Three subdivision business models of artificial intelligence + medical care

There are many ways to combine artificial intelligence and medical treatment. From the viewpoint of medical treatment process, there are applications for each stage before, during and after consultation; from the viewpoint of application objects, there are multi-role applications for patients, doctors, hospitals and pharmaceutical companies; from the viewpoint of business types, there are various modes such as efficiency enhancement and cost reduction.

Virtual assistant. Virtual assistant is a kind of auxiliary robot that can communicate and exchange with human beings. It understands human thoughts and needs through artificial intelligence technology, and outputs various kinds of knowledge and information to assist human beings’ life and work. Generic virtual assistants are relatively familiar to everyone, such as Siri on Apple’s phone, Microsoft’s Cortana, Amazon’s Alexa, Google’s Google Assistant, Facebook’s (Facebook) M, etc. Artificial intelligence virtual assistants use natural language processing technology for speech and semantic recognition, as well as optimized decision algorithms to accomplish interaction with humans. With the help of virtual assistants, people can directly state their questions, wishes and needs and get answers from the feedback of the virtual assistants. There are two general methods of interaction between people and virtual assistants, voice and text, and the machine communicates with humans through voice and semantic recognition. Therefore, speech recognition technology is a very important technology in virtual assistant products. But there is another way of interaction for medical type of virtual assistants, which is multiple choice questions. Because it is difficult for ordinary people to express their problems in accurate language, most of the virtual assistants in the medical health category will use multiple-choice questions to communicate with people. According to the classification of virtual assistants’ service objects, virtual assistants can be divided into three categories, which are virtual assistants whose users are patients, including applications such as personal consultation and medication consultation; virtual assistants that link doctors and patients at the same time, including applications such as intelligent guidance, triage robots and chronic disease management; virtual assistants whose users are doctors, including applications such as voice entry of electronic medical records.

Disease screening and prediction. Modern medicine is to diagnose whether people are sick or not from various biochemical and imaging test results, but there is still a long way to go to achieve more scientific and accurate prediction of diseases in the future. Diagnosis of mental illness. Diagnosis of mental illness In ordinary psychological treatment, doctors first make a preliminary diagnosis of the patient’s mental condition, determine the symptoms through several interviews similar to psychological interviews, diagnose the type of mental illness based on experience, and then draw up a treatment plan for the illness, including what kind of drugs to use and how much to take. In 2015, a group of researchers created an artificial intelligence model based on the linguistic characteristics of schizophrenia patients, and by analyzing conversation transcripts, accurately predicted which group of people were likely to suffer from schizophrenia (the major symptom).

Brain hernia prediction. Brain herniation prediction of massive cerebral infarction is a common and very serious neurological disease that accounts for about 10% of all cerebral infarction patients and has a very high mortality rate of about 80%. [3] A large number of studies have shown that active intervention before the deterioration of symptoms occurs is more effective than later intervention, so early and effective judgment of patient prognosis, and thus selection of an effective treatment plan, is the key to successful treatment of patients with cerebral infarction. A paper titled “Predicting the regression of patients with massive cerebral infarction using artificial intelligence systems” was published in China Health Statistics in 2014. The paper stated that a multifactor prediction model using an artificial neural network multilayer perceptron to predict the prognosis of patients with massive cerebral infarction had the best prediction in the single-factor model with an AUROC (area under the subject’s working characteristic curve) of 0.87 [4]. It was finally concluded that the artificial intelligence random forest model can be used as a medical aid diagnostic system to predict the probability of brain herniation in patients with massive cerebral infarction.

Death prediction in cardiac patients. Prediction of death in cardiac patients British scientists have published a research article in the journal Radiology, which concluded that artificial intelligence can predict when a cardiac patient will die. The MRC London Institute of Medical Sciences, under the UK Medical Research Council, said the artificial intelligence software can detect signs of impending heart failure by analyzing blood test results and heart scans. The researchers got these results from a study of patients with pulmonary hypertension. This technology could save more lives by allowing doctors to identify patients who need more interventional treatment. Increased blood pressure in the lungs can damage part of the heart, and about one-third of patients will die within five years of diagnosis. Current treatments include direct injection of drugs into blood vessels and lung transplants, but doctors need to know how long patients will survive in order to choose the right treatment option. The researchers entered the results of cardiac MRI scans and blood tests from 256 heart patients into artificial intelligence software. In each heartbeat measured by the artificial intelligence software, 30,000 points of movement were marked on the heart structure. Based on this data, combined with the patients’ health records over eight years, the software was able to predict which abnormal conditions would lead to the patients’ death. The artificial intelligence software is able to predict survival for the next five years, and the accuracy of predicting a patient’s survival for only one year is about 80%, while the accuracy of doctors’ predictions for this project is 60%.

Medical imaging. Modern medicine is evidence-based medicine based on experiments, and doctors’ diagnosis and treatment conclusions must be based on the corresponding diagnostic data. Imaging is an important diagnostic basis, and 80% to 90% of the data in the medical industry comes from medical imaging, so clinicians have an extremely strong demand for imaging. They need a variety of quantitative analysis of medical images, comparison of historical images to be able to complete the diagnosis. “Artificial intelligence + medical imaging” is an auxiliary tool for computers to complete classification, target detection, image segmentation and retrieval of images through deep learning on the basis of medical images, and assist technologists and doctors to complete imaging, lesion screening, target area outlining, three-dimensional imaging of organs, pathological analysis, quantitative image analysis, etc.

Imaging. On the one hand, there is a lack of high-level technologists, especially in primary hospitals, and there is a waste of imaging resources caused by duplicate imaging. On the other hand, advanced imaging functions are complex, and sometimes a technologist can have a big impact on image quality by adjusting sequences and parameters. Artificial intelligence can help to achieve standardized quality and make corresponding personalized adjustments according to different people, such as people with different BMI heights when doing CT images, it has adjustments in imaging parameters and dose adjustments, which ultimately allows inexperienced primary care doctors to get the same medical images as in tertiary hospitals.

Read the film. First, to present the information better to the doctor. Now imaging is getting easier and easier, the resolution is getting higher and higher, doctors need to see more and more images, but what doctors need is not data, what doctors need is information, how to present this information better to doctors? Artificial intelligence can complete the localization, classification and segmentation of organs, and mark the suspicious location, which is equivalent to removing the interference items for doctors, and presenting more direct information. Second, it helps doctors to analyze quantitatively. Doctors are very good at qualitative analysis. When seeing a film, an experienced doctor can roughly determine what is wrong within 3 seconds, but needs some tools to make more accurate judgment, and quantitative analysis is hard to do by eyes. The work here includes a variety of multimodal analysis, comparison of historical images, and analysis of patient populations, which cannot be done simply with the eyes, but require image segmentation, image configuration, and functional image analysis.

Leave a Reply

Your email address will not be published. Required fields are marked *