Recently, the application of Artificial Intelligence (AI) in the medical field has become a technological development and commercial hotspot in the health industry, with enormous potential and industrial value. However, how to implement AI as a novel technology in the highly complex field of healthcare still faces many challenges. This article will summarize the current state of AI in healthcare and introduce domestic and international applications and innovative commercial practices in this field, for the reference of experts and practitioners interested in this area.
I. The Development History and Constituent Elements of Medical Artificial Intelligence
Artificial Intelligence (AI) is a branch of computer science that is an emerging science that simulates and systematically applies human consciousness and thought processes. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. The concept of "artificial intelligence" was first proposed by McCarthy at the first Artificial Intelligence Workshop held at Dartmouth College in the summer of 1956, marking the birth of the discipline.
Medical artificial intelligence refers to the application of artificial intelligence in the medical field, involving all aspects of the medical industry. Its ultimate goal is for artificial intelligence to replace humans in diagnosing and treating patients. Currently, its main development directions include assisted diagnosis, medical image recognition, drug development, health management, and gene sequencing.
1.1 The Development History of Medical Artificial Intelligence
Early exploration of medical artificial intelligence
The earliest attempts to explore artificial intelligence in the medical field appeared in the 1970s. In 1972, AAPHelp, developed by the University of Leeds, is the earliest documented AI system in the medical field. This system was primarily used for assisting in the diagnosis of severe abdominal pain and for surgical-related needs.
Throughout the 1970s, numerous new advancements emerged. INTERNISTI, developed by the University of Pittsburgh in 1974, was primarily used for the auxiliary diagnosis of complex internal medicine diseases. MYCIN, developed by Stanford University in 1976, could diagnose patients with infectious diseases and prescribe antibiotics. It contained 500 rules; by answering its questions in sequence, the system could automatically determine the type of bacteria causing the patient's infection and prescribe the appropriate medication. Other systems included CASNET/Glaucoma developed by Rutgers University, PIP and ABEL developed by MIT, and ONCOCIN developed by Stanford University. By the 1980s, several commercially available systems had appeared, such as QMR (Quick Medical Reference) and DXplain developed by Harvard Medical School, which primarily provided diagnostic solutions based on clinical presentation.
In general, most early explorations in medical artificial intelligence have been unsuccessful. However, this situation only illustrates the high complexity of medicine and has not deterred humanity from attempting to explore artificial intelligence in the medical field.
Recent Developments in the Field of Medical Artificial Intelligence Abroad
Currently, IBM Watson is the most well-known AI in the medical field, and it has achieved remarkable results. For example, in cancer treatment, Watson can sift through 1.5 million patient records from decades of cancer treatment history in seconds and provide doctors with evidence-based treatment options. The top three hospitals in the cancer treatment field are all using Watson, and Watson has officially entered the Chinese market.
Besides IBM, tech giants like Google and Microsoft have also made positive progress in the field of medical artificial intelligence. In February 2016, Google's DeepMind announced the establishment of DeepMindHealth, a division that would collaborate with the UK's National Health Service (NHS) to assist in decision-making, improve efficiency, and reduce time. DeepMind also participated in an NHS study using deep learning to design radiotherapy treatments for head and neck cancer patients. DeepMind's collaboration with Moorfields Eye Hospital also applies AI technology to the early detection and treatment of vision-threatening eye diseases.
In 2016, Microsoft announced its Hanover initiative to use AI in healthcare, aiming to help find the most effective drugs and treatments, and to collaborate with the Knight Cancer Institute at Oregon Health & Science University on drug development and personalized treatment.
While planning to develop its own AI chips, Apple has also made several acquisitions of AI companies. Following its acquisition of Siri in 2010, Apple has made several other acquisitions in the field of voice technology in recent years, including VocalIQ and Novauris Technologies.
The history and current status of medical artificial intelligence development in my country
Research and development in the field of artificial intelligence in my country began in the early 1980s. Although starting later than in developed countries, its development has been rapid. In 1978, Professor Guan Youbo of Beijing Traditional Chinese Medicine Hospital collaborated with experts in computer science to develop the "Guan Youbo Liver Disease Diagnosis and Treatment Program," marking the first application of a medical expert system to the field of traditional Chinese medicine. Subsequently, representative AI products in my country include the "Lin Rugao Bone Injury Computer Diagnosis and Treatment System," the "Chinese Traditional Medicine Treatment Expert System," and the "Traditional Chinese Medicine Computer-Aided Diagnosis and Treatment System," which provides consultation and auxiliary diagnosis. Since the beginning of the 21st century, my country's medical artificial intelligence has made significant progress in many more areas.
In October 2016, Baidu officially released its latest achievement in the medical field—Baidu Medical Brain—under the theme of "Opening a New Era of Intelligent Healthcare," aiming to compete with similar products from Google and IBM. As a specific application of Baidu Brain in the medical field, Baidu Medical Brain simulates a doctor's consultation process through the collection and analysis of massive amounts of medical data and professional literature, providing final treatment suggestions based on the user's symptoms.
