AI ultrasound: a rising star in the field of AI imaging

AI ultrasound: a rising star in the field of AI imaging

Release Date: 2020-06-09

In all subdivisions of artificial intelligence (AI) medical imaging, AI ultrasound has received far less attention than AI CT imaging.

In early 2020, FDA approved the first AI-assisted Caption Guidance, with the Caption Guidance that will bring a breakthrough in the FIELD of AI ultrasound, drawing even more attention.

Improve diagnostic access

Ultrasonic testing is a safe and inexpensive medical diagnosis method, but in the actual diagnosis and treatment process, the use rate of ultrasonic is not high.

Currently, there are about 50 million doctors in the world, but only 2 percent of them are skilled in ultrasound scanning.

In addition, according to the statistics of The Chinese Medical Equipment Association, the number of ultrasonic equipment in China is not low, but the number of hospitals at all levels is not balanced.

By the end of April 2018, China's 2,427 tertiary hospitals had 24,270 sets of COLOR doppler ultrasound equipment, with an average of 10 sets.

The average number of color doppler ultrasound equipment in the second and first level hospitals was 5 and 1 respectively, with an obvious difference.

Ultrasonic testing has the characteristics of non-radiation and repeatable diagnosis. With the progress of technology, the cost of ultrasonic testing is becoming lower and lower, and the equipment is gradually miniaturized.

Only by lowering the barriers to use can it become a truly inclusive and portable diagnostic tool.

Ultrasound diagnosis is different from radiology diagnosis -- the radiologist can make diagnosis through still images, while the sonographer needs to collect dynamic images from different sections for real-time diagnosis. The acquisition and diagnosis of ultrasound images are highly dependent on the doctor's experience.

Equipped with ultrasonic equipment, AI can help solve two problems: one is how to better obtain images;

The second is how to better analyze the image.

To complete image acquisition and image analysis in a short time, AI needs to complete three steps.

With Caption Guidance as an example, the software will first use AI to guide the doctor to obtain the image. With real-time Guidance of AI, non-professional doctors will be able to collect the ultrasonic image.

The second step is to use the algorithm to find the best image;

The third step is image analysis.

Generally speaking, when engaged in interpreting and analyzing ultrasonic images, the doctor needs to go through years of learning, while Caption Guidance will automatically measure the ejection fraction through in-depth learning to assist the doctor in assessing the patient's condition.

AI makes ultrasonic detection easier and more accessible, and ultrasound is more conducive to the development of AI value.

In general, areas with economies of scale are more likely to produce AI value depressions.

Ultrasound is used more frequently than CT and NMR. Therefore, the commercial prospect of AI-assisted ultrasound diagnosis is more attractive.

“The big one and the small one”

The combination of AI and ultrasound is becoming a rising star on the AI video track.

In addition to helping with better diagnosis, AI can also perform automatic image quality assessment, image standardization processing, image sketch, automatic measurement and other functions in ultrasonic images.

These functions cannot be realized by a common scheme in ultrasound, so AI ultrasound takes two completely different routes.

One route is in the traditional ultrasonic department, where AI makes large ultrasonic equipment more intelligent, making the ultrasonic equipment no longer just an imaging product, but an intelligent terminal integrating data collection, management and analysis and deep learning.

In 2019, General Medical listed a LOGIQTM E20 equipped with cSound+TM image generator in China. Through image perception, LOGIQTM E20 can realize the functions such as tissue and organ structure screening, intelligent lesions segmentation and intelligent measurement, etc., to help doctors get rid of numerous and complicated image optimization and measurement work and focus on clinical diagnosis and treatment.

This device is mainly used in the clinical fields of intervention, thyroid, breast, musculoskeletal, pediatrics, heart and so on, to assist clinicians in accurate diagnosis.

Of course, AI's role is just icing on the cake for large devices right now, but it's expected that AI will play an increasingly important role in the future.

At the same time, compared with hardware ultrasonic equipment, AI software has a faster iteration speed, and software and algorithm are expected to become the mainstream research direction in the field of ultrasonic in the future.

