Reimagining healthcare with AI

Targeted therapies, personalised medicine and other customised treatments that aim to match an individual’s genetic profile are part of the newly emerging world of healthcare that challenges the common notion of ‘the balance of risk’ when it comes to recommending or administering a drug or treatment.

Understanding why one in 100 people might suffer adverse reaction to a medicine can help us create a stronger patient experience tailored to individual need, thus shifting offerings from blanket therapies to personalised medicines. Having insights to make better decisions when it comes to reforming and improving the overall value chain of health systems at pace is a fundamental component.

Today, the global spend on health is estimated to be around 10% of global GDP, or $8.3tn for the next three to five years, according to the World Health Organisation (WHO).

Public health systems continue to experience insurmountable pressures as a consequence of the Covid-19 pandemic. Hence, there is a need to accelerate change within public and private healthcare providers and across their ecosystems, forcing them to innovate, adapt and respond to the many dynamically shifting needs and demands within a shorter period of time.

Healthcare is, unquestionably, a sector that is in need of innovation. Health plans, health providers, life and bio science businesses, as well as governments, are all facing increased costs and often yielding inconsistent outcomes.

Trend forecasters have highlighted many technologies that are likely to transform the healthcare landscape as we know it. These include, amongst others, technologies such as artificial intelligence (AI), internet of medical things, telemedicine, big data and analytics, cloud computing, immersive technologies, genomics, blockchain, 3D-printed devices and mobile health.

Digital technologies form the backbone of transforming healthcare practices and optimising their systems and responses, thereby improving patient outcomes and lowering overall costs.

The shift in user behaviours to be more accepting of, or even favouring, digital health solutions is also a significant driver that is precipitating the creation and reinvention of healthcare offerings.

Whilst the digital economy is impacting every sector, and represents 22.5% of global GDP, the potential of digital technologies to unlock value, particularly in the health and pharma-related industries, remains far from being fully understood, or exploited, according to research by Accenture and Oxford Economics in November 2020.

Identifying digital healthcare opportunities across the entire patient pathway, from primary prevention and screening through diagnosis to treatment and monitoring, requires a clear assessment and distillation of where the ‘pain points’ and potential ‘gains’ are within the pathway and how digital health solutions might address them.

The confluence of AI with other advances in digital and biological sciences is undoubtedly presenting an exciting wave of innovations. For example, the speed with which scientists have been able to sequence the Covid-19 virus’ genomic structure and its various mutations, in weeks rather than months, is a testament to the new converging advances in computing, large data analytics, AI - including machine learning (ML) - and biological engineering.

The decreasing cost of DNA sequencing – which is currently less than $1,000 and likely to drop to below $100 within this decade – is increasing our ability to understand and engineer biology and resulting in the emergence of new techniques to edit genes and reprogram cells.

There are four transformative biological areas that are experiencing innovation growth, according to a report by McKinsey Global Institute:

• Biomolecules: the mapping, measuring, and engineering of molecules to create processes, devices and products that benefit people
• Biosystems: the engineering of cells, tissues and organs
• Bio-machines: the interface between biology and machines
• Bio-computing: the use of cells, or molecules such as DNA, for computation

All these bio-innovative areas show various rates of progress, from their initial innovation triggers to that of developing proofs of concepts and, ultimately, deploying them in their relevant fields. Such new biological capabilities have the potential to bring about sweeping change to our societies, particularly in areas such as oncology, where the time between diagnosis and treatment is often key to a patient’s outcome.

Of course, such biological advancement would not have been possible without cost-effective computational processing power and the power of AI’s predictive capability. If there is one activity at which ML really excels at, it is identifying patterns and extracting insights about complex systems when given lots of data. Healthcare, therefore, represents an ideal candidate for the AI challenge and opportunity.

The progress in the disciplines of AI and data science, together with a perfect combination of increased computer processing power and speed, having access to larger data collections and data libraries and an expanding pool of AI talent, have all enabled more AI experimentation within life sciences and healthcare.

Noticeably, the advancement in deep learning has had an impact on the way we look at AI tools today. This is the reason for much of the recent excitement surrounding AI applications. Deep learning allows for finding correlations that were too complex to render using previous ML algorithms. Largely, this is due to the expansive developments in artificial neural networks.

For example, deep learning capabilities offered by companies like IBM Watson and Google’s Deep Mind are currently being used for many of the healthcare-related applications. IBM Watson is being used to investigate diabetes management, advanced cancer care modelling and drug discovery, but has yet to show clinical value to the patients. Deep Mind is also being considered for applications such as mobile medical assistant, diagnostics based on medical imaging and prediction of patient condition.

Apple’s well-known partnership with Stanford Medicine could lead to a new paradigm shift from a ‘provider-centric’ to a ‘patient-centric’ healthcare operating model. Meanwhile, the latest application from GE Research demonstrated the use of a sensor which is smaller than a fingertip that could find viruses and pathogens including Covid-19.

AI is proving to be a key enabler for accelerating healthcare innovations both in clinical and non-clinical domains.

Now, AI applications are being used in such areas as cognitive reasoning technologies, which include computer vision, natural language processing and speech recognition.

