2025 Volume 25 Pages 53-70
Big data and artificial intelligence (AI) are now creating a whirlwind in the general methodology of both medical research and healthcare practice. It fundamentally transforms the entire paradigm of medical field, which could be called “the third revolution of medicine”, where the first one was caused by invention of antibiotics, contributed to eradicate the bacterial infection and the second one was brought by the invention of biopharmaceutical, such as molecular-targeted drug and antibody agent, introducing innovative treatment methods for cancer and several incurable diseases. This article discusses the future form of medicine which the third revolution will bring about. As for the methodological innovation of medical research, this revolution will bring about the data-driven approach for medical science, promoting the reverse science, which will be supported by big data and inductive AI. As for the medical practice and healthcare, this revolution will advance the current mobile health and real world medicine, which, incorporating the cutting edge molecular instrumental methods, will realize PM (precision medicine) mobile health and PM real world medicine. Moreover, on account of current progress in natural language processing as seen in the large language model, AI method will expect to develop to understand the interrelationship among the clinical events described in EMR to comprehend disease progression course, which will realize the “predictive control medicine”. Those innovation in both medical research and clinical practice will contribute to reduce the disparity of medical cure level of the clinical practice.
1. Introduction
There has been a recent increase of interest in big data and AI in the medical field; accordingly, it is imbued with a significance to elucidate its underlying reasons for this prosperity. Big data and AI are expected to fundamentally transform the entire paradigm of medicine and healthcare, which, the author anticipates, will bring about the “third revolution of medicine".
The first revolution of medicine occurred in the mid-20th century with the advent of "antibiotic therapy", which drastically reduced the bacterial infection, including tuberculosis, the most popular lethal epidemic at that time.
Subsequently, the second revolution occurred with the advent of "molecular medicine," which was caused by rapid advances in molecular life sciences and the subsequent introduction of molecular therapeutics, such as molecular-targeted drugs and antibody agents. These rapid advances ranged from the discovery of the double helix structure of DNA in the 1950s to the subsequent invention of recombinant DNA technology in the 1970s. However, it took as long as 40–50 years for this second revolution to permeate into actual clinical practice. Similarly, the coming third revolution based on big data and AI is expected to take same period of the second revolution until permeating into the practice of clinical medicine.
The third revolution of medicine is expected to result in radical changes ranging from the basic methodology of medical research to everyday healthcare practice.
This paper discusses the essential features of this third expected revolution, firstly in its scientific aspect of methodology of medical research which will be discussed at following chapter 2, and secondly its influence on the medical practice and healthcare which will be described at chapter 3. In the end of this paper, we will discuss the integrated effects to the future medicine in conclusion.
2. Methodological innovation in medical research in the third medical revolution: "Data-driven" research
2.1 "Hypothesis-driven" versus "data-driven" medical research
First, we discuss the methodological innovation in medical research during the era of big data and AI. The third revolution of medicine is expected to lead to "data-driven" medical research, which differs from conventional (traditional) "hypothesis-driven" medical research. In the conventional approach, certain unclarified research areas in the fields of medical science are selected based on the researchers’ interests or societal demands. Specifically, hypotheses are constructed to clarify the unclarified issues and are examined through the appropriate experiments, which allows the acquisition of bland new medical knowledge. As indicated by the philosopher of science K. Popper [1], the basis of the truth of newly obtained knowledge lies in the “falsifiability of the hypothesis by the experiment”. Accordingly, Popper argued that Marx's materialism and Freud’s psychoanalysis are not science since they lack a proper experimental method that demonstrates or falsifies their truth.
