Deep learning algorithm based on convolutional neural network has become a new kinetic energy of medical revolution since 2012. Based on CT, MRI, X-ray, ultrasound, thermal infrared, smear, electrocardiogram and other medical image of the intelligent auxiliary diagnosis and treatment system, in Clinical use has been proven effective.
Not long ago, in a chat with Zou Guowen, a partner of Dao Investment, Mr. Zou disclosed that there are more than 100 companies in China that are dedicated to the development of intelligent auxiliary diagnosis and treatment systems. So many entrepreneurs or entrepreneurs get out of here. On the one hand, it proved its great contribution to medical advancement, on the other hand it also heralded its potential, huge commercial value space.
The virtual reality behind the investment enthusiasm was basically clear in 2018. Some couldn't get financing, and the capital chain was broken. Some of them had been exposed, and some of them had been deeply practised for years. Innovation is still working hard, and some are still leading due to the strength of technology, capital, and medical resources. The general market structure is basically stable this year, leaving only a few companies in each subdivision.
If 2017 is still spelling out the accuracy of algorithm and image recognition, the fight in 2018 is to get certification and commercialization. How the product will be sold will become the key research topic of this year's enterprises. Since last year, the billion euro think tank and many Entrepreneurs and investors talked about related topics, and also formed some inconsistent views and opinions. They hope to share and discuss with experts.
I. Current mainstream products in the market
The identification and diagnosis of lung nodules is the starting point for training of most enterprise algorithm models. On the one hand because China is a big country with lung cancer, lung image data is the most abundant; on the other hand, the research on lung nodule recognition is the most mature in the world. In China, the rapid progress of lung cancer diagnosis and screening is mainly due to science and technology, Tuma Shenwei, Jianpei Technology, Voxel Technology, Yiyi Medical, Huihui Huiying, Shenrui Medical, Vision Medical, China Resources Wan Liyun et al. In addition to lung cancer, automated detection of stroke, breast cancer, hepatocellular carcinoma, nasopharyngeal cancer is also becoming the focus.
Intelligent CT-assisted screening product (AI-CT) functional interface (Source: Imagine Technology)
The auxiliary screening for cervical cancer, the cervical smear-assisted intelligent screening system, is a field that is attracting the attention of the industry. This application is mainly aimed at the pathology department and assists pathologists in the automatic detection of cellular DNA on pathological sections. There is Landing Medical, DeepCare. The former's core product, Cell DNA Auto-detection Analyzer, is the earliest product to obtain Type II card in China.
Diagnosis and screening of diabetic retinopathy, cases are also increasing. Voxel technology, Airdoc, DeepCare and other companies have launched related products.
There are also a few applications that are less involved in the enterprise:
· Tencent’s hard screening of esophageal cancer screening products;
· AI-ECG-Platform, introduced by Lepu Medical, is an automatic analysis and diagnostic product for ECG;
· The X-ray fracture diagnostic system introduced by Hui Huixin
· The diagnostic screening of neonatal jaundice launched by Paula is an AI product for skin images.
In addition to the diagnostic link, there are products worth looking forward to in the treatment segment. In the field of radiotherapy, companies such as Lianxin, Huiying Huiying, and Universe Medical have introduced automatic target area delineation systems that can help doctors complete the radiotherapy plan efficiently and accurately.
Intelligent target area automatic delineation diagram (Source: Lianxin medical official website)
In the field of surgery, Wei Jian Medical, China Resources Wan Liyun and other companies launched an image-based 3D reconstruction system based on evolutionary algorithms and VR technology to help doctors plan preoperatively more accurately.
Second, the starting point in the top three, the end of the grassroots
The training of the algorithm model requires a large amount of high-quality and diverse image data. This data is the largest in the top-three hospitals. Therefore, the company has sought cooperation with the top-three hospitals through scientific research and cooperation to obtain high-quality data and become the basic route. However, the top three hospitals do not lack quality imaging departments, radiotherapists, and grassroots hospitals. Therefore, from the perspective of practical applications, the real demand is at the grassroots level but not at the top three. The state has vigorously promoted medical informationization in the past two years. , grading diagnosis and treatment, telemedicine, is a major positive factor. Based on digital images, and with the instant interoperability of medical records of multi-level hospitals as conditions, the primary hospitals will have the ability to diagnose and treat serious diseases such as cancer.
Many companies build their products on the cloud, from a private cloud in a hospital to a public cloud in many hospitals, and it is the product form for future applications.
Third, do not do hardware?
Most of the companies mentioned above started as software systems. There are two profit models for doing software: one, selling to hospitals, charging hospitals for annual fees, and two, cooperating with hospitals, charging per patient, but The financial model of the profit model is still unknown. As several mainstream companies get certification and begin to enter the stage of marketization, domestic medical device manufacturers are likely to extend their “olives” and seek cooperation. M&A. The combination of the two, for start-up companies, can use the resource strength of medical device manufacturers and high-value-added services of hardware and software integration to quickly find a path for income generation; for medical device manufacturers, it can add value to products. Higher, but also enhance the technical strength and market competitiveness of enterprises.
Do you want to create hardware for your own company? This is a question worth considering carefully. The advantage of doing hardware is to make money from consumables. In the medical device market, it is not the number of devices that are actually making money, but consumables. For example, in 2017, the revenue was nearly 100 million. The products it relied on were not only the cell DNA automatic detection analyzer mentioned above, but also the cell fixatives, coloring agents, and disposable consumables such as cervical brush. The profitability point for sequencing companies is mainly based on body fluid collectors. Of course, doing hardware is relatively heavy for the company, and it is not in line with the situation in each area. Therefore, it needs to be prudent.
In general, Yiuou think tank thinks that 2018 will be the starting point for the commercialization of medical AI enterprises. The certification progress of CNDA is also accelerating. There are already several companies that have obtained Type II certificates. The companies that applied for the three types have also been in the past year or two. Settle the dust. We are very optimistic about the tremendous improvement brought by artificial intelligence in the next five to ten years for disease diagnosis and treatment. We also hope that we can enjoy technological advancements that bring us health and quality of life.
