Recently, the research paper “Deep Learning Segmentation and Reconstruction for CT of Chronic Total Coronary Occlusion”, which was jointly completed by United Imaging and Shanghai First People’s Hospital, was published in the journal Radiology.
“Radiology” is a top journal in the field of medical imaging, and its impact factor will reach 29.1 in 2022. The fact that the research results can be included in “Radiology” means that the results have been highly recognized by the radiology community.
The Shanghai First People’s Hospital (“Shanghai No. 1” for short), which participated in this cooperative research, was one of the first hospitals in China to establish a medical imaging department, and opened a separate radiology department in 1946. Artificial intelligence + cardiothoracic imaging is the key direction of scientific research in a radiology department in Shanghai. Led by Zhang Jiayin, director of the radiology department of the hospital, the research on AI automatic reconstruction of coronary artery CTO is one of the scientific research projects in the cardiovascular field of the hospital.
Cardiovascular has always been an important field of medical imaging AI layout, but coronary CTO reconstruction is difficult, and no AI products with a high success rate have appeared yet. According to reports, the algorithm jointly researched by United Imaging and Shanghai Yiyi can achieve a success rate of 95% for the automatic reconstruction of coronary CTO.
What is so special about the algorithm used in this study? How to solve the problem of automatic coronary CTO reconstruction? What other research directions are worth investing in in the field of cardiovascular AI? Focusing on these issues, Leifeng.com’s “Medical and Health AI Nuggets” had a dialogue with Zhang Jiayin, director of a radiology department in Shanghai.
Zhang Jiayin, Director of the Department of Radiology, Shanghai First People’s Hospital
The “secret” of high success rate of coronary CTO reconstruction
Coronary CTO (Chronic Total Occlusion), that is, coronary chronic total occlusion, is one of the biggest problems in interventional treatment of coronary heart disease.
According to the “China Cardiovascular Health and Disease Report 2021”, there are more than 11.39 million patients with coronary heart disease in my country, and the prevalence and mortality of coronary heart disease are increasing year by year. Among patients receiving invasive coronary angiography (Invasive Coronary Angiography, ICA), about 20%-25% of the cases were diagnosed as coronary CTO lesions.
In clinical practice, coronary CT angiography (coronary CTA) is currently mostly used as a non-invasive method for assessing coronary artery disease, but for the identification and reconstruction of coronary CTO, it still needs to rely on doctors to follow up with manual drawing based on experience.
As early as 2013-2014, Zhang Jiayin had paid close attention to the complex lesion of coronary artery CTO, and published three papers related to coronary artery CTO lesion in the journal “Radiology”, explaining the imaging identification of coronary artery CTO Diagnose, guide therapy, and evaluate collateral anatomy, among other issues.
Zhang Jiayin introduced that without the use of ICA, CTO lesions are often not well identified on CT imaging. The doctor can only use the post-processing workstation to simulate the growth of blood vessels, manually draw the occlusion segment and the collateral vessels distal to the occlusion, and then perform necessary measurements, including length, calcification load, and negative remodeling.
“Traditional identification of coronary CTO requires huge clinical manpower and time costs.” Zhang Jiayin said that under normal circumstances, it takes doctors more than ten minutes to complete a case of coronary CTO reconstruction.
Based on this pain point, Zhang Jiayin and United Imaging’s intelligent team cooperated in the first half of last year to carry out the automatic reconstruction of coronary artery CTO AI based on deep learning algorithm. Although it took less than a year and a half from the establishment of the project to the submission of the manuscript, this research is actually quite challenging.
Due to the weak or poor visualization of the total occlusion of the coronary artery, it is easy to cause segmentation and incomplete rupture of blood vessels during the reconstruction process, which places high demands on the algorithm.
To this end, the researchers integrated the anatomical structure of the coronary arteries and the heart into the algorithm to provide contextual information for the coronary artery segmentation. At the same time, the local and overall features of the coronary arteries were combined to ensure that the coronary artery segmentation, especially the image development, is not clear. The completeness, accuracy and continuity of the CTO location segmentation.
In the end, the two teams conducted reconstruction experiments on a total of 240 cases of CTO lesion vessels in 211 patients. Experimental results show that the automatic segmentation and reconstruction success rate of the algorithm is 95%, while the success rate of traditional workstations is only 48%. In terms of post-processing time, with the help of this algorithm, the reconstruction time can be shortened by up to 80%, and the result can be obtained in 2 to 3 minutes.
At present, the results of this research have been put into clinical use in a radiology department in Shanghai. Zhang Jiayin said that in practical applications, it is found that AI has a good reconstruction effect on small branches less than 1mm and 1mm distal to the distal end of the coronary CTO, and can almost automatically identify, segment and label the entire occlusion segment, as well as the lateral side of the distal end of the occlusion. Blood vessels, “The image effect is even better than that produced by human hands.”
Zhang Jiayin believes that the follow-up of this study can greatly promote the accurate evaluation of preoperative images of clinical coronary CTO. He hopes that in the future, this coronary reconstruction algorithm can be extended from the radiology department to the cardiac catheterization room, so as to help the cardiology department to better analyze and make decisions about diseases.
AI still has a lot to offer in the cardiovascular field
Zhang Jiayin is an excellent radiologist with both clinical diagnostic ability and scientific research ability. His path of study and career also closely followed the development of cardiovascular imaging, hoping to summarize and summarize his clinical experience And transform it into evidence-based medical evidence, and use practical actions to inject its own strength into the field of cardiovascular imaging diagnosis, so as to benefit more patients.
