Diagnosis is one of the core aspects of medical care, and current Chinese medicine and Western medicine give diagnosis in different ways. Western medicine relies on 1) patient signs, 2) patient descriptions, and 3) examination data, of which 1 and 3 are the main basis for judgment. TCM doctors confirm the diagnosis by checking the physical signs, pulse diagnosis, and patient's description. In this process, the doctor obtains patient information, interprets the information for a longer period of time, and there is also the possibility of misdiagnosis. The current market for assisted diagnostic tools is essentially to solve the above problems.
AI to realize the auxiliary diagnosis is divided into two categories: one is built by experts based on personal experience knowledge base. This is relatively accurate, but the coverage is small, only suitable for specific areas of auxiliary diagnosis. The second is to use machine learning algorithms to extract data from hospitals and online to construct a more comprehensive knowledge graph. However, limited by the lack of data planning and the current state of the art, it falls short in terms of accuracy. In today's society, basically most of the companies providing assisted diagnostic services are the second type.
Based on the type of underlying data, auxiliary diagnostic models can be divided into: text diagnostic type, image diagnostic type. The implementation logic of the text type is, by and large, to use algorithms to obtain and generate structured data from medical literature, cases, and online medical information. The image diagnostic type is obtained from hospital image databases, as well as online development of image libraries.
Comprehensive market players doing text-assisted diagnostics include: Ping'an HaoDo, WeMed, and other online medical platforms, watson, Baidu Medical Brain, and KangFuZi, and other service providers. Image-assisted diagnosis and other players are: Ali ET medical brain, Tencent foraging, etu technology and other companies. As mentioned above, doctors in the diagnostic process is extremely dependent on the inspection or observation of the information obtained from the data, which are basically in the hospital, the availability is relatively poor. This leads to the current stage of text-assisted diagnosis model effect accuracy can not meet the requirements of doctors. Image-assisted diagnosis, due to the processing of business is relatively single, the main use of ct / mri data, at the same time, image recognition technology relative to NLP, more supportive of the diagnosis, reflected in the market results is that there are too many players on this track.
At this stage, the development of auxiliary diagnostic tools is not good, summarize the reasons are as follows:
1) the state has been planning the development of medical data at a high level schedule, and in terms of policy, the opening of medical data, gradually clearing the obstacles
2) hospitals, especially the Tertiary hospitals, in order to develop their own, but also quasi-different upgrade of the internal information systems, unified Data specification
3) The subsequent opening and use of data should be in the form of an alliance, as far as possible to get the ticket
3) Technology upgrades depend on the development of the entire technology field, before this can be combined with the company's resources, to choose a specific field to try to assist in the decision-making