Additionally, in medical imaging technology, a number of firms are now working toward making the role of radiologists more effective by using AI-based computer vision algorithms to identify areas of mammograms that are consistent with breast cancer. The system automatically analyses mammogram images and outlines suspicious areas that indicate potential abnormalities.

Google’s recent launch of ‘Derm Assist’, an AI-powered platform that enables users to self-diagnose hundreds of skin conditions, is another demonstration of a true inflection point in this field. Other players such as Apple, Amazon and Microsoft are also pushing into this potentially lucrative space, offering healthcare solutions for patients, physicians and pharmaceutical and other healthcare related businesses.

Over the next two decades, we will see an expansion of direct brain-to-device communication for paralysed patients who are unable to communicate. In a recent lab demonstration, a paralysed man was able to translate his thoughts of writing into text at a rate of 90 characters per second.

By connecting these implants into his premotor cortex, he was able to envision that he was writing by hand, and the corresponding text was generated. Beyond 2040, we are likely to see a direct link between a brain and computational chip – a true neuromorphic computing capability.

Healthcare, life and biological sciences are entering an exhilarating new phase of growth as a result of digital advancements, particularly in AI, forging a transition from therapeutic to therapeutech. The lab of the future will enable scientific research to become more collaborative and predictive, joined by shared knowledge and digitalised processes, able to discover and respond to challenges, more rapidly.

Dynamic learning and compliance can be achieved to enable faster commercialisation from lab to market. And, of course, the use of AI, AR/VR, IoT-analytics and other forms of emerging digital automation technologies will generate a new patient journey ‘beyond the pill’ and an experience that is more dedicated to a patient’s specific needs, netting better results for all.

Parsing existing traditional business and operating models used by healthcare systems, providers and stakeholders to become more digitally activated and offer channel-agnostic patient/user experiences through integrating and using data from multiple systems and channels is now a necessity, and no longer an option.

Businesses organising for speed in this new era, to claim a first-mover digital advantage within the health landscape, are stimulating a race for new innovative value propositions.

But beyond exploring and reimagining health-related value propositions and creating optimised solutions, a number of challenges remain that an organisation could face and should be mindful of. These include, amongst others:

Data source of trust and truth: Sources of structured and unstructured data including those from healthcare networks (hospitals, clinics and laboratories), technology networks (sensors, monitoring and IoT devices), and social networks will need to be enabled and conditioned for use by AI models and big data analytics applications.

Deep learning algorithms tend to be data-greedy. Big data in biotech is not always well-prepared for modelling, or it might be inaccessible due to privacy reasons. Off-the-shelf pretrained data models with large parameters are now accelerating the uptake of AI and its diffusion.

Google’s Bidirectional Encoder Representations from Transformers has 340 million parameters. Microsoft launched its Turing-NLG model in February 2020 with 17 billion parameters. OpenAI launched GPT-3 in June 20 with 175 billion parameters.

Culture of a digitally enabled organisation: A change in the culture of how digital health is viewed across all involved, including the public, policymakers, providers, and those from the health and care professions, is needed to elevate the value of data accuracy, consistency and currency. Communicating the benefits of shared data and data governance is crucial.

Decentralisation of healthcare models: With remote patient monitoring on the increase (for example, analysis of a patient’s health metrics including vital signs, heart rate, blood glucose, temperature, medication adherence and physical movement), unnecessary visits to healthcare providers will be reduced, negating the need for centralised healthcare models. Such decentralisation must be public health-driven to enable better orchestration of services.

The demand for AI talent: Around 28% of AI/ML initiatives have failed, according to a survey by IDC in June 2020. Lack of staff with necessary expertise, lack of production-ready data and lack of integrated development environments are reported as primary reasons for failure. A growing wave of specialised university courses, geared toward data science and AI applications, is projected to address this issue in the coming years.

Ethical considerations: Building trust, transparency and value to the public as well as profits for commercial organisations is the starting point. In AI-related applications, the notion of AI explainability, where results of the solution can be understood or replicated by humans, thus removing the concept of the black box, will need to be resolved to enable trust to be manifested.

Technology is becoming a regulated industry. The EC will propose a horizontal regulatory guidance in 2021 to safeguard fundamental EU values, rights and user safety by obliging high-risk AI systems to meet mandatory requirements.

Earlier this year, the UK published its AI Council Roadmap to guide self-regulations. And, of course, the element of power consumption related to the processing of AI’s large neural networks will need to be responsibly innovated to mitigate any unintended harm to the environment and society.

National AI strategies continue to grow with real money attached to them. Examples include investment of $47.5bn in the US, $7.2bn plus regional investments in China, $2.6bn in the UK and €1.5bn in France.

In medicine and healthcare, AI algorithms and related digital technologies are already reshaping the bio/life sciences and healthcare roadmaps. Customised offerings that match an individual’s genes, their environment and lifestyle needs, and that are founded on defined clinical baselines and determined efficacy, will become the new normal in the not-so-distant future.

A completely undiscovered world of new possibilities awaits us, where health capabilities powered by a fusion of digital technologies seamlessly interlacing with bio and life sciences will help us create the embodiments of our imaginations.