Silver BlogHow AI will transform healthcare (and can it fix the US healthcare system?)

This thorough review focuses on the impact of AI, 5G, and edge computing on the healthcare sector in the 2020s as well as a look at quantum computing's potential impact on AI, healthcare, and financial services.

Image source above: MarketsandMarkets AI in the healthcare market

It was noted that a study by Accenture estimated that the insurance sector could save up to $7Bn across an eighteen month period through application of AI based technology.

Image Source Markets & Markets, AI in Healthcare by region

Furthermore, Krithika Srivats noted in the Becker Hospital Review that AI has the potential to enable significant cost reductions by reducing the number of improper payments claims that are reported to cost the healthcare sector hundreds of billions of dollars on an annual basis.

Enabling better efficiencies in terms of payments will enhance the cost efficiencies. More fundamentally the insurance sector for healthcare plays a key role all over the world. We are likely to witness innovations across the healthcare system with participants from within the insurance sector keen to take advantage of the innovations that 5G will bring and future generations of wearable technology to help avoid an insurance claim or reduce the claim. The personal data would need to be protected from snooping and misuse.

Tractica Global Forecast for AI in Healthcare

Ultimately the objective of preventative medicine should be to result in quicker and more accurate diagnosis, along with reduced side effects (fatalities and serious adverse reactions) from drugs or medical procedures which in turn should reduce insurance claims and improve health standards for all of us. The potential for AI in personalised medicine, more efficient drug development, and real-time analytics with wearables, smart sensors and medical imaging should have a beneficial impact for all of us during the course of the next decade and result in a more efficient healthcare and insurance sector.

Drug Discovery

AI has huge potential to transform and disrupt the drug discovery process resulting in a dramatic acceleration in the time taken to develop new drugs and enormous cost reductions for the development of new drugs. It is estimated that the pre-clinical stage accounts for 33% of the cost of developing a new drug.

The estimated cost for developing a new drug is US$2.6 billion. Nic Fleming authored an article published in Nature entitled "How Artificial Intelligence is changing drug discovery" that observed "Few people in the field doubt the need to do things differently."

Drug Discovery entails a big data challenge with a vast search space. Olğaç et al. authored research entitled "Cloud-Based High Throughput Virtual Screening in Novel Drug Discovery" and noted that "In 1996, Regine Bohacek, generated predictions about possible chemical compound types that might be chemically accessed. Her estimation was pointing to 10^60 chemical compounds making up chemical space that was virtually identified by using carbon, oxygen or nitrogen atoms, and by considering linear molecules with up to 30 atoms."

"In later studies, the limits of the chemical space was drawn to be between 10^18 to 10^200, according to the results of the analyses by different methods and descriptors... But it is expected that this number will continuously increase by the discovery of new chemical skeletons. Additionally, the number of organic compounds accessed experimentally is 10^8, according to CAS and Beilstein databases which contain records obtained from the scientific papers, those have been published by the scientific community, since 1771."

Polishchuk et al. Estimation of the size of drug-like chemical space based on GDB-17 data estimated the diversity of synthetically feasible chemicals that can be considered as potential drug-like molecules as between 10^30 and 10^60.

Image Source Above The Biopharmaceutical Research and Development Process

An article in Bloomberg entitled "AI Drug Hunters Could Give Big Pharma a Run for Its Money" further emphasised the points made above noting "And science moves slowly: In the nearly 20 years since the human genome was sequenced, researchers have found treatments for a tiny fraction of the approximately 7,000 known rare diseases."

Further, there are approximately 20,000 genes that can malfunction in at least 100,000 ways, and millions of possible interactions between the resultant proteins. It’s impossible for drug hunters to probe all of those combinations by hand.

Artificial Intelligence could be used to scan millions of high-resolution cellular images—more than humans could ever process on their own—to identify therapies that could make diseased cells healthier in unexpected ways.

Recursion, a startup applying Machine Learning techniques to scan images and search compounds that may disrupt disease without harming healthy cells raised $121 million in its latest financing round at a valuation of $646 million, according to PitchBook.

At DLS, we have also worked with medical imaging with Deep Learning techniques applied for cell imaging and believe that computer vision in healthcare is set for continued growth.

Ingrid Torjsen "Drug development: the journey of a medicine from lab to shelf" summarised a part of the drug discovery process and noted that upon identification of a potential target, researchers embark upon a search for a compound or molecule that will act upon the given target. In the past researchers search for candidate drugs focussed on natural compounds for example extracted from plants or fungi, however, the recent trend has been for researchers to apply knowledge obtained from the studying genetics as well as proteins to develop new molecules via computers. The process results in up to 10,000 compounds being considered and reduced to a mere 10 to 20 that have a theoretical ability to intervene in the disease process.

Chen et al. authored a paper entitled "The rise of Deep Learning in drug discovery" where it was noted that the large amounts of data made drug discovery an ideal area in which to apply Deep Learning. Chen et al. also noted the potential for De novo drug design through Deep Learning whereby new chemical structures are generated by Neural Networks.