Contrastingly, in the era of big data and AI, comprehensive data subsuming the unclarified scientific issues will be initially collected, followed by the application of inductive methods such as machine learning and AI to clarify these issues and obtain bland new knowledge. However, “data-driven” research requires “comprehensive data” of uniform depth with "global coverage", which is the concept that has emerged in the life sciences especially after the human genome project. This concept has emerged to measure the nucleic acid sequences of the entire genome obtained using a high-speed next-generation DNA sequencer or whole-spectrum transcriptome (gene expression profiles) determined by using the DNA chip. The broadly-wide coverage of these data has been recognized even in the conventional medical field, which was alternatively called as the "genome-wide” property of the data. Moreover, “data-driven” medical research requires uniformity of data depth and the cross-disciplinary nature of obtained data.
Data-driven research requires a new methods other than conventional ones. This method includes a new machine learning and AI method known as “deep learning”, which is a novel type of neural network that uses the “auto-encoder network” method to automatically extract constitutive latent features lurked behind big data without human assistance [2]. Compared with similar conventional statistical methods such as principal component analysis (PCA), this “deep learning” method allows a better summary of large data represented by lower dimensional "latent space variables". Taken together, the automatic extraction of essential latent features from big data without human assistance is essential to realize "data-driven" medical research.
For "data-driven" medical research, the parent big data must (1) exhaustively cover the data-generating mechanism and (2) have independence from the previous human knowledge. These two conditions allow "data-driven" medical research to investigate the unclarified medical issues without human assistance, and thus facilitate discovery of the brand-new knowledge.
There are large differences between “hypothesis-driven” and “data-driven” research. In conventional “hypothesis-driven” medical research, there has been extensive research in areas with great research interests or societal demands, while there frequently exist adjacent areas with very little research interests or societal demands which are often completely unexplored. This results in a knowledge system with many isolated "gaps” (Figure 1). Contrastingly, "data-driven" research uses exhaustive data with uniform depth, which in turn yields knowledge with uniform depth that is not bounded by human-established disciplinary boundaries.
Figure 1. Difference in the knowledge distribution between (a) hypothesis-driven and (b) data-driven science
(a) A research hypothesis is selected based on the interests of medical researchers or societal demands. Therefore, there has been extensive research in areas with great research interests or societal demands, while there frequently exist adjacent areas with very little research interests or societal demands which are often completely unexplored. Thus, the current knowledge system is comprised of gaps of varying extents. (b) "Data-driven" research uses exhaustive data with uniform depth, which in turn yields knowledge with uniform depth that is not bounded by human-established disciplinary boundaries.
2.2 “Reverse Science” through "Data-driven" research
“Data-driven” research has become a well-recognized approach in life sciences. Traditionally, studies on genetic causes of diseases have first searched for characteristic abnormal proteins to provide clues for identifying disease-causing genes. However, the recently widely used statistical linkage analysis method of identifying the location of disease-causing gene mutations is based on the assumption that disease-causing gene mutations are located near the mutations of the certain genetic DNA marker which is strongly correlated with disease inheritance. Further, it uses pedigree (family tree) analysis of large families with frequent occurrence of target disease. This method is termed as the “positional cloning”. Moreover, since it differs from the methodology of conventional genetics, it is sometimes alternatively referred to as “reverse genetics” method [3].
Conventional “candidate gene approach” for identification of disease-causative genes has been recently replaced by the “comprehensive gene approach”. In the “comprehensive gene approach”, all genes are considered as a candidate instead of restricting the number of candidates. Similarly, genome-wide association analysis (GWAS), which has recently attracted much interest in the genetics, targets all single nucleotide polymorphisms (SNP) without any initial restrictions. Same approach was adopted in UK biobank, where no specific disease was pursued with all possible diseases observed. So, at the inception of the UK Biobank, there was some controversy regarding whether research without restricting target diseases could be considered scientific [4][5].
Conventional “hypothesis-driven” research can be considered as “forward science” since it firstly formulates the hypotheses based on existing knowledge, and subsequently conduct investigation of these hypotheses by doing experiment (Figure 2). Contrastingly, in the "data-driven" research firstly collects exhaustive data related to the unclarified issue and from them inductively discover the new knowledge without human assistance; accordingly, "data-driven" science can be alternatively called as "reverse science" (Figure 2).
