The Artificial Intelligence in Integrated Medicine and Healthcare: Back to the Future

Abstract

The artificial intelligence (AI), in a simple way, is another approach to enhance the powers of normal innate intelligence. AI is a product of continued progress in computer science and technology with massive research and development efforts in industry, climate, social and media and health sectors. Exponential powers of AI have opened doors for unimaginable applications in medicine and personal healthcare. Novel AI tools and algorithms are applied with past medical discoveries and inventions with unpredictable sophistication in precision diagnostics, personalised pharmacotherapy with new drugs, life-saving measures in acute medicine and effective measures to safeguard personal, family and community health. The AI journey from ‘bench to bedside’ is in progress with many bridges and tunnels to cross. AI is the new ‘kid in the block’ with promising tools to bring about revolutionary enhancements in future-integrated medicine and healthcare. AI has the potential for successful global implementation of the ‘traditional and complementary integrated health (TCIH)’ programme.

Introduction

All biological creatures possess variable intelligence, judged by their innate ability to understand the surrounding things and information. Humans have an added advantage to express and display intelligence in different ways, collectively referred to as wisdom. In this context, artificial intelligence (AI) would include any other means to enhance the normal intelligence of an individual or group of people. Thus, clearly, AI should include commonly used audio-visual devices and appliances in the home, at work and in public places. Perhaps the best example should be the calculator that allowed fast and absolutely correct mathematical data, which became an essential tool for students, teachers, statisticians, researchers and many others. Arrival of computers, including the mainframe, personal desktop or laptop computers, revolutionised the power to understand, retrieve and analyse information to unexpected high levels. We are all accustomed to harnessing and utilising gains of computational science and technology practically every single minute of daily life. In this context, AI has huge potential for accelerating all fields of science and socio-economic gains.^[1] There is, as yet, no universally agreed unitary definition of AI since it refers to complex, multi-faceted and dynamic applications of sophisticated theoretical mathematics and computational technology.

Medical science, as in other fields of science, is dynamic with numerous examples of novel inventions and applications. Each new discovery or application offered an opportunity for enhancement in the diagnostic or therapeutic capability of the medical practitioner. Marie Curie’s discovery of X-rays is perhaps the best example, followed by ultrasound, computerised tomography, magnetic resonance imaging and many more new imaging advances. Each one of these is comparable to the present-day AI. Digital audio-visual technologies now allow the medical and health practitioner the opportunity for remote expert consultations and actions for the patient. This is a perfect example of AI in our lives. People and society at large are getting accustomed to the benefits of AI with huge expectations for the future.^[2]

The current review examines the inroads of AI and its impact on the current and future practice of medicine, including shaping the overall healthcare. Modern medicine is already employing advanced stages with cutting-edge radio-diagnostics, advances in laboratory diagnosis, precision gene-sequence level diagnosis, personalised targeted gene-molecular therapy, radiotherapy and immunotherapy for treatment of cancer and many other conditions.^[3] All these modalities are indispensable without the applications of computational science and technology. AI would be just one more step for future advances in all aspects of medicine and healthcare. Future medical and healthcare practitioners will be equipped with AI-generated ultimate precision diagnosis, new personalised drugs, novel invasive or non-invasive surgical interventions, personalised healthcare surveillance, early warning systems of microbial endemic or pandemic disasters and future community and population health planning.^[4] The traditional and complementary healthcare (TCIH) initiatives led by the World Health Organisation (WHO) are linked with AI applications. This is closely linked with India’s AYUSH (Ayurveda, Yoga, Unani, Siddha and Homoeopathy) traditional healthcare mission. The future of medicine and healthcare will require the integration of modern medicine with TCIH and AYUSH. AI tools are anticipated to find inroads for managing the ‘big data (BG)’ essential for precision and personalised integrative healthcare.

Essential Technical Requirements

The term artificial intelligence or AI was first used about 70 years ago when the need for creating intelligent machines was conceptualised.^[5] Essentially, AI analyses and emulates complex human thought processes or information to accomplish tasks or achieve objectives which would otherwise require huge resources and/or time. Technically, AI includes patterns of AI (weak or strong), machine learning (ML), deep learning (DL) and BG.