Highly digital devices generate a lot of data, and how to connect and integrate the data is also a key research issue.

Another route is to apply it to primary care Settings.

There are nearly 900,000 primary-level medical institutions in China. Among these three links, medical treatment, drugs and inspection, more investment in inspection is essential to solve the structural contradictions in the medical system.

With the help of portable handheld ultrasonic equipment, it is a feasible way to empower primary medical institutions with ultrasonic equipment.

At present, the enterprises in the market of basic hospitals are mainly start-ups. These companies mainly apply AI technology to handheld ultrasonic equipment, and more of them are doctors with less experience in ultrasonic testing.

Ultrasonic diagnosis in the past, need to rely on professional doctor to identify the image in the eyes of anatomical structures, and AI by intelligent identification, can automatically find the best image and auxiliary diagnosis, the ultrasonic detection operator is not limited to, through professional training of doctors, the more common doctor of grassroots medical institutions also can make ultrasonic diagnosis.

The "late ripening" of the track stems from the technical barrier

Compared with THE CT images of AI, the AI ultrasonic track is not crowded, and only a few start-ups set foot in it.

Then, why has AI ultrasound become a late-maturing track in the field of AI imaging?

The main reason is that AI ultrasound technology is more challenging than other imaging fields.

There are three major technical challenges for AI application in ultrasound:

The first is to achieve real-time diagnosis. Different from the static images of CT and MRI, ultrasonic images are dynamic real-time images. The difficulty of ultrasonic detection lies in the simultaneous completion of image acquisition and film reading.

The acquisition of CT, mri, X-ray and other images is completed by technicians, while the film reading is completed by radiologists. The ultrasonic examination requires image acquisition and film reading to be completed simultaneously, which puts forward higher requirements for auxiliary diagnostic techniques such as algorithm and calculation force.

Secondly, due to the special data browsing, processing and storage habits of ultrasonic images, the image data is more difficult to obtain than CT images, and the size of the database is limited.

In addition, the degree of standardization of ultrasonic images is low, and the image definition mainly depends on the operating methods and equipment models of ultrasound doctors. AI ultrasound needs to be cleaned and analyzed by a strong expert team.

Finally, there is the limitation of the algorithm framework.It is very important for AI ultrasonic enterprises to have their own algorithm framework.However, at present, most enterprises are using open source algorithms, and very few enterprises have autonomous algorithms.

Ultrasonic AI is different from radiological AI. In order to ensure the accuracy and real-time performance of analysis, ultrasonic AI relies heavily on the algorithm framework independently developed. If the algorithm is too long, the processing speed of the equipment will be slow.

Ultrasonic detection requires high real-time performance, which can produce dozens or even hundreds of frames of images per second. If there is no powerful algorithm, such a huge amount of data cannot be processed.

In addition to the above three points, if AI technology is to be used in handheld ultrasound, computational force constraints need to be solved.

Because the handheld ultrasonic equipment is far smaller than the traditional ultrasonic equipment, the AI software is very testing AI computing power.

Not only do companies need a very accurate model to analyze ultrasonic video, but on top of that, they must ensure that the model works effectively with the limited resources of a tablet or phone platform.

As ultrasound technology evolves and advances, hardware and software capabilities are important.

At present, the technical barrier of AI ultrasonic software system is higher. In the ultrasonic equipment market with relatively homogeneous hardware, whether good AI software can be developed determines the application space of AI ultrasonic hardware to a certain extent.At the same time, for basic hospitals, portable or small equipment is cheaper than large equipment. Therefore, AI handheld ultrasound is more suitable for promotion than large equipment such as AI CT.

How to realize the intelligent ultrasonic flow of handheld AI, so as to better meet the application needs of general practitioners in grassroots hospitals, is the key problem that enterprises need to solve.

(Author: Arterial network)

Source: China Medical News

The copyright of this article belongs to the originator and does not represent the position of this site.

We reprint this article for the purpose of spreading more information, such as copyright matters, please contact delete.