From this perspective, modern science would be considered to have been exclusively developed through positivism based on hypothesis formation and experimentation. However, there have been many scientific discoveries based on “reverse science”. A notable example is the discovery of universal gravitation. Here, the Danish astronomer Tycho Brahe’s long-standing observations of planetary orbits were analyzed by his disciple Kepler, who discovered Kepler's three laws (the orbits of the planets are ellipses, not circles; the sun’s position is at one of the two foci of this ellipse; and the speed of the planet is negatively correlated with its distance from the sun). Newton mathematically investigated Kepler's law by using his law of motion (force = mass × acceleration) to determine what type of "force" can produce such a trajectory. He found that there exists a gravitational force between two objects, which was proportional to the inverse square of the distance between the two objects (Figure 3). Despite the lack of experimentation, the truthfulness is proved because the law could reproduce the trajectory precisely.
Figure 2. Forward science (a) and reverse science (b)
Figure 3. Scholars who contributed to the discovery of universal gravitation
(2) What is the truthfulness of data-driven science?
As aforementioned, the criterion of truthfulness in hypothesis-driven research is corroborated by the “falsifiability of the hypothesis by the experiment” [1]. Accordingly, it is important to describe the criterion of truthfulness in data-driven science. For this, the “autoencoder network” model of deep learning is instructive. Figure 4 shows a generalized version of the autoencoder network model that considers a symmetrical neural network (decoder) in which the network weight combination is the transpose of the encoder network.
Specifically, input data are contracted into a set of variables of the lower-dimensional latent space by a multi-layered autoencoder network. They are encoded into contracted knowledge (truth), which is then decoded by the transposed weight matrix, and thus reproduces the input data. This reduction-reproduction process can be achieved using a multilayer autoencoder-decoder network. For example, if many Rembrandt’s paintings are inputted and their essential features are contracted in the low-dimensional latent space, a similar, but novel, Rembrandt’s painting can be reproduced by defining a certain point in the latent space and decoding it. Accordingly, this demonstrates that the truth criterion in data-driven science is whether reality can be reproduced from new knowledge (the contracted truth). Similarly, Newton's law of universal gravitation is considered correct since it can accurately reproduce the orbits of the planets.
Regarding the future form of medical science, we would say, since unsupervised learning requires a large amount of data, it is best to use both “hypothesis-driven” and “data-driven” approaches to solve problems; it is necessary for humans and AI to collaborate to solve problems and discover "co-created knowledge”.
Figure 4. Truth ontology of data-driven science
3. Innovations in medical practice and healthcare during the third revolution of medicine
It is important to discuss future innovations in healthcare practice during the era of the third revolution of medicine. Recent advances in biomolecular research methods, including next-generation sequencers, have allowed significant progress in molecular biomedicine, which is expected to continue in the future.
The recent introduction of cryo-electron microscopy has allowed the clarification of protein 3D molecular structures; moreover, gene editing technology has facilitated the elucidation of various biomolecular mechanisms. Specifically, recent advances in single-cell omics have provided information regarding diverse cell behaviors across numerous biological phenomena, including cancer [6]. Additionally, the revolution of sequencing technology and its subsequent application in genomic medicine and precision oncology further demonstrates the continuous development of biomolecular measurement technology and the resulting elucidation of various biomolecular mechanisms.
Another expected continuous trend is the advancement of "information technology in medicine," or DX (digital transformation) of healthcare. This is exemplified by the recent rapid progress in mobile health, including therapeutics, which is attributable to advances in smart media. Another medical application of AI is the use of clinical annotation and decision support systems, which use genome-omics information to facilitate precision medicine in clinical practice. Taken together, it is important to identify remarkable progress in molecular bio-instrumentation technology as well as the application of information technology in medicine, including medical AI when discussing the future of medicine.
Since the development of genome-omics medicine following the introduction of next-generation sequencers and other advances has been described elsewhere [7] [8], the present study discusses recent noteworthy trends in the DX of medicine.
