In recent years, domestic high-end medical imaging diagnostic equipment has continuously emerged, allowing people to obtain clearer and richer medical and health imaging information. For example, an abdominal MRI examination can provide doctors with over 3000 images of different levels of examination. By using cutting-edge analysis technologies such as artificial intelligence, the images can also bring deeper diagnostic and therapeutic benefits to patients. In this context, traditional film not only struggles to carry massive amounts of multidimensional data, but also limits the further improvement of diagnostic and therapeutic capabilities. Not long ago, the nationwide medical insurance image cloud cross provincial access was officially launched, marking that the era of countless patients carrying film bags and seeking medical treatment is becoming history. How to seize the opportunity and promote the improvement of medical imaging quality? How to promote cloud based medical imaging to better serve people's health while film exits the historical stage? With the help of artificial intelligence technology, what disruptive scene applications will emerge in the digital world of images? Recently, the reporter of Science and Technology Daily interviewed Wang Zhenchang, an academician of the CAE Member and dean of the School of Medical Imaging of Capital Medical University, and asked him to feel the development of modern medical imaging in the process of digital and intelligent changes. The development of medical imaging technology faces significant opportunities and challenges. Reporter: In what aspects has medical imaging technology made breakthroughs so far? Wang Zhenchang: Currently, 75% -85% of core information in clinical diagnosis and treatment comes from medical imaging. With the development of detection technology, the spatial resolution of medical imaging equipment has jumped to the micrometer level, and the temporal resolution can reach the sub millisecond level. In 1895, R ö ntgen discovered X-rays and took the first X-ray photograph in human history, opening up the application of imaging technology in the field of medicine. After more than a hundred years of development, medical imaging has evolved beyond traditional "planar imaging" forms. For example, the well-known CT (computed tomography) scans the human body by rotating an X-ray beam around it, combined with computer reconstruction technology to generate three-dimensional tomographic images, which improves the accuracy of lesion localization; As MRI (magnetic resonance imaging) technology gradually matures, it becomes possible to clearly present subtle structures such as soft tissues and the nervous system; Nuclear medicine, such as PET-CT (positron emission computed tomography), accurately captures organ function and lesions by tracking the distribution of radioactive tracers, providing key support for early diagnosis of tumors and detection of metastatic lesions. These technologies can non invasively and high-definition present the structure and functional status of organs, which can not only provide strong support and evidence for diagnosis and treatment decisions, but also help doctors accurately identify diseases before symptoms appear. For example, the ear specific CT developed by our team has broken through the spatial resolution to 50 microns, achieving in vivo imaging of micro scale structures such as the stapes floor and vestibular window with a diameter of only 1.5 millimeters for the first time. It can also accurately quantify indicators such as the activity amplitude and bone density of the ossicular chain. Currently, information technologies such as artificial intelligence and big data have made image analysis more in-depth and comprehensive. Artificial intelligence assisted diagnostic systems achieve rapid screening and quantitative analysis of lesions, while multimodal image fusion technology integrates different modal image information obtained from various devices to provide comprehensive diagnosis and treatment basis for difficult diseases. Reporter: Besides preventing patients from carrying film around, what are the advantages of electronic imaging? Wang Zhenchang: Electronic imaging has obvious advantages for both doctors and patients. Traditional film can only display some key tomographic images, while raw electronic images contain thousands of tomographic images. Therefore, film cannot fully present the spatial relationships and detailed features of lesions. In addition, the inability to establish an electronic traceability system for film can sometimes lead to the problem of repeated inspections, resulting in annual inspection costs of over 10 billion yuan nationwide due to film issues. The application of electronic imaging can effectively solve these problems. With the comprehensive implementation of unified data formats and interface standards, the output and storage of standardized electronic images, as well as the degree of integration and interoperability of medical images, have further deepened; The application of various lightweight image browsing software allows doctors to access images anytime and anywhere on mobile phones, computers, tablets, and other terminals, and can enlarge, rotate, measure, and compare images. The patient authorizes the doctor to access electronic images as easily as handing over the film to the doctor. For patients, electronic imaging saves the material cost of film, and with the advancement of mutual recognition and sharing of examination results, some unnecessary repetitive examinations can also be reduced. This saves patients medical expenses and time, and also avoids the inconvenience caused by repeated blood draws, irradiation, etc. Technological innovation makes image storage and retrieval more efficient. Reporter: What challenges do we face in the storage and retrieval of medical electronic images? Wang Zhenchang: We face significant challenges in terms of data storage costs, call efficiency, and other aspects. The application of medical electronic imaging means massive amounts of data. We know that photos take up more storage space than text, and medical imaging not only includes regular format images, but also continuous scans of hundreds or thousands of frames. Taking the current photon counting CT as an example, the single image data volume can reach 10GB-20GB, and the average annual image data increment of a tertiary hospital exceeds 100TB, which creates a huge demand for data storage space. Therefore, hospitals need to invest a large amount of funds to solve the problem of massive data storage. Common solutions include building your own data center and purchasing cloud services. According to the requirement of retaining outpatient images for 15 years and inpatient images for 30 years, the cumulative image data volume of a tertiary hospital over 30 years will exceed 15PB. The hardware procurement and maintenance costs of traditional centralized storage are extremely high, which grassroots hospitals often find difficult to afford. How to build a good information infrastructure and achieve data intensive management still requires exploration and practice in various regions. Data should not only be stored properly, but also easily accessible at any time. But it is not