Main menu


Artificial intelligence use in healthcare 2022

Artificial intelligence use in healthcare 2022

Artificial intelligence use in healthcare 2022

Artificial intelligence technology can be used most hotly in the medical field. Because various and extensive tasks such as finding, treating, observing the prognosis, and researching and applying new treatment methods are being done in the medical field, new methods are always needed to efficiently perform these tasks. am. However, on the other hand, since it is directly related to human health and life, it requires a lot of thought in applying new technology.

In this post, we will briefly review artificial intelligence technologies that are being researched and utilized in the recent medical field. And furthermore, it introduces the current concerns about whether artificial intelligence technology can replace the role of medical staff.

What artificial intelligence technologies will be used in the medical field?

The field of artificial intelligence includes many sub-fields, and so does the medical field. Therefore, to classify artificial intelligence used in the medical field, two methods can be applied. The first is to divide the AI ​​technology field and look at the medical applications to which each technology is applied [1], and the second is to divide the medical field and look for AI technologies for each field.  In this post, we will use both methods, and we will follow the method applied in the reference paper

Classification according to AI technology

·       Computer vision – medical imaging

Artificial intelligence use in healthcare 20.22

Figure 1. Structure of CNN for processing medical images

Deep learning is the most widely used technology in the field of artificial intelligence. And deep learning is most widely used in the field of processing image or video data. Figure 1 briefly expresses the concept of inputting medical image data into a Convolutional Neural Network (CNN), one of the most representative models of deep learning. Various tasks such as classifying types by inputting image data, finding specific patterns in images, or tying images with similar characteristics can be performed using the deep learning model. It can be used in the dermatology department to identify the condition and symptoms of the skin, and in the radiology department, it can be used in the process of finding nodules or tumors in CT or X-ray images. In dental X-ray images, it can help determine the condition of teeth or jaw joint, and in ophthalmology, it can help fundus examination to detect abnormalities in the retina. use in

·        NLP(Natural Language Processing)

The second most representative artificial intelligence technology is Natural Language Processing (NLP). And NLP can be most utilized in the process of processing Electronic Health Record (EHR). EHRs contain a huge amount of information. Not only data, but also the experience and wisdom of doctors are melted. Therefore, it is important to properly extract and utilize important information from the EHR and to properly record the information in the EHR. The process for extracting information from EHR has been tried for a long time in the data mining field, and research on extracting important information from content written in natural language and building a system to understand it is representative. In recent years, efforts are being made to free medical staff who are overloaded with paperwork by converting spoken contents into text and storing them. Research related to extracting and storing information from conversations between doctors and patients is also ongoing. Of course, in this process, not only NLP but also speech recognition technology is used.

·       Robotics

Robotics is different from the fields mentioned above in that it is a field in which various technologies are fused. However, it is undoubtedly an area that can greatly contribute to the development of medical technology. Robotics technology is typically applied to surgical robots, and in detail, laparoscopic surgery can be greatly developed. [3] The key is robotic-assisted surgery (RAS) using reinforcement learning, and allows the robot to learn the doctor's actual surgical procedure through trial & error, or the doctor remotely controls the robot. Attempts are being made to use robots for repetitive tasks such as suturing and knot-tying.

·       Generalized information processing

Artificial intelligence use in healthcare 20.22

2. Machine Learning in Genomics

There are AI technologies that are specialized for specific types of data, such as Vision or NLP, but in real situations, it is often necessary to process various complex data. For example, in precision medicine, which predicts which treatment protocol is most likely to be successful based on patient characteristics or treatment environment, or in the process of analyzing how well a patient usually follows a doctor's instructions, various types of data are used. should be collected and analyzed.

Figure 2 conceptually expresses the process of applying artificial intelligence in Genomics, and the representative example is analyzing what kind of disease a mutation identified in genetic information causes. Artificial intelligence technology is also used in the process of predicting a phenotype (eg body size) or estimating the probability of contracting a specific disease based on genetic information. These analysis results can be used to find the most suitable customized treatment method for each patient by interlocking with the imaging data that records the patient's condition, the data usually collected using a wearable sensor device, and the patient's examination (treatment) history information.

Artificial intelligence technology is also being used in the field of drug development, and it is used for tasks such as finding substances that can be used to treat specific diseases, synthesizing compounds, predicting drug properties, and exploring new ways to use specific drugs. is utilized

Classification according to medical field

·       AI for living assistanc

The goal in the field of making devices or systems that help in daily life is to allow users or patients to efficiently control the devices or systems they use. In particular, this technology is very important for patients who cannot use their hands freely or suffer from paralysis in certain parts of the body. do. Also, in the ambient intelligent system-related technology that tracks the overall human condition or fall-detection systems, the artificial intelligence module that detects the user and environment status from various sensor data plays the most important role. In addition, artificial intelligence robots that help cognitive (memory) rehabilitation while continuously interacting with users in daily life are being studied.