3.1 Recent developments in mobile health
There have been recent notable developments in mobile health, which is brought about by the combination of (1) rapidly developing smart media; (2) advanced "wearable biometric sensors" developed through non-constrained biomonitoring technology; and (3) the concept of “participation of patients”, becoming mainstream of health care. In the 1970s, the idea of patient participation was first introduced as "Quantified Self movement (QS)" on the west coast of the United States, where individuals should assess their physiological status in order to manage their health and disease treatment by themselves (Figure 5). Moreover, the concept of “patient-centered care” has become prevalent in healthcare activities in the U.S., where patients play an active role in measuring their own physical information, gathering medical knowledge, and selecting treatment strategies.
Figure 5. Wearable devices at the onset of the Quantified Self movement in the U.S.
Notably, recent advances in mobile health have moved even towards the field of therapeutics. Specifically, mobile health now seeks to use the measured physiological status to yield medical advices and treatment recommendations. This trend, which is characterized by information-based treatment, has recently been termed as "digital therapeutics" [9].
Treatment of behavioral disorders or chronic lifestyle diseases require behavioral changes in daily life, such as diet and exercise, which are recommended by using information regarding one’s health status. For example, the “Blue Star” system (WellDoc, inc.) is a smartphone application that provides treatment support to patients with type 2 diabetes based on their measured blood glucose levels, medication status, and lifestyle habit. It was found to significantly improve HbA1c levels of these patients and was approved by the FDA in 2017. Moreover, Akili Interactive Labs, inc. developed software for treating children with attention-deficit hyperactivity disorder (ADHD) by helping them focus on playing a video game, which is similarly approved by the FDA (Figure 6(a)).
Currently, there are more than 100 digital therapeutics approved by the FDA, with the most common targets being mental health disorders, including substance addiction, as well as chronic diseases, including diabetes, hypertension, and arrhythmias (Figure 6(b)) [10]. Healthcare smartphone apps have so far mostly focused on health maintenance; however, they has been a recent remarkable increase in apps seeking to treat and cure diseases.
Figure 6. (a) Representative digital therapeutic apps and (b) disease-specific apps by digital therapeutics
Pharmaceutical companies are embracing the "beyond pills" concept and are establishing "non-pharmacological intervention (NPIs)” divisions in collaboration with IT companies to achieve this goal through digital therapeutics. Moreover, there has been increasing research on when to provide advice or behavioral support in order to maximize the effectiveness of mobile health care. Accordingly, to assess the effectiveness of mobile-provided healthcare, it is important to randomize the intervention methods, timing and durations to examine which is effective. It brings about the new investigate method which is reoffered to as “micro-randomized trial method (MRT) [11]. Accordingly, MRT allows the determination of an optimized “just-in-time adaptive intervention”.
A mobile health intervention that measures daily-life disease status (self-monitoring) to inform drug intake (medication self-titration) has been shown to be more effective than conventional treatment of chronic diseases by JAMA [12], which is made of clinic visits at 1- or 2-month interval to receive antihyperglycemic or antihypertensive medications.
3.2 Notable developments in bioinformatic (precision medicine) real-world data
(1) Advantages of “real-world data”
There has been a notable trend in use of real-world data. The emergence of big data in medicine has raised questions regarding the validity of the concepts of evidence-based medicine (EBM) and Randomized Controlled Trial (RCT) in various aspects. For example, RCT includes participants based on pre-established “inclusion criteria”, which would yield a study population that deviates from the one in real-world clinical settings [13] (Figure 7). Specifically, clinical trials in the U.S. have almost exclusively adopted white adults, with few studies adopting elderly individuals, pregnant women, or black adolescents. For example, 79.2% and 7.4% of participants in FDA-approved drug trials were White and Black, respectively [14].