An example is given by DLS whereby our Deep Learning model not only generates novel compounds but also can test those compounds for toxic characteristics hastening its clinical translation.

Deep Learning techniques used in drug discovery include:

The global AI for drug discovery market is estimated to witness high growth during the forecast period. CBINSIGHTS note in "AI Trends To Watch In 2019" that "...traditional pharma companies are looking to AI startups to reduce the long drug discovery cycle. Although many of these startups are still in the early stages of funding...One of the top AI trends of 2019 will be increased investment in the space by leading pharma incumbents."

Image Source Above: CBINSIGHTS AI Trends To Watch In 2019

The amount of equity invested into drug discovery remains relatively small in relation to the amount being spent by the Pharma sector on drug discovery with the ABPI stating "The global pharmaceutical industry invested over $1.36 trillion in R&D in the decade from 2007 to 2016 and forecasts predict an annual investment of $181 billion by 2022."

Hence even if it is just 1 or 2 AI drug discovery companies that succeed in becoming a Google, AirBnB or Amazon of the Pharma sector and in the process bring the change in the sector needs the amount invested will have been worth it given the scale of change and legacy in terms of changing the lives of people for the better at a time when the healthcare system is broken.

Drug Discovery and Personalised Medicine

In addition applications of AI in drug discovery may also be targeted towards the development of personalised medicine whereby the next generation of drugs may emphasis the intended benefits of the drug and reduce the harmful side effects.

Naveen Joshi in an article entitled "Big data in the pharmaceutical industry" notes that there is scope to "develop personalised medicines that are suitable for an individual patient’s genes and current lifestyle. Also, precision medicine can predict susceptibility to certain disorders and enhance disorder detection. With this approach, precision medicine has a higher probability of providing successful treatment compared to conventional medicines. Precision medicine may also save costs that occur in conventional medicines."

Nic Fleming in an article entitled " How Artificial Intelligence is changing drug discovery" observed that "Some think the potential of AI to pinpoint previously unknown causes of disease will accelerate the trend towards treatments designed for patients with specific biological profiles. "

“Personalised medicine has been talked about for a long time,” says Hunter (CEO of BenevolentBio). “AI is going to enable it.”

The end goal of AI in drug discovery should be to accelerate the development of the next generation of drugs to patients at lower cost with reduced side effects. Furthermore, the application of AI in drug discovery may enable greater focus on areas such as rare diseases due to the reduced cost potential of drug discovery relative to the conventional drug discovery approaches used today.

Medical Imaging: Improving the Speed and Accuracy of Diagnosis

The chart above from Signify Research shows the rapid growth rate in AI technologies for image analysis to hit revenues of $2Bn by 2023 with Deep Learning set to play the major role. The advantage of deploying such technology is that it can alleviate the skills gap as it takes years and significant investment to train medical staff.

A study by NVIDIA reported that Deep Learning Drops Error Rate for Breast Cancer Diagnoses by 85%. The study observed that "human analysis combined with Deep Learning results achieved a 99.5 percent success rate. This shows pathologist performance can be improved when paired with AI systems, signalling an important advance in identifying and treating cancer."

An article entitled Using AI to predict breast cancer and personalize care set out recent research from MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), and Massachusetts General Hospital (MGH), entailed developing a Deep Learning model to predict the likelihood of a patient developing breast cancer from a mammogram. The model was trained on outcomes that were known and mammograms from in excess of 60,000 patients who had been treated at MGH with the model learning the particular patterns in the breast tissue that re precursors to malignancy. The paper published 7th May 2019 is entitled A Deep Learning Mammography-based Model for Improved Breast Cancer Risk Prediction.

Research from Stanford University published in Nature in 2017 by Esteva et al. showed that a "Convolutional Neural Network (CNN) achieves performance on par with all tested experts...demonstrating an artificial intelligence capable of classifying skin cancer with a level of competence comparable to dermatologists."

In addition to cell imaging, Deep Learn Strategies (DLS) team have worked on advanced application of such technology for a data set relating to Skin Lesions in order to demonstrate the potential of this technology to help solve the problems that the healthcare system faces.

The aim is to augment the medical staff rather than replace them.

Furthermore, our team have demonstrated the concept of explainable AI with the CNN using visualisations of the decision. This is important in the healthcare sector where the medical staff need to understand and trust the decisions of the machine with confidence.

We took a large dataset labelled with 7 different types of skin lesions one of which is melanoma cancer. We trained a state-of-the-art CNN to process this dataset and learn the patterns in the images that enabled the network to classify an unlabelled image to one of the seven different classes. The end to end deep learning architecture consists of data pre- processing and the Deep Neural Network for training. This is followed by a visualisation step that shows which areas of the image the network is focussing its attention on to take the appropriate decision.

