Most of us have heard it before: with hard work and the right attitude, you’ll be successful. But that’s easier said than done.

In fact, an A.I. system based in Japan became the embodiment of the phrase. Hisashi Kambe, a man with a love for computers, who worked with programs for decades, developed an A.I. system with the ability to identify pastries in bakeries. And this very A.I. system might be the key to identifying cancer cells.

It took a long time for Kambe and the A.I. system to reach where they are now. Back in 1982, when Kambe realized he didn’t want to continue the family lumber business, he began a small software company. Kambe pursued different projects of interest. In 1984, he developed a program called SUPER TEX-SIM, which aimed to fix the problem with planning programs that monitored the looms in the textile business. It didn’t do well in the market until 1985, when he landed a distribution deal with Mitsubishi’s fabric division. It was then that Kambe established BRAIN Co., Ltd. BRAIN worked on various projects for two decades until a restaurant chain sought them out in 2007. The chain planned to start a line of bakeries, but there was a major problem.

One of Japan’s most important imports is bread, and Japanese bakeries hold their reputations to the range of pastries and breads they have. Take a look at Japanese candy. Kit Kats in Japan have more than three hundred different varieties. New bread is being baked all the time, with bakers developing more types and contributing to the vast variety of breads.

While performing a market field study, analysts for the bakery line established that the more variety the bakery offered, the more revenue it would generate. In addition, they discovered that bread in open baskets sold better than wrapped pastries, which led to the problem. The restaurant chain needed to find a way to keep hundreds of pastries unwrapped and to sell them properly. Since the breads wouldn’t have barcodes, a cashier would need to identify them by hand. This would require heavy training of employees to name each pastry and lead to unsanitary conditions from breads being passed around, unwrapped, by hand. Looking for a solution, the chain found BRAIN.

BRAIN had its work cut out for them. The first issue they landed on was that some breads that were different looked similar, but also that some breads that were similar looked different. Not to mention, Kambe’s company found itself in a financial mire in 2008, having invested heavily in the pastry algorithm.

The team tested various cameras and lights, as the algorithm would need to work in different and inconsistent lighting and times of day. Bread also could be given to the machine messily, as in multiple pieces squished together or some torn. When they found that amount of time baked could alter color, they conducted a study on the different baking times and color of pastry. All in all, Kambe and his team spent half a decade on the pastry project. However, it came with results. In 2013, they built a machine that could scan and analyze bread and solve the multitude of problems they had initially encountered. In the same year, BakeryScan was released.

BakeryScan, which currently costs almost $20,000, was implemented into hundreds of bakeries. Of course, BakeryScan ran into its fair share of feedback errors, but with time and effort, those too were addressed. Each time, the device became more accurate than before. BakeryScan also became known in Japan by many bakery fans, appearing in advertisements and even appearing on a high school entrance exam.

It was a while after that that a doctor at the Louis Pasteur Center for Medical Research, based in Kyoto, watched a video on BakeryScan, and it occurred to him that cancer cells, when looked at through a microscope, seemed similar to… well, bread.

Soon after he got in touch with BRAIN, the team got to work on their new project, and the A.I. system would be able to measure the nuclei of cells on a microscope slide. In addition, hospitals began using the A.I. more diversely, such as for applications like sorting medicine. However, hospitals weren’t the only ones using the system. Other people began using it as well, like another company who used it for identifying bolts wired wrong.

After attending a conference in 2018 to speak about the A.I. system and its use for cancer cells, Kambe saw how the attendees were interested in using these deep-learning systems and wanted to contribute. And he and his team did: by creating Cyto-AiScan. Cyto-AiSCAN, the system to identify cancer cells, is currently being tested in two major hospitals in Kobe and Kyoto and is now able to search all cells on the entire slide for cancer cells, rather than looking at one cell at a time.

There is still a lot that can be done with the A.I. system.  Yet, it has the potential to make a difference in not just the field of medicine, but in other fields needing the efficiency of such an A.I. BRAIN’s Cyto-AiSCAN, evolved out of the famous BakeryScan, can now join the effort in fighting cancer.

 

written by Saanvi Gutta

edited by Keerthi Selvam and Tryphena Pilli

 

References:

Somers, J. (2021, March 18). The Pastry A.I. That Learned to Fight Cancer. The New Yorker. https://www.newyorker.com/tech/annals-of-technology/the-pastry-ai-that-learned-to-fight-cancer 

Image Sources:

[Japanese bread] [Photograph]. Bokksu. https://www.bokksu.com/blogs/news/a-guide-to-japanese-bread-kashipan