Figure 7. Differences in the age distribution of patients with hypertension between clinical trials and daily medical care
In the big data era, since it is easy to collect large data, there would be a reduction in sample collection costs. Specifically, large-scale real-world data can be directly collected without considering complicate statistical sampling schema. For example, real-world data is now used to perform post-marketing surveillance regarding the side effects of drugs. When drug named Vioxx (Merck & co.), a non-steroidal anti-inflammatory drug, was suspected to cause sudden death and myocardial infarction, multi-center data obtained from the collected EMRs of the Kaiser Permanente hospital group (real-world data) were used for post-marketing surveillance. Vioxx was shown to increase the incidence of sudden death and myocardial infarction, and thus it was withdrawn from the market.
Based on the 21st Century Cure Act, which was enacted in December 2016, the FDA promotes the use of real-world data in clinical trials. Specifically, this law (1) clarified the regulatory scope of the FDA and recommended the use of mobile medical apps and digital medical products; further, (2) it encouraged the active use of real-world evidence (RWE) to ease the approval regulations. Accordingly, the FDA issued guidance (Figure 8) in August 2017 to allow the use of real-world data and RWE as reference materials for new drug applications [15].
Figure 8. FDA guidance for use RWE (2017)
Additionally, the FDA issued guidance for the FDA RWE Program [16] in 2018 (Figure 9), which describes the verification methods for safety and efficacy to provide a framework for the use of RWE in new drug development, post-marketing surveillance, and indication expansion according to the U.S. regulations (21st Cent Cure Act and FD & C Act). This guidance allowed approval of several new drugs using RWE under relatively less strict conditions. The subsequent section describes a case of the use of real-world data in clinical trials.
Figure 9. FDA real-world evidence program (2018)
(2) Drug trial using advanced bioinformatic real-world data as the control group
Conventionally, clinical trials set up two unbiased groups and assess the effectiveness of the investigational drug through comparison with the control drug. However, for extremely rare diseases or drugs with high urgency for regulatory approval, establishing an appropriate control group may not be feasible.
For example, TRK (tropomyosin receptor kinase) is a receptor-type tyrosine kinase involved in neuronal differentiation and maintenance. The NTRK gene, which encodes TRK, can fuse with other genes to produce fusion proteins, which strongly influences cell growth and cancer development. Accordingly, it is considered a driver gene for cancer. However, there is an extremely small number of patients with NTRK fusion gene-positive cancer.
Another gene causing a similar fusion is the ROS1 (c-ros oncogene 1) gene, which is a receptor tyrosine kinase of the insulin receptor family. The ROS1 fusion gene is a potent oncogene that is extremely rare (about 1%) in advanced non-small cell lung cancers (NSCLC). Although Crizotinib, a kinase inhibitor, has shown very high efficacy in clinical trials, tumor resistance occurs within several years so that the patient’s condition may regress. Accordingly, the FDA granted breakthrough therapy designation for the development of drugs for NTRK or ROS1 fusion genes-positive solid tumors. Entrectinib was expected to be effective for the treatment of NTRK, ROS1 fusion gene-positive solid tumors. The FDA approved the following single-arm study that used real-world data as a control arm.
The trial design was as follows [17]:
Drug arm: N = 94, ROS1(+), non-small cell lung cancer (NSCLC); drug: Entrectinib
Data: Integration of three phase 1/2 trials for single arm
Control arm: N=65, ROS1(+), non-small cell lung cancer (NSCLC); drug: Crizotinib
Data: from the Flatiron Health real-world database
Flatiron Health is a company engaged in a database service that provides structured and anonymized patient information from its OncoEMR (precision oncology EMR). Accordingly, this clinical trial real-world data was extracted from the aforementioned Flatiron database regarding patients prescribed with Crizotinib.