There are two AI patterns: ‘Narrow or weak’ and ‘strong or general’. The narrow AI describes the application of learning algorithms for specific tasks, such as language processing tools (e.g., ChatGPT), virtual assistants (e.g., Siri, Alexa or online navigation) and image recognition. All current AI systems are domain-specific and thus are included in the weak or narrow category.^[6] The future format of AI is aimed at a sophisticated level, surpassing human intellect with the abilities of agile reaction and forward planning. This is labelled as ‘strong’ or ‘general’ AI. To some extent, it would be similar to humans but would offer added potential for AI dominance over human intelligence.^[7]

The ML refers to computer systems that learn automatically from experience without being specifically programmed.^[8] ML systems identify patterns in datasets and create an algorithm for further applications in similar datasets. The DL is the outcome of ML, which uses artificial networks to recognise patterns in data and provide a suitable output. DL layers algorithms into an artificial network inspired by the human brain, similar to neural signal transmission from one step to another, similar to the neuronal transmission of neurons.^[9]

The ‘BG’ refers to datasets too large and complex to be processed using conventional data processing systems. Data can be ‘big’ in terms of the volume, velocity or both. Specifically designed algorithms adapt and improve without using traditional algorithms or human intelligence.^[10] In brief, technically, AI, ML and DL are closely related and part of the whole AI infrastructure ( Figure 1 ).^[3]

Figure 1.

Venn diagram demonstrating the relationships between AI, ML and DL^[3]

Datasets for Artificial Intelligence

In addition to sophisticated high-powered technical infrastructure, generating, storage and fast retrieval of BG is the key for any AI platform. Currently, there are numerous datasets in all key areas such as socio-economic information, complex climate data, all forms of biological life, including microbes, all branches of science and technology, human and veterinary medicine and healthcare. All these AI parameters require BG analytic software for useful AI applications.^[10]

Most current AI applications regularly access large datasets governed by different public and private institutions and organisations. Major datasets, such as Google, chatbots and electronic medical records (EMRs), are regularly used by millions equipped with small electronic devices, for performing a wide range of small AI functions.^[11] Researchers and AI professionals employ various mechanisms for big data analysis.^[12] Advanced BG analytic approaches have the potential to improve medical outcomes and population health through biomedical research efforts, specifically organising and conducting clinical trials for good clinical practice and new drug discovery.^[13]

AI in Clinical Diagnostics

Clinical diagnostics involve detailed and deep phenotyping with judicious use of personal medical history, interpretation of family history, along with considerations for relevant ecological and lifestyle patterns. In addition to conventional physical parameters, phenotyping also includes all formats of imaging data, outcomes of biochemical, immunological and histopathological investigations and inclusion of clinically actionable genomic data. A typical digital health system would be expected to include all this information in an individual patient’s EMR. Rapidly emerging and expanding datasets are being generated for each element of EMR that would be humanly impossible for any level of medical and healthcare professionals to make any logical use in clinical diagnostics and thus effective and safe patient care. It is widely anticipated that the power of AI is already being harnessed by diagnostic and interventional radiologists.^[14] There are several AI algorithms available facilitating high-order clinical diagnostics, especially imaging-based and genomic diagnostics.^[15,16]

AI in Nutrition and Pharmacotherapy

Adequate and optimal nutrition are essential requirements for human health and sickness. This would depend upon food availability and selection in keeping with geographic locations, climate and ethnic preferences. Food processing and cooking as per variable ethnic customs and cuisines are essential components of the personalised diet. Finally, genetic and genomic factors govern the digestion, absorption and metabolomic integration of basic dietary components (carbohydrate, fat and protein) along with essential nutrients.^[17] Food intolerance and specific food allergy are cross-cultural and global challenges that are managed by a range of local community and specialist practitioners. Advances in nutrigenetics and nutrigenomics, including epigenetics, have helped unravel the complexity of food intolerance and allergy.^[18]

Malnutrition, as judged by a broad spectrum including undernutrition, overweight and obesity, is a global problem. Several approaches and avenues are available that are ethnic specific and personalised. Developing personalised diets requires a range of food technologies, including computational hardware and dedicated software for super-connectivity and AI. Technologies developed in other fields can be used in food science research, collectively referred to as ‘foodomics’. Such technologies include genetic analysis, whole genome sequence analysis, nutrigenomics, metabolomics, nutrigenetics, nutriepigenetics, microbiome technology, sensomics and culturomics.^[19] Clearly, large datasets are necessary for logical and practical AI applications for achieving ethnic specific and personalised dietary formulations essential for optimal health.

Every medical and health practitioner, whether practising western-style medicine (allopathy) or ethnic traditional medicine (Ayurveda, Chinese, Unani and others), is expected to ensure personalised, safe and effective pharmacotherapy. Thus, any pharmacological agent (a drug, a specific herb or a herbal cocktail) should have clearly defined and tested ingredients. Most drugs are either derived from a plant or herb, such as digitalis from foxglove, but increasingly many approved drugs are the product of scientific drug discovery and a complex drug development process. Finally, what happens at the cell, tissue, organ or systems level is governed by a range of complex interconnected pharmacogenetic and pharmacogenomic networks. At present, datasets exist generated by food and science technologies. It is not humanly possible to retrieve, select and prescribe the most effective drug or herbal remedy for an individual patient. Clearly, there is a need for food and nutrition-specific AI and ML tools to equip the health practitioner for safe, effective and efficient precision and personalised pharmacotherapy.^[20]