Tencent, leveraging the vast data volume and dimensions of WeChat, is exploring the development of medical artificial intelligence. For example, Tencent's collaboration with Sun Yat-sen University Cancer Center (Nanshan Branch) is piloting an early esophageal cancer screening system in Shantou, Guangdong. Tencent uses AI-powered image processing to assist in early esophageal cancer screening, significantly reducing manual labor while improving the medical capabilities of medical institutions. Furthermore, Tencent's AI lab will collaborate with Zhuojian and Yilian to develop a future-oriented follow-up consultation system.
Alibaba is focusing its efforts on the field of intelligent medical diagnostics. In July 2017, Alibaba Health launched its medical AI, "DoctorYou." The "DoctorYou" AI system includes a clinical medical research and diagnostic platform, a medical auxiliary testing engine, and a physician training system. Furthermore, Alibaba Health has collaborated with local governments, hospitals, research institutions, and other external organizations to develop intelligent diagnostic engines for 20 common and frequently occurring diseases, including diabetes, lung cancer prediction, psychological intelligence, and fundus screening.
1.2 Three Stages of Medical Artificial Intelligence
From the perspective of technological development, artificial intelligence can be divided into three stages: computational intelligence, perceptual intelligence, and cognitive intelligence. In the first stage, machines began to calculate and transmit information like humans. In the second stage, machines began to understand what they see and hear, make judgments, and take actions. In the third stage, machines are able to think like humans and take proactive action.
Chart 1: Three Stages of Medical Artificial Intelligence Development
Chart source: Qipu Research
From the perspective of data effectiveness and business model development, the application of medical artificial intelligence can be divided into three stages: The first stage is the data integration stage. While advanced algorithms such as deep learning already exist, the low standardization of medical data and weak sharing mechanisms limit the application scope and effectiveness of AI in the medical industry. Before the sharing mechanism matures, companies possessing large amounts of medical data have a competitive advantage and bargaining power. The second stage is the "data sharing + perceptual intelligence" stage. When medical data is integrated to a certain extent, auxiliary commercial products will emerge in various fields such as assisted diagnosis and treatment, and image recognition. In this stage, both data and algorithm advantages become important barriers, and effective data will further optimize the implementation of algorithms. The third stage is the "cognitive intelligence + health big data" stage. In this stage, AI as a whole develops from perceptual intelligence to cognitive intelligence, the cost of acquiring health big data will decrease, and humanity will enter an era of personalized medicine. This stage will see AI applications that can replace human doctors.
Chart 2: Development Stages of Medical Artificial Intelligence from the Perspective of Data and Algorithms
Chart source: Qipu Research
1.3 The Three Elements of Medical Artificial Intelligence
The core of artificial intelligence is the algorithm, and the basic conditions are data and computing power. Therefore, it can be considered that the key elements for combining medicine and artificial intelligence are "algorithm + effective data + computing power".
Advanced algorithms are the core of medical artificial intelligence, enabling improved data utilization efficiency. With the continuous development of advanced algorithms, artificial intelligence is moving from computational intelligence to perceptual intelligence, and will further advance towards cognitive intelligence in the future. Advanced algorithms can improve the efficiency of converting information into "knowledge," thereby enhancing the level of intelligence.
Effective medical big data is the foundation for the application of artificial intelligence. The effectiveness of medical data encompasses three aspects: the degree of digitization, the degree of standardization, and the sharing mechanism. The degree of digitization emphasizes the quantity of data and medical records available; the degree of standardization emphasizes the comparability and universality of data; and the sharing mechanism emphasizes the convenience and legality of data acquisition channels. Only when these three conditions are met can medical big data be effectively collected and applied, thus laying the foundation for artificial intelligence.
Computing power is another fundamental requirement for medical artificial intelligence. In the future, the development of quantum computing and faster chips will further drive the application of artificial intelligence.
Chart 3: Three Key Elements of Medical Artificial Intelligence
Chart source: Qipu Research
II. Commercial Applications and Business Model Innovation of Medical Artificial Intelligence
Since the development of medical artificial intelligence must be based on effective medical big data, AI can be useful in any field of medicine where effective data can be obtained. For example, medical AI has made good progress in gene sequencing, assisted diagnosis, and drug development.
Here we will first introduce IBM's Watson, a representative company in the field of medical artificial intelligence. Its main business model is to expand the business model of artificial intelligence in related fields by leveraging its data and algorithm advantages in a certain type of disease.
2.1 IBM Watson: Deeply Focused on Artificial Intelligence Technology in the Oncology Field
IBM Watson, an artificial intelligence system developed since 2007 by the DeepQA team led by David Ferrucci, IBM's chief researcher, is currently a leader in the field of medical AI. Watson's business strategy in the medical field consists of three aspects: first, a deep focus on oncology and expansion into other areas; second, acquiring data resources through acquisitions; and third, expanding use cases and exporting ecosystem capabilities through collaborations.
Watson began using machine learning in the field of oncology in 2011, and after extensive training, developed the WatsonforOncology product. This product helps oncologists quickly understand similar medical records, complete preliminary diagnoses, and shorten diagnostic time. Next, Watson began to provide the ability to identify evidence-based adjuvant treatment options.
By combining attribute data from patient files with clinical knowledge and external data, Watson provides doctors with cancer treatment solutions for reference. Currently, Watson's diagnostic accuracy for different types of cancer ranges from 70% to 90%.
For more information, please visit the Industrial Robots channel.