The CNN network was trained with NVIDIA GPUs and we used TensorFlow as a framework for the Deep Learning. We used 2 different architectures namely Inception V3 and DenseNet as a backbone architecture for the CNN. We found DenseNet to be performing better than InceptionV3. This is as a result of dense connections between the convolutional layers of the CNN. The data set is challenging because it consisted of multiple images that are difficult to classify even for the human eye. Deep Learning makes the process relatively easy here as a result of seeing thousands of images at once.

We used a dataset of approximately 10,000 images with 90% used for training and 10% used for testing. There is published research that the human accuracy of the dermatologist is around 86.6% in being able to detect and identify cancer. Our CNN was able to achieve an accuracy of 90% across 7 classes. We believe our accuracy would have been even higher but for the class imbalance within the dataset and in particular if we had access to a larger dataset.

It is noted that recently that a team from Google have hit 99% accuracy rates on a different and larger dataset relating to lung cancer (we worked on skin lesions).  The results from Google demonstrate the benefit of hospitals and regulators enabling data to be shared with anonymisation available to protect patient identity so as to enable the startup community to access larger datasets and in the process demonstrate how this technology can help save lives.

Our CNN took 3 days to train on a NVIDIA Titan X GPU. We believe the visualisations of the output of our CNN can help the dermatologist focus their attention on the relevant parts of the image that the CNN identifies as having indications of cancer.

Some of the visualisations are shown below for a type of skin lesion:

Image A is the original image that is usually captured with a camera and image B is a heat map showing the visualisation of the network with colours indicating areas of attention at different levels with red meaning the highest attention.

  1. Original Image  - clear for human eye to detect
  2. Heatmap - what the Deep Learning algorithm focuses on
  3. Original Image - harder for the human eye to detect
  4. Heatmap - Deep Learning algorithm detects it

Another example shown in C, is benign keratosis – like lesions, and image D is the heat map. Our network is able to process such images that are difficult even for the human eye to analyse.

In BorntoEngineer "Diagnostic Computers Outperform Human Doctors" noted that "In one example, the computational-imaging system predicted with a 97% accuracy which patients were showing evidence of heart failure. While, two human doctors presented with the same information were only able to predict 74% and 73% accuracy....Does This Put Doctors Out Of Work? The short version is no. "

"I always use the example of Botswana, where they have a population of 2 million people—and only one pathologist that we aware of,” he said. “From that one example alone, you can see that this technology can help that one pathologist be more efficient and help many more people." Anant Madabhushi Professor of biomedical engineering at the Case School of Engineering.

"The massive study examined the electronic health records (EHR) from nearly 600,000 patients over an 18-month period at the Guangzhou Women and Children’s Medical Center and then compared AI-generated diagnoses against new assessments from physicians with a range of experience."

"The verdict? On average, the AI was noticeably more accurate than junior physicians and nearly as reliable as the more senior ones. These results are the latest demonstration that Artificial Intelligence is on the cusp of becoming a healthcare staple on a global scale."

A new study from China has found that an AI system can best some doctors when it comes to diagnosing common childhood diseases. The study, published in Nature Medicine yesterday (Feb. 11), was co-authored by a long list of Chinese and American researchers, and is just one of the latest from China to look at the use of AI in medical diagnostics.

A study from China relating to AI within medical diagnosis was published in Nature Medicine on the 11th Feb 2019 relating to a Deep Learning system trained on 101 million data points generated from the EHRs of 1.3 million patient visits to a medical center in Guangzhou.

Adam Rasmi stated in "China has produced another study showing the potential of AI in medical diagnosis" that "Researchers found that the AI system was able to meet or outperform two groups of junior physicians in accurately diagnosing a range of ailments, from asthma and pneumonia, to sinusitis and mouth-related diseases. The AI was also able to meet or exceed diagnostic performance with some groups of senior physicians, for instance, in the category of upper respiratory issues."

"In some cases, the system was able to diagnose conditions with 90 to 95% accuracy. In no category did the AI model dip below a diagnostic accuracy rate of about 79%, higher than one group of junior physicians, but lower than the other. Senior doctors overall did better than the AI system."

"One limitation to the study, however, is that the researchers pulled all their data from visits to a single medical facility in China, which raises questions about its applicability outside the country and even the Guangzhou facility itself...Still, the study adds to the growing body of research demonstrating AI’s benefits when it comes to diagnostics."

The role of AI in healthcare will be to augment doctors and make their work more efficient.

Aaron Saenz commenting on the same study in an article entitled "The Pediatric AI That Outperformed Junior Doctors" noted that "New research out of Guangzhou, China, published February 11th in Nature Medicine Letters, has demonstrated a natural-language processing AI that is capable of out-performing rookie paediatricians in diagnosing common childhood ailments...The near future of Artificial Intelligence in medicine will see these Machine Learning programs augment, not replace, human physicians. The authors of the study specifically call out augmentation as the key short-term application of their work."

