Aside from Taylor Swift, there are few topics that have dominated more recent headlines than artificial intelligence (AI). This is especially true in the realm of healthcare, where AI is being used in everything from medical transcription to detecting diseases such as cancer.
Here at Breaking Cancer News, we’ve covered AI extensively – from applications of the technology in diagnosing breast cancer at early stages to a model that helps determine the origin of unknown cancers. To the casual observer, it may seem as though AI’s potential to transform healthcare is virtually limitless.
Not unlike other technological innovations, many AI tools got their start in industries and settings outside of the world of healthcare before expanding to applications such as the detection of cancer.
One AI system in particular demonstrates how certain core concepts of machine learning can be applied in the fight against cancer, and how the cycle of innovation remains in constant motion.
The Beginning…and Bear Claws
Arguably the most fascinating story around AI and cancer originates in Japan and begins with an AI system first developed to distinguish between various pastries at a bakery counter. You read that correctly, pastries. The system, known as BakeryScan, utilizes deep learning to correctly charge customers for their selections.
The innovator behind this revolutionary pastry identification system is Hisashi Kambe, a Japanese computer systems engineer and the CEO of BRAIN Co., Ltd.
With more than 40 years of experience in software and systems development, Kambe and his team have created everything from early systems designed to display the score of a baseball game to programs that improved the textile production process. But in 2007, Kambe and BRAIN took on an unlikely project that would prove to have much larger implications for the application of AI.
Nearly 20 years ago, a large restaurant chain approached Kambe with a very specific problem. The group was planning to open a line of bakeries that would offer an exceptionally wide variety of pastries, with new products added on a weekly basis.
The push for such an extensive pastry lineup was based on consumer data that showed that Japanese customers much prefer a plethora of options, and establishments that meet this need significantly outperform those with limited menus. The consumer study also found that pastries sold much better without a wrapper or packaging, as they were perceived to be fresher.
However, the inherent problem with a strategy built on these findings was that the unwrapped pastries offered no place for a barcode and easy scanning. Furthermore, offering such a large selection with no barcodes made it virtually impossible to train staff on all the various types of pastries – in some cases more than 100 – and long lines ensued as employees tried to distinguish between a bear claw and one of many glutenous concoctions the chain of bakeries offered, such as the “ham corn.”
Enter BakeryScan.
After five years of development and various prototypes, Kambe and his team managed to build a device that could identify even the most subtle differences in baked goods by taking a photo of the pastry and analyzing its features. They called it BakeryScan.
While the idea that machine learning can be used to identify and accurately price a pastry with a simple photo is remarkable, this application of the technology was only the tip of the iceberg.
BRAIN soon began expanding the fundamental technology behind BakeryScan – which the company branded as AI-Scan – to various other applications, including the identification of cancer cells.
Scanning for Cancer
Nearly 10 years after Kambe began developing the technology that would become BakeryScan, a doctor in Kyoto contacted him after seeing an ad for the system. He noticed that the pastries being scanned by the device were remarkably similar to the cancer cells he had been studying. When the doctor asked if BRAIN would be interested in testing the technology to see if it could detect cancerous cells, Kambe and team jumped at the opportunity.
The testing proved that not only could BakeryScan identify cancerous cells, but it could also do so with a high level of accuracy.
The BRAIN team found that BakeyScan could detect a cancerous cell under a microscope with 98% accuracy. Initially, the device was limited to viewing one cell at a time. But following modifications and additional rounds of testing, the technology was able to look at an entire slide of cells under a microscope and identify the cancerous cells at the same 98% rate of accuracy.
BRAIN’s cancer identification program, now dubbed “Cyto-AiSCAN”, has since been adopted in two major hospitals in Kobe and Kyoto, Japan. In Kobe, the technology is being trained and fine-tuned to detect bladder cancer.
According to the company, today Cyto-AiSCAN is 99% accurate and allows technicians to review up to four times as many cases per day.
This increased efficiency and speed in identifying cancer could have significant implications for patients, as early detection greatly increases the survival rate.
“This is groundbreaking work and another example of the use of AI in healthcare,” said Dr. Gerald Grant, Chief of Neurosurgery at Duke University and a Teen Cancer America board member. “We look forward to follow up studies across all cancer types.”
What’s next?
As Cyto-AiSCAN heads to market, it is not without competition, as various other AI solutions have been unveiled recently boasting the ability to detect cancer early and offer insight into its treatment.
But BRAIN remains focused on offering an innovative solution, especially within a region that has a particular need for early detection.
According to the World Health Organization (WHO), 10 million people died of cancer in 2020. The organization states that nearly half of these deaths occurred in Asia, where historically the disease has often been detected at more advanced stages.
In Kambe’s home country of Japan, for example, the incidence of cancer is increasing, yet the cancer screening rate remains low compared to that of other countries.
For its part, BRAIN continues to innovate – refining its AI and training Cyto-AiSCAN to detect and diagnose more cancer types. In recent interviews, Kambe and his team have also been quick to note that their AI solution is intended to be a tool to help healthcare providers, not replace them.
Perhaps the machines aren’t coming for us after all.