·       AI in biomedical information processing

Clinical information is large-capacity, and biomedical data in the form of images and video information as well as text are complexly connected. Therefore, if a process for extracting and merging knowledge from such data or automatically finding and resolving conflicts between information is developed, it is possible to do things that were not possible in the past. A QA system to provide appropriate information to patients or caregivers, and an intelligent workflow system to reduce the administrative burden on medical staff can also be included in this field.

·       AI in biomedical research

The process of searching for and utilizing a vast amount of newly published academic research materials in the medical field is a time-consuming and difficult task. Therefore, attempts are being made to automatically perform search, analysis, summary, and arrangement of large-scale academic research data based on artificial intelligence technology. As the COVID-19 virus entered the pandemic stage, artificial intelligence researchers attempted to automatically analyze the research results accumulated so far. In addition, artificial intelligence technology plays an important role in the process of developing a system that simulates medical experiments that are difficult to perform in practice.

·       Disease diagnostics and prediction

Artificial intelligence can be applied to the process of diagnosing or predicting a disease. Typical tasks include detecting abnormalities in genetic information, diagnosing cardiovascular diseases based on ECG data, finding diseases from medical image data, or predicting the survival rate of cancer patients. This process can greatly assist the doctor in the diagnosis and treatment process.

Will artificial intelligence replace doctors?

There are three main views on whether advanced artificial intelligence will replace human doctors. These are the position that it will be replaced, the position that it cannot be replaced, and the position that it will coexist.

The position that artificial intelligence will replace humans is based on its massive data processing capabilities, tireless learning and monitoring capabilities, and predictions that it will cost less than using humans. Of course, the relationship between doctors and patients is a matter of trust and empathy, but it is a view that even this can be imitated by artificial intelligence, and patients will want accurate and quick diagnosis rather than empathy. It is also suggested that the patient will not feel ashamed when artificial intelligence gives treatment, unlike when a doctor gives treatment. Some argue that eventually human doctors will play a role in helping AI.

On the other hand, the view that artificial intelligence cannot replace human doctors argues that health care is not simply achieved through data. The clinical context plays a big role in the diagnosis and treatment process, and it is argued that artificial intelligence cannot grasp the context and situation like humans. Human doctors treat patients as human beings, and it is argued that artificial intelligence will not be able to implement medical services based on trust, responsibility, and dedication based on overall understanding and insight into the patient's life and environment. In addition, in many cases, it is not possible to explain the patient's situation only with technical knowledge and quantitative numbers, and sometimes it is necessary to select a treatment method or drug while knowing the side effects or risks. It also raises the question of whether Furthermore, the treatment process involves listening to the patient's questions and giving a feeling of understanding the patient's feelings. From another point of view, it is argued that it is difficult to perfectly link the clinical workflow and EHR system with artificial intelligence technology. [4]

From the standpoint of coexistence [5], we have the view that the best scenario is to acknowledge and complement each other's limitations

Artificial intelligence use in healthcare 20.22

Figure 3. Multi-step automation concept

Figure 3 considers the utilization level of medical artificial intelligence as the 'automation' level, and compares it with the automation level of the driverless car. In the figure, it is judged that complete automation is impossible. There are certain things that AI can do better than humans, but considering the fatality of the problems that may be caused by the bias that may occur in the process of implementing artificial intelligence and the errors that may be caused by artificial intelligence, the ultimate purpose of medical artificial intelligence is argues that it is not complete automation, but corresponds to conditional automation. A high level of automation or complete automation is impossible, and it is a view that always requires human review and monitoring. The notion that doctors will be completely replaced by AI is a belief that artificial intelligence is driven by excessive expectations and fantasies, because AI always needs to be modified and verified. It is also argued that it is wrong to think that the 99% accuracy recorded by artificial intelligence during the R&D process will be shown in clinical practice, and it is also wrong to think that a judgment based on large-scale data will always produce the best results in the treatment process of individual patients.


In the process of artificial intelligence being put into actual medical practice, conditions other than technology - legal environment, economic, social conditions, ethical considerations, social acceptance, etc. - will also promote or slow down the use of artificial intelligence. Artificial intelligence is clearly appearing little by little in the medical field, and will replace humans first in dealing with information and data rather than directly dealing with patients. And there is ample room to contribute in the development of preventive, participatory, and personalized healthcare technologies. Thus, anyone who rejects AI completely and refuses to use it at all may lose his or her job in the medical field. However, on the other hand, it is also important that medical AI technology must be gradually incorporated into the actual clinical environment, and that it must go through a thorough verification process in the process. Also, even if the verification process has been completed, thinking that these technologies will completely replace medical staff may be a misjudgment based on excessive illusions about technology. This is because the medical process is an interaction between the medical staff and the patient, and the fact that the knowledge, experience, judgment, insight, and prompt and accurate actions of the medical staff are absolutely necessary in the medical field will not change.