AI in Surgical Interventions

Apart from the basic clinical surgery, modern surgery includes many avenues driven by sophisticated digital technologies. Minimal invasive surgery (MIS), robotic and radiological interventional surgical interventions are commonly used for routine and complex surgical treatment.^[21] AI and ML have the potential in surgery since modern surgery is more complex for human control.^[22] Surgeons are employing AI in many operative procedures, aided with robust analysis of preoperative information, imaging and navigational techniques. Novel DL algorithms enhance the power of AI and ML with anatomical classification, detection, segmentation and picture registration. The power of radiomics, an essential tool for computer-aided intraoperative guidance for MIS, is likely to be augmented with AI.^[23] Maxillofacial surgeons have begun to exploit AI and ML for efficient use of robotic assistants in complex intraoral MIS procedures.^[24] Deep-seated tumours, such as intracranial or intrapancreatic, can be precisely detected, diagnosed and successfully treated with an AI-assisted deep convolutional network-based approach.^[25,26]

AI in Genomic and Omic Applications

The fast-emerging field of genomic medicine offers the clinician to practise all key areas of medicine: Precision diagnosis, personalised care with targeted therapeutic interventions and prevention for long-term health consequences or recurrence in asymptomatic close family members.^[27] Novel genomic technologies, specifically the microarray chromosome analysis, next generation genome sequencing based genetic and genomic diagnosis, ribose nucleic acid (RNA) sequencing, mitochondrial genome analysis and complex epigenetic analysis (methylation profiling, histone modifications and others), offer the highest possible sensitivity and specificity for safely practising precision and personalised evidence-based medicine.^[24] The integration of genomic laboratory techniques with omics-based techniques has exponentially enhanced the power of precision and personalised clinical practice.^[28] There are a number of life sciences fields with the omic suffix that include transcriptome, proteome, metabolome, methylome, microbiome, exposome and many others, which aim to characterise and study the collection of proteins, metabolites, microbes, environmental chemicals and transcripts expressed in an organism. Omic technologies are at different stages of development in terms of their translational and clinical application in precision diagnosis and personalised therapy.

Genome sequencing techniques, particularly whole genome sequencing (WGS), generate a huge amount of an individual’s raw genomic data with variations in six billion nucleotides, including single-nucleotide polymorphisms, copy number variations and evolutionarily conserved sequences. Typically, WGS in a small nuclear family of four members will generate genomic information in 24 billion sequences. Clearly, big genomic raw data would require a very large digital storage facility, sophisticated fast retrieval computational hardware and software, specifically designed algorithms for targeted genomic analysis or information applicable in many aspects of experimental and applied clinical medicine. Many such tasks would require an enormous workforce and resources for efficient and effective applications, such as clinical medicine. The individual genomic data is distinctive and heterogeneous in terms of specificity and complexity that necessitate the need for specially manufactured computational tools and algorithms that leverage advances in ML and neural networks. AI has the potential to carry out these tasks with speed and accuracy.^[29]

Successful integration of AI with genomic medicine requires dedicated infrastructure development of a comprehensive genomic medicine unit alongside the AI-based ML neural networks facility. In addition to the genomic data, the ‘BG’ would include clinical data, electronic health records and information on the healthcare systems ( Figure 2 ).^[3,30]

Figure 2.

Integration of clinical genomics and healthcare information with AI^[30]

Personalised cancer care is considered to be a major challenge amongst many challenging fields in precision medicine. AI has emerged as a powerful tool in the field of cancer genomics, revolutionising our understanding of cancer and its treatment.^[3] Integration of AI approaches, specifically ML, DL and natural language processing, offers platforms for meeting challenges of BG and transforming into clinically actionable knowledge, as the core foundation of precision medicine. The current status and future directions of AI applications in cancer genomics and clinical oncology are promising for precision cancer care. New algorithms are now available for AI applications in cancer genetic testing, precision cancer diagnostics and targeted therapeutics.^[31,32]

AI and Critical Care Medicine

Critical care medicine is heavily dependent upon fast, accurate and unambiguous information for swift and precise care of critically sick patients. Recent advances in accessing electronic health records, digital information portals and fast digital devices have made a significant impact on the patient survival in major critical conditions such as sepsis, trauma and major postoperative care. During the last decade, awareness and recognition of AI in critical care have significantly increased.^[33]

Anaesthesia is a complex medical speciality that is the backbone of surgical specialities, with the potential of major applications in non-surgical fields such as pain management, palliative care and critical care. Anaesthetists play a crucial role in critical care medicine, working alongside emergency physicians, nurses and technicians. Recent developments in anaesthesia include advanced clinical decision support tools based on ML with increasing efficiency and accuracy. New promising opportunities for anaesthesia are on the horizon with AI developments and DL.^[34]