(3) Oncology EMR systems
Oncology EMR systems, which are specialized for precision cancer treatment, have been marketed by several companies. In the U.S., the term “community oncology” is used to describe a unique cancer treatment service system in which specialized clinics, rather than big hospitals or cancer centers, are responsible for advanced cancer treatment [18]. Cancer treatment is a long-term, individualized, frequently changing process. Moreover, distance barrier can impede access to a specialized cancer hospital and center. Accordingly, the establishment of numerous vicinal community oncology offices with sufficient state-of-the-art equipment and personnel may improve cancer patient care. In the United States, 55% of patients with cancer are treated in community oncology settings; for example, in Texas, there are 400 community oncologists practicing in 25 offices. Additionally, oncology EMR systems support state-of-the-art treatment practice and provide updated knowledge and information within community oncology settings. This updated knowledge includes the guidelines and chemotherapy standard models (templates) provided by the National Comprehensive Cancer Network (NCCN), which is a collaboration of 31 leading cancer centers in the U.S. engaged in cancer patient care, research, and education. Furthermore, the oncology EMR system allows the latest classification of cancer progression state according to the American Joint Committee on Cancer (AJCC), which is a collaboration of 20 cancer-related academic societies in the U.S., using an easy-to-understand display of cancer treatment workflows, various templates, and various indicators and markers.
Notably, the clinical practice data regarding cancer treatment documented in oncology EMR are compiled in databases of real-world data on precision cancer treatment. Additionally, these databases are manually supplemented with other important data or indicators that are not included in the oncology EMR system.
Advances in cancer treatment have led to extremely diverse approaches; moreover, the efficacy and prognosis of anticancer drug therapy varies according to the affected site, cancer stage, other cancer progression signs, and type of concomitant therapy. Given these diverse conditions, a dynamic treatment plan based on experienced real-world data should reflect the complex conditions. Accordingly, it is important to accumulate oncology EMRs in order to establish a precision oncology database that reflects the clinically complex therapeutic course of patients with cancer.
3.3 Future Developments in Medical Artificial Intelligence
AI is being applied across various medical fields, including medical imaging, which is now being applied most successfully, as well as genome-omics life sciences, medical informatics, mobile health, and public health [19].
The success of unsupervised learning by AI is dependent on the availability of the structured high-quality data. Since medical images provide high-quality structured signals, there are large amounts of accumulated data within clinical image databases. Moreover, the “convolutional neural network” has provided machine learning methodology of the image data. These factors brought about the high initial success of AI application in imaging diagnostics.
(1) Application of AI in interpreting biomedical multidimensional time-series signals
The next promising area for AI application is the interpretation of biomedical time-series signals to identify dynamic course of changes in disease progression. AI has already begun being applied to interpret biomedical signals such as electrocardiography. Moreover, in the circumstances such as intensive care unit (ICU) or coronary care unit (CCU), simultaneous measurement of multiple biological time-series signals should be treated. There is currently ongoing research on the use of these multidimensional time-series data to understand dynamic disease course and predict prognosis in the disease status [20]. Highly effective AI methods for analyzing time-series data, including recurrent neural networks [21] and long short-term memory neural networks [22], indicate the utility of AI in this research field.
(2) AI-based dynamic and qualitative-systemic understanding of relationships among clinical events
Next, AI is expected to model the relations among the clinical events described in EMRs by using” knowledge graphs” (event graphs) [23]. In the EMR, clinical events are described using natural language and are not often structured. Therefore, to understand the relationships among clinical events, it is important to elucidate these correlations in terms of the clinical relation phenotypes shown in the knowledge graphs. To achieve this, natural language processing must reach a certain degree of proficiency. Natural language processing is predicted to reach the human level, seeing the rapid advances of large language model (LLM), in the near future [24]. Therefore, the qualitative relations among clinical events will be able to be expressed in knowledge graphs in the near future.
Typically, there can be temporal (cause-effect relation) or dependency (master-slave relation) relations among clinical events, which AI should be able to understand. Accordingly, AI is expected to be able to accurately understand the systemic and dynamic changes underlying behind the disease progression.
The aforementioned knowledge graphs can facilitate medical diagnosis based on temporal understanding of changes of the patient's condition. This process will require symbolic reasoning. Therefore, reasoning on the disease progression of patient will eventually involve a combination of reasoning by connectionism (e.g., neural networks) and reasoning through symbols (symbolism). Taken together, AI will play a crucial role in understanding the overall dynamics of disease progression, predicting the disease prognosis. We could call this way of practicing medicine “predictive control medicine” (Figure 10).
