Overview
An artificial intelligence scanner system made to differentiate more than 50 types of pastries and bread is now being used in the healthcare field to detect cancerous cells.
Artificial intelligence (AI) has found its way into robotics, sports, drones, advertising and more as businesses discover how the technology can improve processes and address operational challenges.
In Japan, an unexpected use case of AI in bakeries inspired a doctor to look into adapting the system to detect cancerous cells.
Differentiating a Bun from a Donut
In 2007, analysts conducting market research on bakeries in Japan found that a bakery would sell more bread the more varieties it offered, and that pastries sitting in the open instead of in packaging sold better. Perhaps they appeared fresher to the consumer. The ensuing difficulties — with cashiers trying to memorize hundreds of pastry types without any labels or barcodes, and checkout processes delayed — led a bakery chain to turn to an AI company to automate the checkout process.
The resulting AI system can scan bakery items on a tray, identify each unique pastry and correctly price it, expediting the checkout process while also making it more hygienic. Initial setup and a feedback mechanism in the system allows employees to specify items or correct the AI so that it achieves the highest levels of accuracy.
This is BakeryScan. It is equipped with a backlight to eliminate shadows that can obstruct the pastry shape, a magnet system that can tell when a pastry has been ripped and whether it is actually one item, and another system that can classify two separate items touching. Through these functionalities and more, BakeryScan affords business efficiencies and allows bakeries to more easily broaden their offerings.
Inspired by Bread
Seeing BakeryScan on the news in 2017, a doctor at Kyoto’s Louis Pasteur Center for Medical Research realized how, under a microscope, some cancer cells closely resembled some of the pastries and baked goods that the AI system was scanning.
The now specialized Cyto-AiSCAN uses the same technologies that can differentiate a donut from a cream puff to identify cancerous urinary cells with 99% accuracy. By measuring the nuclei of the urinary cells and picking out certain visual cues, the deep learning mechanisms can determine if the cell is diseased.
Over time, Cyto-AiSCAN can analyze an entire microscope slide of cells at once instead of each cell individually, dramatically speeding up the diagnostic process and helping doctors more effectively treat patients.
The AI technology is now also distinguishing different pills and locating problems in mechanical engineering. From helping small businesses expand their operations to saving lives, AI is proving to be an increasingly critical capability for processing large amounts of data and addressing highly complex challenges.