Domestic first AI "image-assisted diagnosis" software three types of certificate issued, MR brain tumor first out of the heavy siege

Guide: "The value of medical AI is essentially to solve clinical needs, how in the future layout to clinical needs as a starting point, the use of AI technology enabling medical, is our foothold in the industry, but also the next always need to stick to the initial heart"the epidemic test, the first AI image assisted diagnostic field of three types of evidence recently ushered in a breakthroughOn June 12, the State Drug Administration announced that The Intracranial Tumor Magnetic Resonance Imaging Diagnostic Senod Diagnostic Software of Ander's BioMind "Tian medical wisdom" had passed the approval of NMPA three types of medical devicesthe product was developed by the "Neuropathic Artificial Intelligence Research Center" jointly established by Tiantan Hospital and Ander Medical Intelligence, andis the first medical AI software approved by the Fda to be named "Image Assisted Diagnostics"it is understood that Ander Medical Technology Co., Ltdis headquartered in Beijing, and set up a wholly-owned subsidiary in Singapore and LuxembourgThe approval makes it the first of three types of evidence approved in China in the field of AI imaging assisted diagnostics, based on deep learning technologythe market has been silent for a long time, why can Ander medical wisdom at this time break the situation? What inspiration does the approval of intracranial tumor AI products bring to the industry?What exactly arelayered products?this approved product is BioMind "Tian medical wisdom" in the MR intracranial tumor AI-assisted diagnostic software, it is for hundreds of thousands of pathologically verified brain tumor data, the use of deep learning algorithm, can achieve intracranial tumor (such as meningococreoma, hearing neuroma, myeloma, glioma, etc.) of the accurate diagnosis of artificial intelligence, diagnosis accuracy of more than 90%, some diseases diagnosis accuracy is more than 96%A structured report, including accurate information such as tumor location and volume, is automatically generated to help doctors diagnose quickly and improve radiologists' ability to diagnose brain tumorstalking about developing intracranial tumors, Wu Zhenzhou, chief technology officer of Anderin Medical, recalled the decision at the time: "It was mainly due to the market size and technical complexity." Because MR is a multi-sequence image, the analysis is difficult, coupled with the intracranial situation is more complex, the domestic study of MR intracranial tumor enterprises are few, this is our opportunity"
from the market point of view, urbanization, industrialization, aging and globalization process, as well as ecological environment deterioration, lifestyle changes, biological and genetic factors, the frequency of risk factorexposure and water average of malignant tumors continue to grow, the incidence and mortality of malignant tumors in the world are on the rise, seriously threatening human health and socio-economic developmentthe image of the central nerve tumor has "same-phantom disease, the same disease and other appearance", so we can not rely solely on the image signs to infer, must be based on the patient's medical history, clinical symptoms, signs and other related auxiliary examination This means that a qualified medical imaging worker, in addition to mastering the disease's video diagnosis knowledge, but also to understand and master the cause of the disease, pathology, clinical symptoms, signs, laboratory test indicators, and even treatment principles, such as a series of clinical knowledge, and to form a comprehensive analysis of morphology and functional indicators, multi-factor trade-off, comprehensive judgment of the logical thinking, in order to gradually improve the accuracy of brain tumor diagnosis, comprehensive Such accumulation and learning process, resulting in excellent image workers training time cycle is very long, the growth of the number of imaging physicians is far from meeting the growing intensity of diagnostic work , therefore, Ander's medical wisdom to solve, is actually the lack of medical resources Through AI-assisted diagnosis, Ander medical wisdom can empower hospitals and doctors to help solve the serious imbalance in the level of diagnosis and treatment As recently, Ander Medical Zhi as CCTV "dialogue" the only invited to the arrival of medical AI business representatives, passed to the people of the country's social mission: medical artificial intelligence can be used as a high-quality medical resources sinking tools, improve the level of clinical doctors in primary hospitals, so that ordinary people can enjoy a high level of medical services at home, which is the real significance of medical artificial intelligence "The 'intelligence' and 'can' of medical ai-artificial intelligence products should have a special connotation that distinguishes them from Other Areas of AI, " said Li Jingxuan, CEO of Ander medical and zhi China, 'Intelligence' should represent the wisdom and clinical experience of top hospital experts The input determines the output, which