Clinical Trials and remote monitoring

Whilst the pre-clinical stage of drug discovery is estimated to account for around 33% of the cost of developing a new drug the clinical trial phase is estimated to cost between 50% to 63% of the cost of developing a drug. The high cost includes a great deal of wastage due to cash being spent on the 90% of candidates that fail to proceed between Phase 1 trial and regulatory approval. I personally believe that for AI to demonstrate drug discovery breakthroughs will also require disruption of the clinical trial process.

The consequences of failure at the clinical trial stage can be very costly for example CBINSIGHTS article "The Future Of Clinical Trials: How AI & Big Tech Could Make Drug Development Cheaper, Faster, & More Effective" noted that "Switzerland-based Novartis, for instance, attributed a 15% drop in its Q1’17 net income to a failed Phase III drug intended to treat heart failure. In the US, two months after pharmaceutical company Tenax Therapeutics’ main drug failed in a Phase III trial, the CEO resigned, and the company was reportedly considering a merger or sale."

The advent of 5G along with the anticipated growth in edge computing with AI on devices all around us ranging from intelligent sensors at home to wearables will result in a transformation in the way that clinical trials are currently conducted. Medical staff will be able to monitor patients on trials in relation to key biology readings in near real-time and if anomalies are detected algorithms could trigger automated alerts to both the patient and the doctor.

Furthermore, AI could be used to match appropriate candidates for a trial with CBINSIGHTS noting that "Matching the right trial with the right patient is a time-consuming and challenging process for both the clinical study team and the patient. Roughly 80% of clinical trials fail to meet enrolment timelines, and approximately one-third of Phase III clinical study terminations are due to enrolment difficulties, according to a Cognizant report on recruitment forecasts."

"Patients may occasionally get trial recommendations from their doctors, if a physician is aware of an ongoing trial. Otherwise, the onus of scouring through — a comprehensive federal database of past and ongoing clinical trials — falls on the patient."

"An ideal AI solution would be Artificial Intelligence software that extracts relevant information from a patient’s medical records, compares it with ongoing trials, and suggests matching studies."

Furthermore Jane Reed in "Can NLP Text-Mining Transform Drug Discovery & Development for Biopharma?" notes that there is scope to leverage Natural Language Processing (NLP) for text-mining to improve patient health. "The explosion of clinical health data has created a trove of information that can be leveraged to accelerate the discovery and delivery of new drugs and therapy that ultimately improve the health of patients." This includes using data from clinical trails as well as data from real world evidence that may be collected from outside clinical trials in order to "to advance the innovation and delivery of their products."

The role of 5G: Hospitals turn into data centres and remote monitoring

Telecoms company Ericsson have made the point that hospitals will turn into data centres.

"In order for the transformation of patient applications to happen, patient data will need to be stored centrally, effectively turning hospitals into data centers and doctors into data scientists. Patients will get online access to a central repository of medical records to help them easily manage the quality and efficiency of their care. "

"5G will certainly bring the ability to stay in touch with the patient through these monitoring technologies, whether it’s out in the field with an ambulance or in the patient’s home." Hospital IT, US.

"When it comes to battery life, 42 percent of cross-industry decision makers expect 5G to enable devices to consume less power which is key in remote monitoring situations."

"Current consumer-grade wearables are widely used for preventative measures but are not considered sufficiently accurate or reliable for diagnosis. In addition, for liability reasons, patients’ smartphones cannot be relied upon for connectivity. For their part, wearables require high-frequency updates from a central repository but at low-data rates. 5G connectivity is not limited to wearables; it also enables patients to carry a medical-grade 5G router that then connects to various wearables using Bluetooth."

Steven Armstrong in an article entitled "How 5G can revolutionise healthcare" quoted Anne Sheehan, director at Vodafone Business “In a 5G-enabled ambulance, paramedics will be able to share critical patient data with A&E staff in real time while en route to the hospital, speeding up the triage process. And high-quality video streaming will enable A&E staff to begin diagnosis as early as possible.”

Brian Buntz in "AI, 5G and IoT in Health Care: Evolution or Revolution?" observed that telecoms firms could play an increased role in the healthcare sector stating that" 5G could boost the influence of telecom firms on the industry."

"The alliance of such firms could converge around Artificial Intelligence, which could serve as umbrella technology. “Physicians don’t have the time to sift through massive amounts of data,” McCann said. And AI can broadly help to address the current shortage of doctors and caregivers while helping streamline and inform medical decision-making, flagging anomalous health care, potentially fraudulent patient data or sift through large troves of research data or genomic information. But again, in order for AI to live up to its promise, the technologists developing it must work closely across the health care ecosystem."

It is my belief that the AI startups that will truly scale and transform healthcare will be those that leverage 5G, edge computing and AI in combination including in the drug discovery world so as to enable clinical trials to gather large amounts of near real-time data and real world evidence so as to enhance and accelerate the process.

I agree with the view that drug discovery cannot be viewed in isolation of working in the real world. Drug discovery startups will still have to prove to the likes of the FDA that their pipeline is effective and pass the stringent requirements of clinical trialing.

