Since the birth of medical artificial intelligence, the pace of exploration by Chinese researchers in the medical field has continued to accelerate.
At first, algorithm scientists started from the lung nodule with the most abundant data and developed a program that can extract pixel shadows in CT images. Automated nodule detection was the first achievement.
Deviating from the consideration of pure computer vision, a group of medical artificial intelligence companies began to think about the specific pain points of clinical departments from the logic of medicine itself, and then integrated artificial intelligence to solve problems. Take Shukun, an AI imaging company, for example. This company cuts into the industry from cardiovascular CTA inspections to create an intelligent CTA solution. This solution can be integrated into the existing CTA process of the hospital, reducing the time it takes for an inspection from 40 minutes to 5 minutes.
After CTA was widely deployed in thousands of hospitals, Shukun further deepened its research. On the one hand, Shukun cooperated with authoritative hospitals such as the Friendship Hospital, Anzhen Hospital, and Eastern Theater Hospital to create several of the country’s largest CT-FFR multi-centers; on the other hand, Shukun continued to use its original and leading advantages and once again won The head and neck CTA realizes a multi-task process, and finally realizes a one-stop stroke, advancing medical imaging to the AI3.0 era. When the time enters 2021, Shukun will advance the imaging intelligence to the extremely complex MR field of liver disease, and conduct comprehensive auxiliary analysis with the liver as the target.
Due to its complex characteristics, there are many complex lesions in the liver, which means that artificial intelligence needs to learn the MR characteristics of various lesions. It is very difficult to construct such an AI, but the current difficulties are the future barriers of the enterprise. Today, Shukun realizes the intelligentization of complex, multi-task and multi-process fields, which can automatically check the lesions in one inspection, analyze the lesions, and output reports. This means that medical imaging has entered the AI4.0 era of cross-modality, complex tasks, and composite processes.
AI continues to iterate and has reached a certain level of maturity, capable of making comprehensive judgments on various images. On this basis, Shukun began to try new ideas. At the 2021 World Artificial Intelligence Conference, Shukun Chairman Mao Xinsheng named his new achievement “Digital Human Body”.
Explore the next era of medical AI
The digital human body is Shukun’s platform-based vision for image AI. In the past exploration, Mao Xinsheng found that coronary CTA, one-stop stroke plan, and liver MR all made breakthroughs in specific scenarios. The next generation of artificial intelligence must break the limitations of a single department and a single mode, start from the patient’s human body function, and bring more accurate and convenient intelligent medical services to patients from a holistic perspective.
Taking this as an entry point, Shukun integrates digital products in the “digital heart”, “digital brain”, “digital chest”, “digital belly”, and “digital muscles” to form the “digital human body” “infrastructure” “.
So far, Shukun’s digital human body uses multiple modalities such as heart CT plain scan + vascular imaging, chest CT plain scan, head and neck CT plain scan + enhanced imaging, head CT perfusion imaging, liver MRI plain scan + enhanced scan and other multi-modalities as the engine, including Intelligent imaging of multiple parts such as cardiovascular/cerebrovascular/pulmonary disease/liver disease/musculoskeletal system can be applied to multiple clinical scenarios such as coronary heart disease/cerebral infarction/tumor/trauma/chest pain center/stroke center/cardiac specialty, etc. A complete set of hospital-level imaging comprehensive intelligent solutions has been formed.
Under the integration, the solution can combine morphology and function to detect diseases, form disease risk prediction, and apply anatomical structured databases, physiological structured databases, and pathological structured databases to form a multidisciplinary diagnosis and treatment supported by an imaging platform mode. In practical applications, patients only need to scan a specific organ once to detect a large number of common diseases, which will effectively reduce the cost of diagnosis and screening, and use AI to achieve true “treatment of the disease.”
At present, the research results of Shukun’s digital heart, digital brain, and digital chest have been included in top scientific journals, and their image factors have all exceeded 12. Shukun’s hospital’s scientific research capabilities have been recognized by the academic community, and he is also actively using the company’s AI capabilities for medical research exploration. Under the empowerment of AI, doctors may be able to dig out more valuable information from hospital data. .
Over 4 billion neurons, the digital human body opens a new era of AI
From AI1.0 to AI4.0, we can clearly see the changes in AI capabilities, but more importantly, lies in the meaning behind it.
Because the patient usually cannot predict the symptoms of the disease before the examination, even some tissues have already begun to disease, but the patient itself does not feel it. Therefore, in the AI1.0 era, what AI can do is to make a “yes or not” judgment on this result under the precondition that the patient has a lung nodule.
In contrast, AI4.0 represented by Shukun Digital Abdomen can conduct a comprehensive investigation of abdominal tissue. For example, if the patient’s routine examination shows that there may be a problem with the liver, then using the digital belly to analyze the MR image, AI can help the patient find out if he has liver cysts, hemangioma, FHN, liver cancer, liver cirrhosis, fatty liver. , Iron overload and other common liver diseases. The logical difference here is that AI does not need to pre-determine patient results, and can assist patients in diagnosing the possibilities through a single image result. Such an auxiliary method has more comprehensive clinical significance.
Of course, to achieve cross-modal and complex task processing, this requires a fundamental improvement in technology. Mao Xinsheng said in his speech at the 2021WAIC Health Summit Forum: “The AI1.0 era is usually dominated by single-modal, single-task AI, processing less than 200 images. At this time, the number of deep learning neurons is 300 million. But as the target continues to become more complex, the required artificial intelligence structure is also constantly evolving. Taking Shukun’s MR liver as an example, the algorithm has more than 4 billion neurons and can process more than 4,000 images at once. It is very difficult to build capacity and has become one of Shukun’s core barriers.”
Mao Xinsheng also said: “Connecting equipment, systems, and people such as CT and MR through AI will eventually generate standardized, automated, and intelligent image data, and then to the disease and patient-centric smart hospital, it is possible to realize that doctors + AI is greater than doctors. , The hospital + AI is greater than the effect of the hospital, and finally formed a new way of working for doctors from target detection, to automatic process, to multi-dimensional imaging, and then to precise diagnosis and treatment. This is the next step for Shukun, but it must Implementation is not easy. In the next few years, we will continue to explore the’digital human body’ and achieve digital coverage of more parts.”
Next step, into commercialization
For a long time, Shukun continues to lead the industry with original technology innovation. From the digital heart to define a new way of smart work, to realize Scan2Report, to the digital brain to create a stroke center multi-department application scenario program, to realize Scan2Decision, to the digital belly to create a new model of AI image-guided clinical treatment, to realize Scan2Therapy, to the digital human body to lead the integration of image AI Constantly improve its product system in the modernized era. Over the years, the advantages brought by this technological exploration have continued to accumulate, prompting Shukun to continue to expand its advantages in AI.
The commercialization of Shukun has accelerated significantly. On the one hand, it can cooperate with ecological companies to expand the reach of its products to doctors with the help of equipment that has already landed in hospitals; on the other hand, it can also rely on its own “digital human body” platform to expand the market and rely on multi-modal imaging. The integration of the building of new value.
With the continuous increase of the three types of certificates for medical AI products and the gradual adoption of the price list, the willingness of the entire industry to pay has been very clear. At this time, localized deployment, pay-per-use, and platform models may all be options for Shukun. With technology, products and market in hand, Shukun is leading the commercialization process of the industry.