is the basis 'Can' should represent the able clinicians, especially in primary hospitalclinicians can improve the level of diagnosis and treatment BioMind"s theuse of Tiantan Hospital, using deep learning algorithm models, has systematically trained more than 10 million cases of tumor imaging in the past ten years, and has incorporated the clinical experience of top hospital experts to the whole disease in addition to the above-approved AI-assisted diagnostic software, BioMind "day medicine wisdom" series of products also cover the head, neck, heart, blood vessels, breasts and other multi-organ, multi-disease accurate diagnosis, is committed to become a whole-body CT, MR imaging AI-assisted diagnostic products, to achieve AI-assisted diagnostic early warning, risk assessment and other functions, can deeply participate in the hospital core medical services It is reported that the current BioMind "day medicine" other series of products are also in clinical trials and three types of registration declaration "based on single-site, single-disease and other single-scenario AI application, clinical application value is limited, but the transition of medical AI development The research and development of 'Day Medical Wisdom' starts from the neurological disease of the head, to the heart, breast to cardiovascular and other chest and abdominal diseases, to achieve multi-site, multi-disease AI application, is the overall direction of our future product research and development In an interview with the media, Ms Li said this systematic, process-based product form, on the one hand, to common diseases, multiple morbidity, acute and serious real application scenarios as a foothold, really meet the needs of the doctor's daily work, on the other hand, not only to assist doctors on the qualitative diagnosis of diseases, on the other hand, to achieve the future development of the disease accurate risk assessment and clinical assistance decision-making "It's only when we turn our attention to clinical needs, applications and values that we really begin to understand AI," paul Chang, president of radiology at the University of Chicago Medical Center, said in a 2019 RSNA In this regard, Li said, BioMind "day medicine wisdom" to risk assessment as the core, process optimization features, can be in radiology imaging, emergency, in-God, outside God, neuro-intervention, intra-cardiac, breast surgery and other hospitals in a number of departments to meet different clinical needs perhaps, BioMind's constant integration of AI in the update iteration is the best practice of the aforementioned nature of AI high standards of artificial intelligence, what needs to be solved? , why did Ander's intracranial tumor AI diagnostic product simply break through the re-enactment through the approval, the artery network of the interview content was collated, trying to find out the logic behind the approval "In the treatment of medical AI approval, the State Drug Administration Registration Department and the relevant personnel of the Center for The Trial of The Professional and Efficient service, can be said to be impressive." Li jingle told Arterial Net in fact, many AI medical products have been delayed in passing approval, most likely stuck in clinical trials It is reported that many medical AI enterprise products have not carried out effective clinical trials, has not entered the approval process is precisely because of clinical trials, products can not reflect non-poor efficacy and efficacy in evaluating the efficacy of clinical trials, non-toxic tests are tests to test whether one drug is inferior to another, and is used in clinical studies with objective efficacy indicators, such as clinical endpoints of antimicrobial drugs, adverse events in cardiovascular treatment, death or progress in tumor treatment, etc The superiority test is an experiment to test whether one drug is superior to another, and is generally used in a placebo-controlled trial in the approval of medical artificial intelligence devices, the definition of non-poor efficacy and superiority is different, but God is similar An AI product must prove that it has the value of application or has an advantage over an existing product if it is to prove that it has the advantage of being effective or non-effective As a result, clinical trial design has undoubtedly become a key part of approval Another problem is the source of the data In the stage of the rise of artificial intelligence, enterprises through various channels to obtain medical data and used for AI learning, but with the "deep learning auxiliary decision-making medical device software review points", "medical device production quality management specificationappendix independent software" release, the gradual standardization of network security, enterprises must explain the effective source of training data to the center of examination and approval, which led to some enterprises have to retrain AI products this way, the successful approval of BioMind "Tian medical wisdom" intracranial tumor MR-assisted diagnostic software, Ander medical wisdom gave a good practical answer - not limited to the early screening of the disease, but can make further accurate diagnosis of the disease, graded early warning, risk assessment, so that it has a real participation in the clinical value of diagnosis and decision-making to recognize the situation, in the way can find the problem