Artificial Intelligence and Electronic Health Records (EHRs)

The vast amount of data being generated means that it is impossible for every physician to read and indeed recall the entirety of all the data published.

The Word Economic Forum (WEF) published an article entitled "How to unleash the enormous power of global healthcare data" and observed the vast scale of healthcare data stating "Two thousand three hundred and fourteen exabytes."

"This is approximately how much space it would take to store the total volume of global healthcare data by 2020, according to a report by the EMC and IDC. To put this in context, it would take approximately five exabytes to store all the words ever spoken by humankind. If all the data stored in 2,314 exabytes were to be stored on tablet computers and stacked, it would reach 82,000 miles high or circle the earth 3.2 times and would equal all the written works of humankind, in every known language, 46,280 times over."

Davenport, Hongsermeier and Mc Cord authored an article entitled in the Harvard Business Review (HBR) entitled "Using AI to Improve Electronic Health Records" and noted that the following areas have potential for AI in relation to EHRs:

Data extraction from free text – the ability to extract structured data or review the notes from a provider with AI applied to recognise essential terms, discover insights and increase productivity

Diagnostic prediction models from big data to warn clinicians of high risk conditions such as sepsis and heart failure.

Clinical documentation and data entry -  the use of natural language processing (NLP)  to capture clinical notes  freeing up medical practitioners to focus on patients instead of typing notes into keyboards. The example of Nuance given that offers AI-supported tools that integrate with commercial EHRs to support data collection and clinical note composition.

Clinical decision support - using Machine Learning that learn as additional data is gathered to recommend treatment strategies that enhance personalised care.

Ayn de Jesus observed the potential for potential AI enhanced AR use cases for EHRs with Google Glass in "Augmented Reality in Healthcare – 2 Current Applications" whereby "Google Glass may have missed the mark with consumers, but the healthcare industry has expressed interest in making it work for physicians. Although it’s a nascent use-case, healthcare networks are looking to equip their doctors with Google Glass software that would allow them to see patient notes without taking their eyes off their patient."

"This could allow physicians to be more present with patients during visits. In the future, AI startups may develop natural language processing-based (NLP) software and/or facial recognition technology that automatically pulls up a patient’s record with a voice commandand transcribes notes about the visit into an electronic health record system."

Robotic Surgery

Roger Smith in "How Robots and AI are Creating the 21st-Century Surgeon" noted that "More than 5,000 surgical robots were used in more than 1 million procedures worldwide in the last year."

"AI delivered advice would be derived by Machine Learning algorithms from thousands of previous cases and stored in the cloud for access when needed. For example, a visual overlay inside the surgical space could indicate where critical blood vessels lie behind the current operating plane, with the AI suggesting that the surgeon steer clear of those areas. It could also show how thousands of previous successful surgeons traversed the anatomy, and where they took action. The robot would also be aware of the specific tools loaded into the robotic arms, and might suggest previously successful alternatives."

Nguyen et al. "Manipulating Soft Tissues by Deep Reinforcement Learning for Autonomous Robotic Surgery" introduced a multi-point approach using Deep Reinforcement Learning for surgical soft-tissue cutting task and benchmarked the accuracy of the multiple pinch points.

Source Image Above Nguyen et al. "Manipulating Soft Tissues by Deep Reinforcement Learning for Autonomous Robotic Surgery"

It is envisaged that Deep Reinforcement Learning will also assist in the training of surgeons in the future for example in relation to Minimally Invasive Surgery for laparoscopy and as Liu and Jiang explored in "Deep Reinforcement Learning for Surgical Gesture Segmentation and Classification" for objective surgical skill assessment and for improving efficiency and "quality of surgery training in relation to segmenting robotic kinematic data or video sequence and to classify segmented pieces into surgical gestures, such as reaching for the needle, orienting needle and pushing needle through the tissue, etc."

Robotics and AI to help take care of the ageing population

An article in Forbes by Shourjya Sanyal How Is AI Revolutionizing Elderly Care observed that economies including the US, Canada, Japan, China and across Europe face a growth in the percentage of the ageing population with the percentage of those over 60 to increase from 12% to 20% (60m to 2 billion) by 2050. "During the same period, the number of people aged 80 years and older will quadruple. In the USA, 14.5% of the population is 65 years or older, but by 2030 these number is anticipated to grow to 20%."

An article in Synced entitled "Grandma’s Robot: How AI Is Revolutionizing Elder Care" noted that "The AI in elder care market is expected to exceed US$5.5 billion by 2022, and will grow into one of AI’s most important support roles in societies of the future."

The article also gave the example of Intuition Robotics developing ElliQ, an AI powered robot that can hold conversations with patients and remind them when it is time to take their medication as well as gentle physical activity to remain healthy.

Synced also note the work of IBM’s elderly care solutions that comprise of movement sensors in corridors, flush-detectors in toilets and bed sensors for monitoring sleep, etc. If the sensors detect a significant deviation from the normal patterns of activity then it may result in an automated alarm to the nurses or doctors of the person.