easy to call target information in massive databases. This is like searching for a specific image in a phone's photo album, and flipping through them one by one can be time-consuming. Therefore, it is necessary to use tagging technology to make the storage path of data conform to the characteristics of medical imaging, so that doctors no longer have to "rummage through boxes and cabinets", but can easily access it through a convenient path. In addition, compatibility and interoperability at the data level are also prerequisites for improving the efficiency of medical imaging utilization. Currently, there are significant differences in data storage formats among different devices and medical institutions, leading to difficulties in integrating data from different sources and a scarcity of high-quality annotated samples. Therefore, it is necessary to lay the foundation for the application of medical imaging data through quality control and standardization construction. Reporter: What technological innovations have we made to achieve efficient storage and retrieval of medical electronic images? Wang Zhenchang: Obtaining an abdominal image through regular network transmission may take more than ten minutes. Even if the call is successful, due to poor format compatibility and inconsistent image formats and encoding methods among devices from different manufacturers, there may still be situations where the image data of primary hospitals cannot be opened or viewed clearly in the system of tertiary hospitals. To solve these problems, continuous technological innovation is required. For example, the "second level transmission" of images can be achieved through 5G+edge computing to improve the transmission speed. Currently, relevant departments have innovatively adopted gigabit private networks and intelligent preloading technology when deploying medical insurance imaging cloud work. When patients register, the hospital system can send preloading requests based on medical insurance information, and the called images enter the hospital's deployed "front-end machine" through a private network, greatly reducing the time for doctors to access cross regional images. To achieve efficient retrieval, the National Healthcare Security Administration has launched a 78 digit national unified image code, assigning a unique "ID card" to each image. The 'ID card' contains basic patient information as well as key information such as examination institution, equipment type, and examination time, solving the problem of cross institutional data being 'not found'. At the same time, relevant departments have established a standardized metadata system to support rapid retrieval based on multiple dimensions such as patient ID, examination site, and disease type, resulting in a retrieval efficiency improvement of over 80%. How is the application of artificial intelligence technology in the field of medical imaging? Wang Zhenchang: Currently, artificial intelligence is being increasingly applied in the field of medical imaging. In some artificial intelligence competitions, artificial intelligence has demonstrated its ability to solve imaging diagnosis problems, sometimes even surpassing the accuracy and diagnostic speed of professional clinical doctors. According to the data, as of June 2024, 103 Class III medical device registration certificates have been approved in the field of artificial intelligence medical imaging, covering diseases such as cardiovascular disease, pulmonary disease, cerebrovascular disease, orthopedic examination, fundus disease, and breast disease. However, the application and promotion of artificial intelligence in the field of medical imaging still face obstacles. A nationwide survey conducted on clinical doctors, radiologists, and radiologists shows that 50% of respondents believe that artificial intelligence is difficult to embed well into current medical diagnosis and treatment processes. Reporter: In clinical trial verification, artificial intelligence has proven to be fast and accurate in calculation. Why do some respondents not have confidence in its application? Wang Zhenchang: The current artificial intelligence medical imaging equipment is developed based on single disease, but in clinical applications, medical imaging is not focused on single disease. The medical report issued by a radiologist is an overall evaluation of the patient's organs, rather than an evaluation of individual diseases. For example, conducting a lung examination not only involves identifying lung nodules, but also assessing emphysema to see if there is atelectasis or pneumonia. However, current artificial intelligence devices only evaluate lung nodules, which does not meet the requirements for clinical application or radiology reporting. In addition, current artificial intelligence training data is based on image information and annotated through deep learning. However, artificial intelligence's interpretation of images is difficult to reach a professional level, and its interpretation of the elements presented in images is incomplete, losing a lot of information, which makes it appear "useless" in practical applications. Therefore, in most clinical cases, current artificial intelligence applications have not effectively improved doctors' work efficiency, but have instead brought confusion and increased labor intensity to doctors. Reporter: So, how can we promote the integration of artificial intelligence and medical imaging to better meet clinical application requirements? Wang Zhenchang: Firstly, it is necessary to achieve standardization, normalization, and quality control in data collection and quality control. If the results of the same examination are different between tertiary hospitals and primary hospitals, and the performance of the same equipment varies greatly due to the different levels of operators, it cannot be considered a shared application. To address this issue, Beijing Friendship Hospital affiliated with Capital Medical University, as the supporting unit of Beijing Medical Imaging Quality Control and Improvement Center, has taken the lead in drafting multiple national health industry standards in the construction of medical imaging standard system, providing authoritative basis for the standardized operation of imaging inspection in medical institutions at all levels. The unification of measurement has promoted the exchange of goods and services between different regions, and the same goes for image data. The standardization has promoted the overall improvement of medical image quality and data security sharing. At present, Beijing has established the "Technical Requirements for Shared Data Transmission on Medical Imaging Cloud Platforms" and the "Quality Control Standards for Medical Imaging Examination Reports". Based on the data gathered by the Beijing Medical Imaging Cloud Platform, the Beijing Medical Imaging Quality Control and Improvement Center has compiled image quality control indicators applicable to the cloud platform and released monthly reports to 253 medical institutions, achieving dynamic monitoring and continuous improvement of image quality in medical institutions within the Beijing region. On the basis of data exchange and sharing, the entire chain of medical imaging can be interconnected