then, for the above mentioned many problems, how should the enterprise solve? How will the future develop? From Ander's experience, we can sum up three paths , find the right application for a long time, we have been calling a weAI in medical applications "AI-medical", the definition of which seems to be to see AI as a theme to promote the development of healthcare But in practice, as more and more scenes start using artificial intelligence, but not entirely dependent on artificial intelligence, people gradually find that the scene determines the application of AI, AI empowers the existing scenario - "medical s-AI" is the right way back to imaging, many of the existing imaging devices - CT, MRI, Color Super, Electrocardiogram, Electroencephalograms, etc - are more or less artificial But for artificial intelligence to really work, companies must not fall into the trap of "one function equals one product" example, if a patient has a fever headache, the doctor can't actually judge the patient's condition After the patient has done MRI, if only a single function of the product, such as brain haemorrhage detection, can not meet the requirements of the doctor What doctors need is an artificial intelligence that can determine the state of a patient's brain In industry parlance, doctors need artificial intelligence products that target at least one part of the "whole disease" this is a trend and one of the alternative paths for enterprise design clinical trials From the current situation, the medical AI enterprises led by Ander medical wisdom to choose and practice this direction - try to create a multi-site, multi-disease products two, choose effective data find the right scenario, enterprises need to use data to polish products, but the acquisition of data is not so simple from the existing algorithm mechanism, if the effective data of primary care to cultivate AI products, then the highest level of this AI product may only stay in the general use of primary care, can not be more inanosis to extend to large hospitals "For the diagnosis of breast cancer, brain tumors and other diseases, different levels of hospital differences are too much, if you choose data at will, there is a good chance that the more training, the worse the accuracy." Wu Zhenzhou said , in order for medical artificial intelligence products to land in The Sana Hospital, we must use high-quality data from top hospitals and deeply learn from the "golden standard" clinical experience of top experts in order to ensure the accuracy of AI Third, the threshold to an effectiveness the medical threshold of artificial intelligence may not have been so obvious for a long time - as long as high-quality data is available, businesses can quickly rise to the top, and now everything has changed "Many ai companies will soon find that open source algorithms of the past are starting to work less well As we gradually move towards the whole disease, the single-task depth learning algorithm is no longer able to cope with the demand, multitasking algorithm will be the general trend Wu Zhenzhou said that since the beginning of his establishment, Ander Medical Zhi has a "learning bull" research and development team, all from Harvard, MIT, Singapore National, Tsinghua University, Chinese Academy of Sciences and other areas of deep learning top universities, especially good at the simultaneous analysis of multiple diseases so, in addition to continuing to compete for high-quality, effective AI data, the next stage, medical aI enterprises must find breakthroughs at the algorithm icing level wrote in the final in general, 2020 has a good start to the approval of the medical artificial intelligence devices of Ander Medical, but the development of medical artificial intelligence is still a long way to go Overlooking the existing medical imaging AI market, there is even competition between businesses and businesses - few are able to offer mature artificial intelligence products , it is important to carry the three types of device approval that are the efforts of many medical ai-artificial intelligence practitioners, but it is more important for enterprises to continue to work deeply on algorithms, find data, and base themselves on medical care, in order to fundamentally change the existing medical artificial intelligence pattern And and has received three types of certificates, Ander,odo, will continue to accelerate on this track as Ander Medical Zhi said: "The value of medical AI is essentially to solve clinical needs, how to use AI technology in the future layout as a starting point, the use of AI technology to empower medical care, is our foothold in the industry, but also always need to adhere to the first heart" medical artificial intelligence has a long way to go, but fortunately through the efforts of many medical AI practitioners, including Ander medical wisdom, dawn, the future can be expected."