The acceleration of the aging process is causing profound changes in the disease spectrum in China. Relevant data show that in 2020, the number of new cancers, cancer incidence and mortality in the mainland will rank first in the world. In the next 5-10 years, the number of cancer patients in mainland China will increase significantly.
Screening and early diagnosis and early treatment are one of the effective ways to prevent and control cancer. The key to "early diagnosis" is to expand the accessibility of pathological diagnosis, improve the accuracy of pathological diagnosis, and improve the level of pathological diagnosis in primary medical institutions.
Pathological diagnosis is the "gold standard" for disease diagnosis, but it takes a long time to train a qualified pathologist. Data show that by the end of 2021, there will be 25,000 registered pathologists in mainland China, of which less than 8,000 doctors can independently complete pathological diagnosis. From graduation to independence, pathologists often need "ten years to sharpen a sword". Therefore, it is difficult to fill the vacancy of more than 100,000 pathologists in the short term.
In terms of regional distribution, 80% of pathologists are concentrated in large cities and 70% in tertiary hospitals. On the one hand, the pathology departments of tertiary hospitals are overloaded and doctors are overtired, which has become the norm. Taking the pathology department of tertiary hospitals as an example, it needs to undertake more than 3,500 pathological diagnosis tasks every day, and the total number of annual slices exceeds 1 million.
On the other hand, the resources for pathological diagnosis at the grassroots level are seriously lacking, and the diagnostic compliance rate is low. At present, the pathological diagnosis compliance rate of secondary hospitals is 35%, and that of county-level hospitals is only 26%. Insufficient "gold content" in pathological diagnosis is the main reason for overtreatment or undertreatment of tumors and other diseases.
The pathology community in mainland China has realized that promoting digital pathology is the key way to solve the imbalance between supply and demand of pathological resources: converting glass slices into digital sections, and then digital storage and digital diagnosis are conducive to remote pathology, artificial intelligence-assisted diagnosis, and pathology big data research. In order to improve the diagnosis efficiency, reduce missed diagnosis and misdiagnosis, and solve the problems of uneven distribution of pathological resources.
Looking back on the past ten years, the digitalization process of pathology department lags behind that of laboratory department and imaging department. The reason is that the key technology of digital pathology needs to make breakthroughs. These key technologies include: fully digital pathological information system, digital pathological storage compression algorithm, multimodal intelligent pathological auxiliary diagnosis algorithm, etc.
Take the data storage system as an example, it is the basic IT foundation for providing digital slice reading services. Since the scanned digital slides are compressed based on traditional algorithms, the slice file is generally around 1-3 GB. -2PB orders of magnitude are increasing, need to be stored for 15-30 years, storage costs are rising rapidly.
What's more, object storage used by traditional storage solutions has a high protocol overhead. The retrieval speed of pathological sections is slow, and problems such as delay and splicing are prone to occur. It even takes tens of seconds from dragging the image to the normal display. Affects the efficiency and accuracy of pathologist reading. At the same time, the large file size leads to slow network transmission speed, making it difficult to carry out services that require high real-time, such as remote intraoperative freezing pathological diagnosis.
The construction of "digital intelligent pathology" cannot be achieved overnight, and it needs to go through three stages: informatization, digitalization and intelligence.
The first step is informatization, that is, a comprehensive digital upgrade of the traditional pathological workflow on the basis of the information management system.
The second step is digitalization, that is, based on digital sections, to realize pathological diagnosis and related digital pathological applications. The key issues to be addressed are the freezing of digital slice reading, and the high cost of storage.
The third step is intelligence, which is based on the organic integration of artificial intelligence and pathological workflow, and now efficient artificial intelligence-assisted diagnosis. At the same time, artificial intelligence is used to deeply study and mine pathological data to discover new disease characteristics and treatments. Development and innovation offer new ideas and approaches.
Based on the "three-step" path blueprint, the "Digital Smart Pathology Department" can completely digitize and intelligentize all pathological information generated and processed by the department on a daily basis, and realize the digital construction of the overall pathological ecosystem.
"To build a digital intelligent pathology department, pathology departments or IT teams alone cannot do it. This can only be achieved by working together as a team to address the key pain points of digital pathology. To this end, the key tasks are broken down and divided into four key tasks: digitization of slide reading, department management informatization, efficient data management, intelligent diagnosis and quality control, requiring pathology research, hardware support, information management system, storage and other teams to participate in compression algorithms and artificial intelligence technology.
Improve diagnostic efficiency. In the conventional biopsy diagnosis with large number, low difficulty and high reproducibility, digital image reading can effectively reduce the repeated work of doctors and improve the diagnosis efficiency. Take a small specimen from endoscopic biopsy of the digestive tract as an example, traditional microscope reading.
The reporting period was 3 working days, which the hospital has now shortened to 2 working days. In terms of solving difficult problems, with the help of the deep learning ability of AI technology, the AI algorithm has high sensitivity and specificity in judging small lesions in digestive tube biopsy.
