves. Of course, we just cannot create images at random. Based on some characteristics found in three-dimensional ultrasound fetal imaging, we produce a photo of the baby.

The company uses GAN algorithms to read and extract necessary characteristics from prenatal ultrasound images and create an estimation of the baby’s face. Alethio was the first company in the world to try out this new technology. 

In most cases, fetal ultrasound images do not show the entire face clearly. That means we have to restore the hidden part of the face using only what we can see. This was technologically challenging, and we experienced much trial and error. 

We made more than 100 models during a six-month period. A model here refers to a cluster of learning algorithms. Among them, only about four turned out well. We repeated the process to find more successful models or learning methods. 

Alethio organized its research staff with professionals who had extensive working experience in deep learning technology, and used GAN algorithms, which can learn to generate and restore images. As a result, the company was able to develop its own technology to create a full ultrasound image of a baby, even if their face is covered by the placenta or hands. Indeed, there is not much difference between Alethio’s AI generated photo and the real child after it is born. Thanks to this remarkable technology, Alethio received investment from Naver, the largest portal operator in Korea. 

When I was a graduate student, Naver held an investment briefing session at my school followed by a dinner. An employee at the company, who sat down across from me, asked me if I had any plan to start my own business. I answered yes and showed him a photo of an unborn baby. He seemed surprised, almost incredulous. He asked “Is this possible?” Soon after, Naver offered to financially back my company. They thought that my idea would meet the needs of would-be parents and they made their investment decision. 

Korea’s most widely-used web portal Naver has supported promising start-ups with high technological skills since 2015. It is far from easy to be selected, however. Out of 2,300 start-ups considered over past three years, only 20 firms, including Alethio, were fortunate enough to qualify for financial backing.

Alethio also received special attention at the 2018 Japan Bootcamp in September. The event connects Korean start-ups with Japanese investors and market access opportunities. During the event, Alethio’s technology elicited exclamations from the participants, who were amazed to see that a baby picture converted from its prenatal ultrasound image was almost identical to the photo of a real infant. 

But it was concluded that Alethio’s technology would not find traction in the Japanese market, as Japanese parents do not often take ultrasound pictures of their babies. Still, the company continues to explore opportunities overseas. 

 

During the Japan Bootcamp event, we researched the Chinese market. Fortunately, we found that there was great demand for our technology in China since new parents in China, like Korea, are eager to see 3D ultrasound images of their unborn children. About 18 million babies are born in China each year. It’s certainly a huge number with great potential for Alethio.

Alethio will launch its commercial services this month and has secured contracts with large maternity clinics. “Challenge” is probably the best word to describe the young start-up, which has taken its first step toward a bright future.