Artificial intelligence accelerates the arrival of precision medicine
Release date: 2016-09-23 Can you imagine that every day, three large planes full of passengers fall into the death of all the members? This sounds terrible, but the number of people who have been killed by malaria every year is as high as 600,000 to 800,000, which is equivalent to the probability of such a crash. Although malaria has almost been eradicated in developed regions, malaria is still a disaster in some underdeveloped areas, and one of the challenges in dealing with malaria comes from the lack of adequate professional pathologists, making it difficult for patients to get timely diagnosis. And treatment. The technology currently being developed will help determine whether a patient is infected with malaria, which type of malaria is infected, and which channels may be infected. Compared with traditional methods, it requires a lot of manpower to look at samples and analyze. This technology greatly enhances the efficiency of doctors. Even in areas where medical staff are scarce, it can be less of a stretch. Secondly, after the image recognition at the cell level is solved, the recognition of the diseased gland is realized. The so-called gland can be simply understood as a collection of multicellular cells, which is closer to the concept of "organs." Compared with cytopathic lesions, the complexity and possible combinations of glandular lesions increase exponentially, but the accurate identification of glandular state can greatly improve the accuracy of cancer analysis, and the significance is more profound. Source: Microsoft Research Soft Candies,Soft Chews Candy,Soft Chewy Candy,Soft Spot Candy Guangzhou Etechange Biological Engineering Co.,Ltd. , https://www.etechange.com
Therefore, the combination of computer and medical care is far more than smart bracelets, blood glucose meters, or intelligent hardware such as Xbox and HoloLens that may be associated with medical treatment. The coverage ranges from front-end equipment to back-end systems to hidden in the end. Each type of algorithm can be an independent discipline. In fact, within Microsoft, there are close to 100 medical-related projects, both of which are very forward-looking and have already entered the practical application level.
In my opinion, the progress of computers in the medical field is actually based on the same foundation, that is, "data change medical care". Whether it is Chinese medicine or Western medicine, it is essentially practical science. Doctors have summed up and counted the laws through countless times of practice, and finally achieved the effect of saving people from illness and illness. As the ability of humans to collect, process, and analyze data grows with the development of technologies such as cloud computing, big data, machine learning, and the Internet of Things, the ability to use big data like a doctor to analyze or aid in analyzing a disease will naturally Increasingly.
Artificial intelligence helps promote precision medicine
Cancer has always been one of the most difficult medical problems that humans need to solve urgently. Since each patient of the same type of cancer has different performances, it can be said that each patient's cancer is an independent disease, even if the doctor is rich. The experience is also difficult to make 100% accurate analysis and judgment, let alone relatively personalized precision medicine. Therefore, Microsoft Research Asia has been using digital medical image recognition as one of the main directions, and hopes to accelerate the promotion of precision medicine through the latest technology in the field of computer vision.
Since 2014, the team of Microsoft Research Asia has begun to study the identification and judgment of pathological sections of brain tumors, and to assist in the analysis and judgment of the stage of cancer in the patient through the shape, size and structure of the cells. In the past two years, we have made two breakthroughs in this field based on the "neural network + deep learning" model:
First, image processing for large-scale pathological sections is achieved. Usually the size of the picture is 224 * 224 pixels, but the size of the brain tumor record slice reached 200,000 * 200,000, or even 400,000 * 400,000 pixels. For the identification of large-scale pathological slice images, we have not used the digital medical image database commonly used in the industry. Instead, we use as many images as possible on the basis of ImageNet, the most mature image database in the computer field, through our own neural network and depth. The learning algorithm is continuously trained in a large amount, and finally the image processing of large-scale pathological slices is realized. 
The identification of diseased glands is mainly based on three medical indicators that can measure the degree of cancer spread and prognosis: cell differentiation, gland status and mitosis. From these three perspectives, we hope to help doctors to predict and judge the possibility of postoperative, rehabilitation and even recurrence in the future through multi-channel data collection and analysis. 
The gland image is abstracted into different structures after computer processing, so that the computer can further identify and judge the past two doctors rely on the "eye" and experience to observe the pathological slice image and judge the condition. Now two core technologies in artificial intelligence: Neural network and deep learning enable the computer system to automatically learn the difference between malignant tumor cells and normal cells and the analysis and judgment criteria of cancer conditions. At the same time, after the pathological sections are scanned, the judgment results are given for doctors' reference. The powerful computing power of the computer makes up for the misjudgment caused by some doctors due to lack of experience, or the inconsistency of rare diseases and incurable diseases. Moreover, the computer can also find small details that are not easily noticeable to the human eye, and sum up some rules that are unexpected to the doctor, thus constantly improving the knowledge system of doctors and computer systems. Therefore, it is artificial intelligence that allows precision medicine to continue to advance.
At present, Microsoft Asia Research Institute has been at the international leading level for the accuracy of 2D medical image recognition results. In addition to brain tumors, the findings can be extended to the identification and judgment of two-dimensional medical images of other diseases, such as intestinal cancer we are studying. In addition, we are also studying CT three-dimensional images of patients with liver tumors. Although the three-dimensional images are fundamentally different from the recognition techniques of two-dimensional images, based on the in-depth accumulation of Microsoft Research Asia in the field of artificial intelligence, we believe that we are in 3D CT images. The breakthrough in identification is also just around the corner.
Super electronic medical record, doctor's " dictionary "
In addition to medical image recognition, we have done a lot of research on medical word processing.
When communicating with foreign counterparts, we found that the medical records written by doctors all over the world were the most difficult calligraphy. Due to limited time, doctors had to dance with their medical records. After the medical records are electronically, although the problem of writing is solved, the various descriptive languages ​​recorded in the medical records - some concise, some embarrassing, some even incomplete - for the doctor to follow up, review or study the disease It is very inconvenient for reference.
So our team studies speech and natural language understanding techniques, allowing doctors to dictate medical records, and then the computer converts the speech into text and then structured it to form a tree that contains all the keywords, summarizing all useful and clear. Information, so that patients or other doctors can see all the pathological processes at a glance, such as what medical history, what drugs have been used, which diseases are excluded, what diseases are to be investigated, and so on.
Based on such electronic medical records, the replacement of doctors will no longer affect the different doctors' complete understanding of the patient's complete condition; young doctors can also grow rapidly by learning various medical records; structured electronic medical records can even automatically summarize the details ignored by doctors and Inferred, get a new clue to the understanding of the disease; of course, greatly reducing the workload of doctors to write medical records is not necessary.
AI (Artificial Intelligence) + HI (Human Intelligence) = Super Doctor 
It can be seen that many technologies in the computer field can be closely integrated with medical applications, whether it is image recognition or natural language understanding. With the increasing computing power and the steady development of artificial intelligence technology, the future computer will be able to process more complex and advanced signals, and the human medical level will surely enter a new era.
However, doctors will never be replaced. In the field of medical science and art integration, artificial intelligence technology will become the doctor's "right arm", help doctors to obtain information more easily and assist doctors to make more accurate diagnosis, and doctors in addition to accumulated a wealth of professional knowledge, but also Need to play more high emotional intelligence to communicate with patients. In the end, the artificial intelligence of the computer and the human intelligence of the doctor will be combined to become a "super doctor" with precise professional judgment and emotional communication. Let us look forward to a new era of medical development led by artificial intelligence!