Improve management efficiency. The digital quality control of the whole process has improved the specimen circulation efficiency, production quality and production efficiency of the pathology department to varying degrees.
Authorized clinical research, pathology informatization, digital construction to improve the quality of pathological data. A digital intelligent pathology auxiliary service platform has been established, covering multiple pathology AI technology fields, establishing a subspecialty database, and expanding the AI model of prostate cancer, lymphoma, lung cancer and other subspecialty diseases, laying the foundation for the "next generation of pathology technology".
The use of full-process automation and digital radiation to branch hospitals and medical institutions in the medical alliance provides reference for the construction of national/provincial regional medical centers, drives grassroots hospitals to improve the level of pathological diagnosis, further improves the accessibility of ordinary patients to obtain high-quality pathological diagnosis resources, and realizes that the hospital is next door to you and the pathology department is by your side.
Three major innovations have brought about a scientific and technological revolution in pathological data management
The construction of the "digital intelligent pathology department" needs to solve the problems of analysis efficiency, reading experience and storage cost of massive digital slides. A technological revolution in pathological data management is taking place.
Adhering to the tenet of "finding technology for scenes, finding scenarios for technology", the data storage product line deeply integrates digital technology with pathology business scenarios, and launches the first "digital pathology storage system" that supports hospital-grade reading and pathological losslessness. The compression algorithm has achieved a qualitative leap in the high-performance acquisition and data reduction of pathological data.
The first is "fast reading"
The overhead and access latency of the object storage protocol are greater than that of the file storage protocol, and it is not the best pathological storage protocol. Data storage independently developed "distributed parallel file client (DPC)" technology, which runs distributed parallel client as a storage client on the computing node, exchanges data with the back-end storage node through a high-speed network, and adopts intelligent algorithms to greatly reduce access latency, improve throughput, and enable upper-layer applications to access storage space more intelligently. This solution eliminates the need to upgrade the network and read clients, effectively protecting existing IT investments.
By introducing DPC technology into the reading of pathological digital slides, the system can automatically match layers according to browsing needs when reading slides. This method solves the problem of large object storage protocol overhead, achieves 80 times better retrieval performance, and realizes a smooth experience of viewing 1000 pathological slides per second.
From the feedback of pathologists, it is really "look wherever you want, no need to wait, very smooth", the same slice supports multiple people to read the slide at the same time, supporting all scenarios including clinical, scientific research, teaching and so on. Without the need for hospital-level stop films, the collaborative efficiency of pathological diagnosis can be increased by more than 70%.
This is followed by "low storage costs"
The average size of traditional pathology images after compression is still close to 1GB, and the correlation between the features and "tiles" of pathology images cannot be fully utilized. On the basis of fully studying the characteristics of pathological data, the "secondary lossless compression algorithm of pathological data" is innovatively proposed, which greatly improves the compression rate of pathological sections through intelligent recognition algorithm, semantic segmentation, reference compression and other technologies.
Different from the "general compression technology" that can only compress TXT text files and the "repeated deduplication compression technology" that compresses data blocks, the essence of this "sick secondary lossless compression algorithm" is accurate scene compression based on sick images. On the basis of the original compression algorithm, it can save 30% of storage space and network bandwidth. In the future, as algorithms continue to evolve, they will bring greater space savings.
In addition, pathological data have obvious periodic access characteristics. Classification is based on data access frequency and hot, warm, and cold, which can not only take into account the performance requirements of image reading, but also effectively reduce storage costs. Tiered storage technology allows different types of physical nodes in the same storage pool to be divided into different hard disk pools, realizing tiered storage and management strategies for hot, warm, and cold data. Taking the Blu-ray storage of cold data as an example, it can help hospitals reduce the cost of Blu-ray storage of 30-year pathological data by more than 57%.
The third is "wide coverage"
For the primary medical institutions that need digital pathology empowerment the most, the deployment model launches the "digital pathology all-in-one machine".
Based on the integrated function of FusionCube 500, the all-in-one machine integrates six functions of scanning, storage, computing, network, security and artificial intelligence, and integrates multiple systems such as data acquisition system, pathological reading system, pathological information system, artificial intelligence-assisted diagnosis system, and remote pathological system. For the grassroots level, only one set of equipment is needed to obtain the functions that can be realized by the traditional six sets of equipment, so as to realize the digitization of the primary pathology department at the fastest speed and the lowest cost, and provide a basis for the sinking of high-quality pathological diagnosis resources.
If a worker wants to do a good job, he must first sharpen his tools. What the "digital intelligent pathology" of the future will look like? "Towards the Next Generation of Pathological Diagnosis Theory" not only relies on digital slides, but also relies on the fusion application and analysis of genetic data, imaging information, clinical information and other data. With the deepening of intelligent pathology research, the technological revolution of pathological data storage and management continues to continue, and the trend of application integration, data fusion and storage integration has become increasingly clear.