MRI “AI 4.0” version.

On May 25th, ge Medical released a brand-new magnetic resonance artificial intelligence platform-“Zhizhen AI+”, which is also after the release of “Zhijian AI+” in early 2020. After 3.0″ platform, another breakthrough in this field.

It is reported that, technically, the “Zhizhen AI+ Platform” integrates the flash radio frequency imaging chain of GE Magnetic Resonance to ensure high-definition, high-quality Image data; in clinical applications, it is equipped with the “liver artificial intelligence analysis platform” application — it was developed by GE Healthcare and its Edison partner “Sukun Technology”.

“Zhizhen AI+” platform

The two parties will extend the application of AI technology in the field of magnetic resonance to the reading and assisting diagnosis cycle, thereby helping to achieve automatic extraction of multiple lesions in the liver and automatic description of lesion signs , Intelligent analysis of the nature of the disease, etc.

In addition, combined with the extremely fast enhanced DISCO LAVA technology of GE magnetic resonance equipment, “Zhizhen AI+” will increase the scanning speed of liver magnetic resonance dynamic enhancement by 60%. Accurately obtain the multi-arterial phase of the liver.

According to the introduction of GE Healthcare China’s magnetic resonance clinical application team, the DISCO LAVA technology is used to collect 6 phases of images in one breath-holding condition, which improves the scanning speed and efficiency and allows clinical The multi-arterial phase can be captured to ensure the success rate of the enhanced scan. The realization of this technology is a prerequisite for the application of AI-assisted analysis tools to the clinical application. It brings accurate image data foundation for the application of intelligent liver analysis after imaging, thus promoting the clinical application of AI tools. span>

Before this, GE Healthcare has carried out three iterations of the magnetic resonance AI platform in 7 years:

  • In 2013, GE Healthcare launched the MRI “AI 1.0” intelligent operating platform, which realized the automatic recognition of anatomical parts, automatic positioning and automatic continuous scanning functions of MRI equipment;

  • In 2015, the “AI 2.0” intelligent post-processing platform was launched, and the intelligent post-processing operation flow of one-key multi-mode fusion was realized on the MRI post-processing end;

  • At the beginning of 2020, the “Intelligent AI 3.0” platform was released, a breakthrough in the realization of AI deep neural network algorithms in the core link of the image imaging front-end, removing artifacts from the imaging source, optimizing front-end images, and increasing imaging speed1 The scanning success rate is increased by 20%.

Now on the basis of the “Smart AI 3.0” platform, the “Smart AI+” platform includes image chain AI, imaging platform AI, and clinical application AI. It integrates AI’s digital technology and magnetic resonance diagnosis in an all-round way, and further Unleash the potential of AI in the field of magnetic resonance technology.

The key point is that the continuous innovation of the imaging equipment itself provides richer imaging information while obtaining sufficient and high-quality data in a shorter period of time for the development of AI applications in the post-imaging part. , Landing lays a solid foundation and brings more possibilities.

Actually, there are many challenges in the onset and diagnosis of liver diseases.

First is the challenge of liver disease. According to the data from the China Cancer Registry Annual Report 2019: In 2015, there were about 3.929 million malignant tumors nationwide, and about 365,000 new liver cancer cases nationwide, accounting for 50% of the global new cases. Liver cancer is the second leading cause of death from tumors in China. Among people under 60, liver cancer is still one of the most common and lethal tumors. (The data from the National Cancer Registry usually lags about 3 years)

The second is the complexity of the disease. According to Yang Zhenghan, director of the National Clinical Research Center for Digestive Diseases and the Department of Radiology, Beijing Friendship Hospital, Capital Medical University, there are many types of liver diseases, which are divided into focal lesions and diffuse lesions according to their etiology, pathogenesis, and pathological status. Focal liver lesions can be further divided into cystic or solid lesions, and there are dozens of common types. These focal lesions can be divided into tumors orNon-tumor lesions, and tumor lesions can be divided into benign and malignant. Diagnosis and differential diagnosis are not only difficult, but also laborious.

Thirdly, it is difficult to diagnose diseases and lack of talents. Clinically, liver MRI can solve most of the problems in the diagnosis of liver diseases. However, Director Yang Zhenghan introduced: “There is a shortage of imaging doctors who can operate MRI equipment in a standardized manner and have strong ability to interpret images. Due to the complex characteristics of the liver, there are so many kinds of lesions in a single organ, and MRI has a lot of imaging information to learn. Judgment, so imaging doctors have a long process of learning and growing.”

In fact, several years have passed since AI and other digital technologies combined with medical treatment to assist diagnosis and improve efficiency, but in the field of magnetic resonance, the pace is relatively slow. This is related to the development of magnetic resonance technology, the complexity of imaging, and the large amount of data.

After MRI liver scan, various sequences and various angles produce at least thousands of different images. The huge amount of information, AI algorithms, clinical diagnosis basis, process, etc., require more accurate source data and more The foundation of an intelligent MRI platform, at the same time, also requires an in-depth understanding of the clinical and the development of AI algorithms for the diagnosis process.

In this regard, according to the team’s introduction, the above-mentioned “liver AI analysis platform” has realized the “post-full-process AI” technical concept, breaking the application of AI technology in magnetic resonance through the imaging cycle of magnetic resonance and entering the cycle of image reading and diagnosis , And through one-click automatic sequence recognition, focus extraction and sign description, to achieve intelligent analysis “reading pictures, easy to diagnose”; AI automatically extracts the characteristics of the anatomical features and information of the focus, and also injects momentum into scientific research.

As a clinical partner in the development of this clinical AI-assisted system, Director Yang’s team of Friendship Hospital pointed out, “The segmental analysis of the liver is based on blood vessels. After we extract the blood vessels, we can do 3D reconstruction. In the next step, there are standard models for liver transplantation or hepatobiliary surgery. After the vascular structure is outlined, the non-vascular structure is the focus. We tested the currently trained model and the segmentation accuracy Dice value is above 0.9 , The highest has reached 0.96 or more, which can meet the current general clinical applications.”