Source Image Above: IBM research, Synced Grandma’s Robot: How AI Is Revolutionizing Elder Care

The video below from the BBC below gives an example of the use case of robots for assisting the elderly:

Melanie Walker authored "Healthcare in 2030: goodbye hospital, hello home-spital" with the vision that in the longer term future around the 2030s the hope is that the fourth industrial revolution will mean that disease itself will be disrupted by technology with the result that there are significantly less diseases to manage. Melanie Walker envisions a world where hospital wards will prioritise instant diagnosis with one device used for scanning metabolic, functional and structural aspects entailing physics of spectroscopy, magnetic resonance and radiation. This will mean you only need one scan, and no biopsy.

We can speculate whether in the longer-term research into Biotechnology, Stem Cells and AI will allow us to delay the ageing process itself. Research work is already underway today in this field. For example a paper by Zhavoronkov et al. "Artificial intelligence for aging and longevity research: Recent advances and perspectives", relates to the work of Insilico Medicine, Inc., a company engaged in ageing research, which designs and uses AI-based algorithms for de novo molecules generations and is also involved in biomarker development. The CEO of Insilico, Dr Alex Zhavoronkov observed in "F036 How is AI decoding aging?" that there are three areas of how AI can be applied to ageing:

  1. The construction of aging clocks - guessing age;
  2. Generation of novel chemistry — enables the design of interventions to test your hypothesis;
  3. New data generation - creating models in which moving one feature in time shows changes in other data types.

What is Quantum Computing?

I've been asked whether Quantum Computing will result in Artificial General Intelligence (AGI) or whether we can actually undertake De Novo drug design without Quantum Computing. Quantum Computing has potential to transform sectors such as capital markets and asset management, healthcare and cybersecurity. It may result in greater accuracy of prediction for Machine Learning models, however, it will not of itself result in AGI.

Quantum Computers can more efficiently process enormous and complicated datasets relative to classical computers. Quantum mechanics are applied for solving highly complicated computations much faster. Potential applications relate to healthcare with genomics and finance. Cybersecurity is another area that could be disrupted due to the ability of Quantum Computers to break encryptions. Currently tech majors including Google, Amazon, Intel, IBM and Alibaba are all conducting research on Quantum Computing and cybersecurity.

CBINSIGHTS "What Is Quantum Computing?" note that Quantum AI could enable Machine Learning that is faster than that of classical computers. "In the distant future, universal Quantum Computers could be used to accelerate the field of Artificial Intelligence."

"Quantum Machine Learning could create AI that more efficiently performs complex tasks in human-like ways. For example, enabling humanoid robots to make optimized decisions in real-time and under unpredictable circumstances."

"Training AI on Quantum Computers could advance computer vision, pattern recognition, voice recognition, machine translation, and more."

However CBINSIGHTS also observed that "Recent work has produced algorithms that could act as the building blocks of quantum Machine Learning, but the hardware and software to fully realize quantum Artificial Intelligence are still as elusive to us as a general Quantum Computer itself."

Financial Services

Quantum Computing has the potential to reduce the chances of the adoption of flawed financial assumptions that cause financial losses due to blind sports in the data being missed in the probabilistic algorithms that financial analysts currently use.

In particular specific areas of finance that maybe affected by Quantum Computing related to portfolio risk optimisation whereby the construction of optimal portfolios that comprised many assets with decencies that interconnect could be further enhanced and the ability to detect patterns of fraud more efficiently.

Quantaneo quote the Qubit report "Many major banks, including Barclays and JP Morgan Chase, are looking to quantum computing to secure their future. Both of these banks have begun to experiment with prototype quantum computer technology via the cloud and also to conduct important research as to how this cutting edge technology can be utilized successfully in the banking world."

"Currently, this new technology is still in its infancy and quantum computer systems are still not at the level of effectiveness as traditional computers. "

"This important development could totally replace other forms of secure data protection, such as blockchain within a period of ten years using quantum encryption."

"Quantum encryption will enable banks to send data which is almost unhackable over a quantum network. Quantum cryptography uses a system called quantum key distribution also known as QKD which ensures encrypted messages and its keys are sent separately. If these messages and keys are tampered, or edited in any way, they are automatically destroyed. At this point, both the sender and the receiver are sent notifications."

A further area where Quantum Computing may impact the finance sector relates to Monte Carlo simulations. Monte Carlo simulations are often applied in financial models to enable analysts to consider the potential outcomes of a particular investment or event and to generate a probability distribution or risk assessment in relation to the particular investment. Quantum Computers would be able to run such processes in parallel relative to classical computing that can only run a single Monte Carlo simulation of a portfolio at a time.

Quantum Computing and Healthcare

Furthermore, Quantum Computing may be able to combined and search through all the possible variations of genetics.

"Rapid quantum genome sequencing could allow us to pool the world’s DNA into a broad population health database. Using quantum computers, we would also be able to synthesize patterns in the world’s DNA data for understanding our genetic makeup at a deeper level, and also potentially uncover previously unknown patterns of disease."