Amongst many promising applications of AI in critical care, management of sepsis is by far the most effective application.^[35] AI offers promising solutions to improve sepsis care through earlier microbial detection, risk stratification and precision-personalised management. AI ‘BG’ for sepsis care includes clinical and biological data, specifically of the microbiome and cascade of molecular biomarkers. Clearly, it includes many challenges for the AI team, including data quality, dedicated algorithms, limited external validation and ethical considerations.^[36] Other major AI applications in critical care medicine include trauma and postoperative surgical care.^[37] However, successful applications of AI and ML in emergency critical care would require a systematic and integrative healthcare ecosystem to ensure transparency, patient safety, accountability and ethical-legal-social issues.^[38]

AI and Ethics

Fundamental principles of ethics in medicine include benefit to all (beneficence), cause no harm (non-maleficence), ensure personal respect and dignity (autonomy) and act within the law (justice).^[39] AI and ML in medicine and healthcare have triggered ethical debate and discussion.^[40] Main areas of healthcare with the potential of AI include prenatal and paediatric medicine, internal medicine and clinical pharmacology, radiation oncology, psychiatry, ophthalmology and clinical surgery.^[41,42] Currently, there is no specific guidance for practitioners on the ethical aspects of AI in medicine and the integrative healthcare system. Since AI and ML are continuously evolving, achieving maturity and becoming extremely reliable, the accountability for the construction and deployment of AI tools, such as algorithms, should tend to fall on healthcare practitioners, institutions such as hospitals and clinics and healthcare systems.

AI in Global Healthcare

AI and related technologies are rapidly moving into modernising healthcare systems that can predict, grasp, learn and act to complement and carry out humanly difficult complex tasks.^[38] However, there is some disquiet around due to the hype surrounding fictional expectations from AI in healthcare. Nevertheless, AI and ML have huge potential in three healthcare sectors- AI-led drug discovery, clinical trials and patient care. Scientists and the pharmaceutical industry engaged in new drug discovery and development have benefited from AI input in healthcare research and development by speeding up the drug discovery process and automating novel drug target identification. AI can eliminate and select the most pertinent, efficient data selection and monitoring methods, saving valuable time and resources. Evidence emerging from AI applications for the COVID-19 pandemic sets out an example for AI-led healthcare.^[43] There is no doubt that AI tools are capable of handling exponential volumes of data and producing highly accurate results. The future AI-led healthcare system would include many facets offering the highest possible clinical outcomes with improved quality of life ( Figure 3 ).^[30]

Figure 3.

The model of AI in the future healthcare^[30]

Integration of TCIH has generated a lot of interest and debate in many countries. The WHO agreed on the implementation of TCIH in its 2019 Charter and set a completion date of 2023, which has now been moved further to 2027. WHO has acknowledged and established a dedicated centre in the State of Gujarat sponsored by the Government of India as part of the traditional medicine and healthcare strategy comprising AYUSH. The potential of AI and ML has been critically explored for the successful implementation of the WHO’s traditional medicine initiatives.^[44] Has the WHO formally included India’s AYUSH AI innovations in its TCIH recommendations? India is the first country to launch a Traditional Digital Knowledge Library. India is leading the collective drive to harness the potential of AI to advance and amplify the strengths of its AYUSH systems to position the country as a global leader in digital healthcare innovation and the integration of traditional medicine.^[45] It is extremely important to have faith and trust in AI-driven healthcare systems for merging traditional medicine and health systems with modern medicine. AI holds great promise in enhancing healthcare delivery, personalising treatment plans, preventive care and patient engagement.^[46] Close and dynamic collaboration is crucial between AI experts, TCIH practitioners and policymakers for harnessing the full potential of this integration.^[47]

Conclusions

The dynamism of medical sciences is undisputed. Historically, several past scientific and technical inventions led to major new applications to bring the modern science of medicine for the benefit of mankind. AI revolution is yet another technical advancement that is likely to enhance precision, personalised and preventive future medicine and healthcare. Concerted global efforts are required for developing efficient and effective modes of AI in shaping the future practice of clinical medicine and preventive healthcare. The science of medicine is on the verge of reinvention with new AI tools. All medical and health practitioners need enhanced skills for blending AI with old, practical and pragmatic ways of thinking about the bedside. AI is already with us, irrespective of geographic location, climate, people variation and socio-cultural environment. AI projects can offer major inroads for the successful implementation of the WHO-led ‘TCIH’ programme. In this context, India’s AYUSH traditional healthcare mission is leading the implementation of WHO’s TCIH. It is widely acknowledged to harness AI tools for the successful implementation of TCIH and AYUSH. This is the AI era that has the potential to bring global changes with invariable impact on all aspects of human life, including other animals and the climate. However, the society and the state need to carefully set out statutory guidance for the implementation and governance of AI for achieving and harnessing the huge potential of digital health.