Figure 10. Future artificial intelligence: temporal and systemic understanding of dynamic trends in diseases to realize predictive control medicine
3.4 Future of medical care in the era of the third revolution of medicine: The true digital transformation (DX) of medicine
Based on the aforementioned trends driving the future medicine, this section will summarize three big revolutionary changes expected in the future of clinical care.
(1) Large-scale accumulation of EMRs for precision medicine (PM EMR): Building a nation-wide information system to support popularization of the precision medicine based on a “PM real-world database” infrastructure
Although there is some significance to ordinary digitalization and sharing of conventional EMRs, typically exemplified by MID-NET [25], it cannot drastically impact the future medicine. Instead, there is a need for specialized deep EMR systems, utilizing the extensive bioinformatic information provided by advanced bio-instrumental methods. It supports the implementation of precision medicine (PM).
In this paper, precision oncology EMR and genome EMR are collectively called as PM (precision medicine) EMR. To facilitate future digital healthcare, there is a need for PM EMR utilizing advanced molecular information as well as a medical ordinary care support system.
(2) PM mobile health as a mainstay in chronic disease treatment
As aforementioned, mobile health is expected to become a mainstay of treatment requiring changes in daily-life activities, such as the treatment of chronic diseases and mental health disorders. However, as already mentioned, mobile health should advance towards the inclusion of bio-molecular information for conquering the defect of ordinary mobile health, which is the superficial observation with only using classic physiological items. It is expected that the iPOP using various omics measurement proposed by Snyder [26] will be implemented in the future without exorbitant costs.
Additionally, mobile health is expected to improve the prediction accuracy of short-term disease progression by integrating advanced omics tests, such as liquid biopsy. Future form of mobile health will be achieved by incorporating the multi-omics data, which will be called PM mobile health.
(3) AI understands the dynamic and qualitative-systemic relation among clinical events to bring “qualitative system-biologic” understanding of disease progression and enables “predictive control medicine”
As aforementioned, advances in natural language processing will allow AI to identify qualitative temporal-systemic dynamics controlling the relations among the clinical events documented in PM EMR and to construct a knowledge graph representing course of disease. Accordingly, AI will be able to capture the qualitative dynamic and systemic trends of disease progression and facilitate “predictive control medicine”.
5. Summary and conclusion for future medicine
So far, we have been predicting and discussing the future form of the next generation medicine. Here, we would summarize our prediction and discussion so far.
At first, we discuss about the driving forces for the next generation of medicine. One is inevitable advance of molecular precision omics which is brought by rapid advance of bio-instrumental methods. It will be bringing about brand-new knowledge of biological mechanism and disease, enabling the precision medicine (PM). The second is inevitable progress of information technology including AI, bringing about dynamic and qualitative-systemic understanding of course of disease, enabling “predictive control medicine”.
Those two inevitable driving forces changes drastically present form of medicine.
As for the methodological aspect of medical science, as we have discussed in section 2, it will bring about the data-driven medical science. Big data and inductive AI method will support this data-driven medical science, which would be called “reverse science”, as described in section 2.
As for the practical aspect of medicine, such as revolution in clinical practice and health care, the most remarkable characteristics of future medicine is digitalization of clinical medicine and healthcare in the era of the third revolution of medicine. Those are the introduction of PM mobile health and PM real-world database as described section 3, which combine extensive biomolecular information with innovative information technologies in smart media and AI.
Together with deep molecular knowledge which brings about the precision medical research, information technology can allow the construction of nationwide database infrastructure, which yields a decision support system that facilitates the implementation of precision medicine in routine clinical practice. In conclusion, the popularization of “precision digital medicine” which is brought by deep molecular omics knowledge and information method innovation by big data and AI, is expected to eradicate disparities in the level of medical care in the future.