A major challenge for the era of powerful Quantum Computing relates to universal Quantum computers. Universal quantum computers are the most powerful and most generally applicable, but also the hardest to build. A truly universal quantum computer would likely make use of over 100,000 qubits — some estimates put it at 1M qubits. Remember that today, the most qubits we can access is not even 128.

Image Source CBINSIGHTS What Is Quantum Computing?

CBINSIGHTS notes that "Experts agree that by 2030, we could see quantum computers outpace classical counterparts."

"Significant technical barriers must be surmounted before Quantum Computing achieves its potential. This will require the development of more stable hardware, commercial platforms for software development, and the development of cloud computing capabilities for the distribution and access of quantum computing resources."

Deloitte "Quantum computing: The next supercomputers, but not the next laptops" believe that "The first commercial general-purpose quantum computers are likely to appear in over a decade’s time - in the 2030s. The 2020s will likely be a time of progress in quantum computing, but it is not likely to be until the 2030s that the larger market is able to develop."

Deloitte also note that "In the 2020s quantum computing will generate revenue, but it will be on a lower scale...Few CIOs are likely to be submitting budgets for quantum computing in the next two years. But that does not mean that leaders should ignore this field. Quantum computing is advancing rapidly, and its impact is likely to be large. So business and technology strategists should monitor progress on the evolution and potential implications of quantum computing starting now."

It has been widely reported that there is a race for Quantum Supremacy between the US and China with the Chinese government specifying billions to play catch up with the US in relation to Quantum Computing technology. It maybe that during the next decade the efforts around Quantum Computing are more driven by a race around National Security issues and the need to be a the leader in the field.

It was reported that Quantum computer ‘backbone’ developed in China but Guo Guoping professor at the Hefei-based University of Science and Technology of China stated "...the gap between China and the world will not be narrowed overnight...China launched its research on quantum computing some 10 years ago while world leading technology companies such as Google, IBM and Canada-based D-Wave have been working in this field for some 30 years. They produced the best quantum chips."

The 2030s appears to be a more realistic view for the potential arrival of truly powerful Quantum Computing. For example in an article by David Manners entitled "Quantum computing ten years away – again" quotes the US National Academies of Sciences, Engineering, and Medicine "Given the current state of quantum computing and recent rates of progress, it is highly unexpected that a quantum computer that can compromise RSA 2048 or comparable discrete logarithm-based public key cryptosystems will be built within the next decade,” says the report."

"Cracking encryption is seen as one of the key applications, and dangers, of quantum computing."

“The average error rate of qubits in today’s larger devices would need to be reduced by a factor of 10 to 100 before a computation could be robust enough to support error correction at scale, and at this error rate, the number of physical qubits that these devices hold would need to increase at least by a factor of 10^5 in order to create a useful number of effective logical qubits,” says the report."

"It goes on to say that it is unable to predict when that capability might be achieved."

As a technologist and futurist, I remain optimistic about the potential for Quantum Computing supremacy in the 2030s and the exciting changes that the technology will bring for Financial Services and areas of healthcare.

Whilst Quantum Computing Supremacy may or may not arrive by the end of the next decade, AI will have made a huge impact upon our lives by the time we celebrate New Year for 2030.

Some articles have proposed on benefit of Quantum Computing will relate to the ability to model protein folding. However, Google DeepMind have already demonstrated important progress as noted in the article "AlphaFold: Using AI for scientific discovery"

Why is protein folding important?

The blog by DeepMind explains that "The ability to predict a protein’s shape is useful to scientists because it is fundamental to understanding its role within the body, as well as diagnosing and treating diseases believed to be caused by misfolded proteins, such as Alzheimer’sParkinson’sHuntington’s and cystic fibrosis."

Source for gif above: DeepMind AlphaFold: Using AI for scientific discovery

"It’s exciting to see these early signs of progress in protein folding, demonstrating the utility of AI for scientific discovery. Even though there’s a lot more work to do before we’re able to have a quantifiable impact on treating diseases..."

The work by Google DeepMind on protein folding is an ongoing work in progress but it has demonstrated the exciting potential of AI to make a real difference to our lives in a positive manner.

AI and ethics

Machine Learning requires data to work effectively and the healthcare sector currently has silos of datas scattered around. As efforts are underway around the world to increasingly digitalise data and to broaden the application of AI to the healthcare sector there are also calls about AI and ethics in relation to privacy. There may also be a temptation for some in the healthcare system to hold the data back in the hope that they can extract a high price for the said data. This also then leads to the issue of asking the question about whom does the data belong to? Does the data belong to the individual patient or the hospital (or clinic) or both?

There are those who would argue that the greater the amount of data that was made available then the greater the likelihood that we will manage to find solutions to various diseases and result in improved patient outcomes. At the same time the individual patient maybe concerned about their data being shared with insurance providers and others without their consent.