In 2006, Zhang Jiayin received a master’s degree from Shanghai Jiaotong University School of Medicine. After graduation, Zhang Jiayin started his career in the Radiology Department of the Sixth People’s Hospital Affiliated to Shanghai Jiaotong University (referred to as “Shanghai Sixth Hospital”). During his doctoral period, he studied under Professor Li Minghua, the then director of the Radiology Department.
Professor Li Minghua is the academic leader of the Department of Interventional Imaging, specializing in neuroimaging diagnosis and neurointerventional therapy technology, and is a pioneer in non-invasive imaging and minimally invasive interventional therapy of cerebral aneurysms in China.
Not long after practicing, Zhang Jiayin encountered an important technological turning point in the medical field. In 2007, just as 64-slice CT became popular in China, the CT imaging technology in the cardiovascular field made a breakthrough, making it possible for CT to be used in routine clinical diagnosis of coronary heart disease.
Before the popularization of 64-slice CT, due to the limitation of inspection equipment, although the incidence of coronary heart disease was high and the harm was great, the use of CT to diagnose coronary heart disease was almost a field that no one cared about in the radiology department.
Zhang Jiayin is very interested in cardiovascular research, and he applied to his supervisor, Professor Li Minghua, to engage in research in this direction. Although this is not Professor Li Minghua’s field of specialization, he chose to fully support Zhang Jiayin’s choice in the face of the enthusiasm of the students.
As a scholar who entered this direction earlier, Zhang Jiayin has witnessed the rise of the cardiovascular field so far. Up to now, he has published 6 cardiovascular-related research papers in Radiology.
Around 2017, with the establishment of major domestic medical imaging AI manufacturers, since then, the introduction of AI has become a topic of great concern to the radiology departments of major hospitals.
Zhang Jiayin’s association with medical imaging AI also began at this stage. When he was in the Sixth Hospital of Shanghai, Zhang Jiayin felt the utility of AI from a rib fracture AI product produced by United Imaging, and has since held an open and embracing attitude towards AI.
In January 2021, Zhang Jiayin joined Shanghai First People’s Hospital as the director of the Radiology Department of the hospital. He officially led the national-level project based on imaging diagnosis of heart disease and the development of new AI technologies, and started cooperative research with United Imaging.
Zhang Jiayin believes that in a series of fields that require a lot of manual intervention by doctors, such as the detection of cardiovascular, head and neck blood vessels, and rib fractures, AI is just needed.
The first three-category AI medical certificate in China falls in the cardiovascular field. So far, in the field of cardiovascular, especially in the direction of coronary artery, many medical imaging AI companies have entered the market.
According to the “China Medical Imaging Artificial Intelligence Development Report (2021-2022)”, as of May 31, 2022, among the products that have obtained NMPA Class III certificates, the number of cardiovascular products is second only to pulmonary nodules.
There are many entrants, but in Zhang Jiayin’s view, there are still many directions to be tackled in the research and development of cardiovascular AI.
The first is anatomical orientation. The AI automatic reconstruction of coronary artery CTO in cooperation with United Imaging has effectively solved the problem of identification and reconstruction of completely occluded vessels. In terms of structure identification, there are still great technical challenges, and many AI products cannot be applied to this type of structure.
Zhang Jiayin revealed that in these more challenging research areas, Shanghai Shiyi and United Imaging have achieved good recognition and segmentation results, and will disclose the latest research results soon.
The second is the direction of functional imaging. Since the coronary plaque is very small, it is not easy to be automatically segmented, and manual post-processing is difficult, so the automatic analysis of coronary plaque components also needs to be broken through; The research is also very important. At present, Shanghai Yiyi is conducting a large-sample, multi-center clinical verification based on the products of United Imaging.
In addition to the above-mentioned functional evaluation on the vascular level, there is also an evaluation on the myocardial level.
Zhang Jiayin said that in clinical practice, the myocardial perfusion imaging process is very cumbersome and involves multiple steps. It takes 20 to 30 minutes to complete the examination and imaging analysis of a patient, and the clinical pain points are very prominent. The cooperative research between Shanghai Yi and United Imaging also covers this link. At present, the fully automatic quantitative analysis software for CT myocardial perfusion developed by the two parties has been put into clinical practice, which can automatically complete the diagnosis process in a one-stop manner.
To sum up, Shanghai Shiyi and United Imaging have carried out cooperation projects from anatomy to functional science, covering the difficult areas of cardiac CT imaging examination in multiple directions. And such a full-stack and full-spectrum medical AI is the future development trend.
As Professor Liu Shiyuan, chairman of the Radiology Branch of the Chinese Medical Association and director of the Radiology Diagnostic Department of Shanghai Changzheng Hospital, mentioned at the 3rd China Medical Imaging AI Conference, AI medical imaging companies should speed up the construction of a vertical and horizontal business layout, based on clinical Scenarios solve multi-task and multi-dimensional problems, and output overall inspection results based on parts and organs.
Judging from the overall trend of the industry, medical imaging AI is developing in a full-process, multi-task, and multi-dimensional direction, and the horizontal and vertical layouts will become more and more in-depth. Leifeng.com Leifeng.com
This article is reproduced from: https://www.leiphone.com/category/healthhai/0XvSPFJp6NAglfjX.html
This site is only for collection, and the copyright belongs to the original author.