[Industry News] AI drives deep changes in the medical imaging industry

In recent years, with the gradual application of artificial intelligence (AI) technology in the field of medical imaging, the continuous implementation of relevant policies to encourage the innovation and development of the medical device industry, and new medical imaging formats dedicated to improving the service level of medical institutions have emerged, especially AI medical imaging Diagnostic equipment represented by products has become a hot spot for the development of the medical device industry.

　　AI empowers the extension of the traditional industrial chain

　　 AI medical imaging equipment is a high-end medical device, with the characteristics of multi-disciplinary, knowledge-intensive, and high value-added. The medical imaging industry chain involves multiple industries such as basic industry, manufacturing, imaging, medical institutions, and the Internet. When the image data accumulates to a certain scale, the industry chain will extend to the AI field, and the development of intelligent image diagnosis applications will further promote the development of medical image diagnosis facilities and services.

　　 The upstream of the medical imaging industry chain is the chemical, metal, communications and other industries. The technological progress of these industries will promote the development or change of the medical imaging industry.

　　 The middle reaches of the industry chain are medical imaging diagnostic services and infrastructure industries, including medical imaging imaging equipment companies. At this stage, the midstream market has the largest scale.

　　 The downstream of the industrial chain involves medical institutions and derivative service institutions at all levels. Among them, public hospitals are the main customers of medical imaging equipment companies, and online imaging platforms and independent imaging centers are the main growth forces of the downstream market in the future.

　　 At present, China's high-quality medical resources are generally concentrated in tertiary hospitals, and the development of independent imaging centers is conducive to the realization of the integrated distribution of high-quality medical resources and the rationalization of the distribution of medical resources. Restricted by factors such as policies and costs, some primary medical institutions have limited ability to deploy large-scale medical imaging equipment, and some clinical needs cannot be met. Independent imaging centers can solve this problem well. On the one hand, independent imaging centers can reduce the burden on tertiary hospitals, on the other hand, they can also improve the service capabilities of primary medical institutions and further promote the rapid development of the medical imaging industry. At this stage, China's independent imaging centers are still in the initial stage of development. In the future, with the support of policies, the increasing demand for chronic disease management and the increase in the number of elderly people, the market will usher in a broad space for development. The independent imaging center belongs to the asset-heavy model and requires the purchase of a large number of medical imaging equipment, which is conducive to driving the development of the medical imaging equipment industry. This will become a key factor in the growth of the midstream market.

　　The online imaging platform provides services such as remote reading through the cloud platform, which satisfies the diverse needs of patients and clinicians. Relying on new technologies such as AI, cloud computing, and big data, online imaging platforms have grown rapidly in recent years, and consumer demand in downstream markets is strong. The addition of the new model extends the traditional industrial chain and also expands the overall scale of the industry.

　　 It is imperative to establish a standardized and large-sample database

　　Data is the core resource of AI medical imaging equipment. Only the algorithm is mastered, but the quantity and quality of the data are insufficient, and good training effects cannot be obtained. At present, factors such as blocked data acquisition channels, vague industry standards, and unclear data usage mechanisms have limited the development of the AI medical imaging industry to a certain extent. In order to better promote the healthy and rapid development of the industry, it is necessary to establish a series of effective solutions and establish a standardized and large-sample database through a reasonable data sharing mechanism.

　　 One is to unblock the channels for obtaining effective standard training data. Compared with other industries, the medical imaging industry has natural disadvantages in the acquisition of high-quality data: on the one hand, high-quality imaging data is concentrated in tertiary hospitals, and there is a lack of effective data exchange mechanisms between different medical institutions, and data is difficult to share; on the other hand, On the one hand, although the amount of medical data in China is huge, 80% of it is unstructured data, which limits the further application of AI in the medical imaging industry. In addition, the training data set should cover images of different scenes such as physical examination, screening, outpatient and laboratory, etc. according to the scope of application, and there is currently no effective database with standardized design in China. Therefore, smooth access to effective standard training data plays a key role in the high-quality development of the AI medical imaging industry.

