Intelligent medical imaging diagnostic system "reality is very skinny"

Since 2017, there has been a lot of discussion about the artificial intelligence bubble. I know that the last question, “How will this wave of artificial intelligence bubbles be shattered?” has received wide attention. Among them, PENG Bo, the co-founder of Weilin Technology, has the highest opinion. Tickets, he believes, "Artificial intelligence is a bit dangerous, because now it seems to be a problem." "At present, AI may not be enough to support an independent company, it is more suitable as a department of other companies, or acquired by other companies. ."

Many smart image front-line practitioners may strongly agree with their views. IFLYTEK wisdom of the medical department of medical imaging products, head Ma Wen-chun told reporters, "Today's smart medical images like the Internet a few years ago, we came in a rush, but the next step how to do, is a problem." Hui Hui Medical shadow Liang Enzhen believes that "in general, intelligent imaging diagnosis really goes into clinical diagnosis very little. At present, the industry tries to cooperate with doctors to do research or improve efficiency, but there is still a big gap in realizing the diagnosis rate."

You can hear the words "AI replaces the doctor" and "the accuracy of AI exceeds the doctor". At the same time, the "ideal is very beautiful, the reality is very skinny" sentiment frequently, the ideal rhetoric can be seen everywhere, what is the reality?

Some people say that the 21st century is the era when data is king; some people compare algorithms to engines, and data to oil; others emphasize that industry data, expert resources, and core technologies are the three elements that make intelligent images indispensable. No matter how important the emphasis is on the importance of data, we use image data as the path. The operation of the intelligent imaging company is a latitude, and we can see the real life of the intelligent imaging company.

Data end: guarantee quality, quantity and benefit

Although there are third-party imaging centers in China, the vast majority of medical imaging data comes from hospitals. It is reported that the image data generated by the big top three hospitals in one year is above 10T. Wu Bo, CEO of Yiyuan Intelligent, said, “The stock of image data in a single hospital is very large, and hundreds of new cases are also common every day.”

In the medical information system, the PACS system is responsible for medical image acquisition, data transmission and storage, image analysis and processing, and can be docked in different DICOM international standards between different PACS systems.

In general, hospital image data is mostly standardized and easy to read by machines. For this reason, smart medical imaging is considered by the industry to be the first to commercialize.

Yang Xiujun, director of the Imaging Department of Shanghai Children's Hospital, once said, "Many medical imaging fields are particularly suitable for artificial intelligence/image recognition technology. There are many manufacturers at home and abroad who are engaged in this aspect and have made some achievements."

Wu Bo told reporters that for AI, the image data itself has the advantage of standardization. However, data anomalies are also easy to come by. "In the case of CT, some patients are not lying but scanning; some are not advanced, but the feet are advanced; CT is 512 pixels or 768 pixels wide, the difference in layer thickness of different rows of machines and Thin-layer reconstruction algorithms can affect clarity,” he added. “As long as the process is fully considered and compatible with these changes, the available ratio of raw data is still very high.”

For an AI system, the data is more pre-conditional, and it is meaningful to increase the quantity while ensuring the quality of the feeding data. Judging the quality of image data depends mainly on the clinical purpose of the intelligent diagnostic products created by AI. In addition, for intelligent image diagnosis, the image data needs to be associated with more accurate diagnosis and later result correlation, otherwise the garbage enters and the garbage is discharged.

Liu Zaiyi, an associate professor at Southern Medical University, once said, "We have a lot of data. Our imaging department produces a lot of data every day, but how much data can be used? 1% is not there, there are a lot of error messages." He added, "Data There are no ways to control the problem of regulation, and clinical information is often wrong."

Take the intelligent diagnosis of lung nodules that is now very hot. For AI, it is valuable to have images of lung nodules. "In the resulting image data, only 10% or 20% of patients have problems. Even so, not all image data with lesions can be used." Liang Enzhen said. Inside the hospital, the medical imaging system and diagnostic report are two separate systems, and the two are not related. "It is very important to use the data to train AI: the system needs to judge whether an image has nodules and whether there is a lesion. The hospital shoots hundreds of images every day, and there is no sign of nodules. For AI, this is It is data without value." Liang Enzhen told reporters.

Lock Channel

The lock channel is one of the most important thing for greenhouse, it can lock the film ,shading net ,insect net .

It is hot galvanized or aluminum material.

We have different thickness.

Welcome to contact us

Lock Channel,Greenhouse Lock Channel,Channel Lock,greenhouse profile

JIANGSU SKYPLAN GREENHOUSE TECHNOLOGY CO.,LTD , https://www.engreenhouse.com