Integrating AI into the entire clinical decision-making experience by the cumulative medical records of hundreds of millions of Chinese patients; each diagnosis and treatment will be based on patient-specific information, combined with the experience of countless identical cases. Higher precision, easier, cheaper, faster and personalized diagnosis; these are the driving force for the integration of AI into modern medicine, and it is particularly what China's healthcare system needs desperately.
The application of the early models, loosened privacy regulations and financial incentives in China bring about the optimistic attitude towards the “soft landing” of AI in Healthcare. Many diseases are often diagnosed in the middle or later stages. If there is a medical intervention in the early stage, it would be of an excellent benefit for the treatment and rehabilitation process. Therefore, several AI embedded solutions for the early diagnosis of several types of cancer and tumour are only being implemented after the disease had already seriously proliferated amongst human body; which brings about the inefficient use of early-diagnosis med-tech solutions in healthcare.
Fifty years ago, the Wall Street Journal boldly predicted that in the future, computers would replace doctors in specific situations to complete patient diagnosis, but as of today, the core of the healthcare system is still human-dominated. We have witnessed loads of medical data circulations, but it seems there is still a long way to go till AI-powered business succeeds.
In the early days, healthcare was considered to be one of the most promising applications of AI. At the time, researchers proposed and developed several clinical decision support systems. However, the construction cost is too high, and the rule-based on it must be logical and interpretable. The rules in the medical field are too broad and complex, and it is difficult to extract relevant information.
Several healthcare branches, such as radiology, ophthalmology, dermatology, and pathology, rely on medical-image analysis support software. In 2018, the FDA approved the first deep learning system to diagnose cardiovascular disease using cardiac MRI (MRI) images, which helps doctors diagnose heart problems. It uses a self-teaching artificial neural network. Cases will continue to feed the amount of data which makes the diagnosis process smarter. The system takes approximately 15 seconds to get the results of an analysis, whereas a professional doctor needs between 30 minutes to an hour for the same process. However, the system has several prerequisites to be utilized efficiently, and it requires patients to go to a hospital where they can find an and MRI or MI device, affordable and readily available, which is not the case for most of the time.
Accurate clinical interpretation of genomic data is key to understanding individual differences and reaching a higher efficacy for precision medicine. Compared with traditional methods, deep neural networks can better annotate the malignant gene mutations and identify non-coding DNA functions. Although the potential social and financial benefits that will be arising from the genomic data analysis are explicit, it is currently not widespread to collect these genetic data from the individuals; which hinders and postpones the practical mass application of the technology. Discovery biomarkers and clinical outcome prediction systems are yet another med-tech solutions that are supported by AI; which face significant regulatory challenges. Health monitoring via wearable devices
Contemporary wearable devices record a large number of biomedical signals, including heart rate, sound, motion and several others. These records can be used to detect possibles diseases. For instance, heart rate and skin temperature data recorded by wearable devices can recognize signs of infectious diseases and inflammatory reactions. A photoplethysmography sensor is added to the wearable device to monitor cardiovascular disease, lung disease, anaemia, and sleep apnea; the latest version of the Apple Watch provides such functions. Although they seem complicated, these sort of technologies are quite basic and affordable; several photoplethysmography sensors supported wearable devices can be found with prices as low as USD 20. A photoplethysmography sensor even enables these wearable devices to take an EGC; which can be a breakthrough in healthcare.
Cardiovascular diseases are not the only ones that are addressed by wearable devices. Several solutions also address the diagnosis and monitoring of diabetic conditions. Medtrum (移宇科技), a Shanghai-based diabetes management company, provides sets of wearable devices addressing insulin management.
According to WHO’s report published in 2016, diabetes is listed as 1 of 4 NCDs (Non-Communicative Diseases) targeted by world leaders since the losses brought by diabetes hinder human social development. The estimated GDP losses caused by diabetes from 2011 to 2030, directly and indirectly, will total USD 1.7 trillion. Besides the billion-sized markets of CGM and artificial pancreas, the welfare of millions of diabetes patients could also be better off. Medtrum presumable is a significant company in this gigantic market.
The company announced the series C funding in the amount of CNY 200 million led by Sequoia China and BOCIG (Bank of China Investment Group) and followed by 3E Bioventures (本草资本) and Nuokai Capital(诺恺投资) on December of 2018.
Besides, due to the limited medical resources, it is impossible for doctors to interact with all patients who need treatment; particularly in China. If patients are guided and incentivized to wear sensors or activity tracking devices and convey medical data via their smartphones to obtain a diagnosis, many can be avoided.
As of 2019, health data collection and processing via wearable devices provide the fastest and the most available AI-powered solution within the healthcare industry.