　　 The second is to improve industry standards. The nature of AI technology determines the data set used for algorithm training and product testing, which is of great significance to the quality control and risk management of the entire life cycle of AI medical imaging equipment. On the basis of obtaining valid data, deep learning technology can combine prior knowledge to correctly train the model, and the training set needs to be labeled in advance. The data quality and scale of different institutions are uneven, and the lack of unified labeling and scanning technology, processing methods, standards and consensus can easily lead to product quality and safety risks and "unacceptable" phenomenon. Therefore, it is necessary to strengthen guidance and regulation through unified industry standards, such as strengthening the qualification identification of the team of labelers of training data sets, unifying the recognition of image signs, labeling methods, segmentation methods, quantification methods, etc., to avoid inconsistent standards in the actual application of products Wait for the situation.

　　 The third is to establish an effective data protection and supervision mechanism. Medical units are reluctant to open and share data, largely because of information security considerations. At present, the types of data that can be used openly, as well as the attribution and ethical issues of the data are still unclear. At the same time, the use of data lacks effective protection and supervision mechanisms. Therefore, it is necessary to improve relevant laws and regulations, clarify data usage specifications, and ensure data quality is standardized and traceable.

　　The industry is facing upgrading and reshaping

　　In the future, with the empowerment of AI technology, the medical imaging industry will further transform and upgrade. At the same time, with the development of AI technology and the maturity of data application mechanisms, industry concentration will gradually increase.

　　The upgrading effect of AI technology on the industry will be more significant

　　 At present, the amount of data in the medical industry is increasing rapidly, and the speed of AI medical imaging product technology optimization will be improved, which will further promote the upgrade of the medical imaging industry.

　　 AI technology can identify complex data and automatically make quantitative assessments, which can assist clinicians in diagnosis and help to form more accurate radiological assessment results. As far as the medical imaging field is concerned, based on the technology category, AI technology has derived two basic applications: one is data perception, that is, medical images are analyzed through image recognition technology to obtain effective information; the other is data training, that is, through deep learning massively The image data and clinical diagnosis data of the company continue to train the model to optimize its diagnostic ability. Therefore, compared with traditional medical imaging equipment, AI medical imaging equipment has obvious advantages, and doctors' demand for the use of AI medical imaging equipment continues to increase.

In addition, in the field of AI medical applications, medical imaging is one of the hottest areas with the highest investment amount, the most investment rounds, and the most mature applications. The capital market’s high recognition and strong support for the AI medical imaging industry will accelerate the maturity of related technologies. The implementation of the application scenarios will promote the transformation and upgrading of the medical imaging industry.

　　Industry concentration is gradually increasing

　　 At present, the competitive landscape of the AI medical imaging industry is fragmented, mainly due to the following aspects:

　　 One is data dispersion. As mentioned above, most of China's medical imaging data comes from hospitals, and the image data is scattered in different hospitals, which makes it difficult for AI medical imaging models to be effectively trained and affects the actual application effect.

　　 The second is the scattered disease. Although the underlying code can be reused, different types of diseases require different annotation data to train different models. Although there are also companies that focus on product research and development for multiple diseases, in general, different disease models have different characteristics, leading to a relatively scattered industry pattern in the early stages of industry development.

　　 The third is changing scenes and business models with diversified characteristics. Taking the medical image intelligent recognition technology as an example, the potential monetization methods include: selling to medical institutions as a separate software module, selling to medical institutions in combination with image archiving and communication systems (PACS), and cooperating with image equipment to form a software and hardware integration There are many ways to sell customized solutions to medical institutions, serve primary medical institutions through telemedicine, and directly serve patients through Internet medical treatment. The diversification of scenarios and business models has led to many subdivided tracks in the AI medical imaging industry, and companies compete on their respective tracks.

In the future, with the establishment of industry data integration and sharing mechanisms, the maturity of model training methods, the establishment of business models, and the approval of more and more products, first-mover companies with advantages in "scene + data + algorithm" will With the gradual establishment of technical and commercial barriers, the AI medical imaging industry will gradually move towards concentration.