Dr Alex Zhavoronkov stated that “It should be a fundamental law, for all the medical data to be donated for medical purposes until we aren’t capable of curing diseases that kill people. People think about data privacy but they fail to remember the pain and suffering caused by diseases. We need data to find cures.”

It maybe that the solution would be to enable anonymization where applicable and ask the patients to consent to grant access upon this basis.

The ethical issues continue and there is a need for us to ensure that we understand how algorithms are applied for example in social media so that they are fair and transparent.

Seth Redmore in "AI in Healthcare: Data Privacy and Ethics Concerns" explains the issues that arise in relation to Facebook and the fact that its activities are outside the jurisdiction of the HIPAA.

Healthcare AI startups will fail. Who will succeed?

Many startups fail. There are those who debate what the actual percentage is. The 90% failure rate has long been cited. For example Neil Patel referenced the 9 out of 10 startups failing before the current trend of AI startups in healthcare, and Bram Kroemenhoek in "Why 90% of Startups Fail, and What to Do About It" observed that "Most entrepreneurs think they’re building the next big thing. In reality, over 90% of them fail." The source for the figure was attributed to the Startup Genome Report Extra on Premature Scaling which noted that "More than 90% of startups fail, due primarily to self-destruction rather than competition. For the less than 10% of startups that do succeed, most encounter several near death experiences along the way."

Erin Giffen "Conventional Wisdom Says 90% of Startups Fail. Data Says Otherwise" argues that the real failure rate is lower stating "...a global investment firm based in Boston, tracked the performance of venture investments in 27,259 startups between 1990 and 2010. Its research reveals that the real percentage of venture-backed startups that fail—as defined by companies that provide a 1X return or less to investors—has not risen above 60% since 2001."

Whether one opts for the 60% or 90% figure, one may argue that many AI healthcare startups will fail for one reason or another. It is also worth noting that there has been an unfortunate element of hype made by some of the startups in the AI and healthcare space and in the AI startup space in general. For example James Vincent authored an article entitled "Forty percent of ‘AI startups’ in Europe don’t actually use AI, claims report" noting that "According to the survey from London venture capital firm MMC, 40 percent of European startups that are classified as AI companies don’t actually use artificial intelligence in a way that is material to their businesses. MMC studied some 2,830 AI startups in 13 EU countries to come to its conclusion, reviewing the “activities, focus, and funding” of each firm."

Furthermore, given the state of trade war between the US and China, and potential economic ramifications of a hard Brexit could result in a challenging environment for startups in 2020. I recall JP Nichols (@JPNicols) speaking during an event hosted by @FintechPower50 on 14th February 2019 that startups may face their first real challenge in a decade during 2020 if the economy contracts in the US and elsewhere and that those who succeed will be those who have a truly viable business model. The same parallel can be drawn for those in the healthcare space.

As noted the ABPI forecasts an annual investment of $181 billion by 2022 by the Pharma sector in R&D and I believe that irrespective of what happens to the economy in 2020, drug discovery will remain an area of substantial investment whilst other areas may experience some degree of market turbulence.

Furthermore it is submitted that the startups that succeed will be those that possess a blend of skills that span genuine AI and technical capabilities (Machine Learning, data engineering), domain knowledge of medicine and biology and are genuinely adding value by solving a problem that can scale. It is a common view in the VC community that AI drug discovery startups should focus purely on the pre clinical stage. There is a logic that startups that succeed are those that focus on solving a problem. However, I agree with the critics who argue that drug discovery needs solutions that are demonstrated to work in the real world. As noted above 2/3 of the problem is at the clinical trial stage and cashflow will be generated once the new drug is demonstrated to work and approved by the regulator.

CBINISIGHTS in "The Future Of Clinical Trials: How AI & Big Tech Could Make Drug Development Cheaper, Faster, & More Effective" observed that "Testing new drugs is a slow, expensive, and manual process. Artificial intelligence has the potential to disrupt every stage of the clinical trials process — from matching eligible patients to studies to monitoring adherence and data collection."

In order to solve the problem of delivering next generation drugs cost effectively and more quickly to the patient, drug discovery startups working with Deep Learning will need to work with the clinical trial stage. In the near future as 5G and edge computing scales out this will include working with IoT sensors and wearable devices to disrupt the entire Pharma vertical. The arrival of 5G and scaling of Edge Computing will enable and remote medicine to grow and also allow for continuous monitoring of the patient even whilst they are out of sight of the clinician.

Jessica Davies authored an article entitled "5G is coming in 2019: Here's what hospitals should know about it" noted that "What’s interesting is 5G’s use of a computing model that pulls insights from data with billions of devices."

"By 2020, the 5G network will support more than 20 billion connected devices, 212 billion connected sensors and enable access to 44 zettabytes of data gathered from a wide range of devices from smartphones to remote monitoring devices... researchers estimate that this connected ecosystem will make it possible to utilise a much larger percentage of digital data (35 percent) than before (5 percent)."

"The much greater ability to continuously gather patient-specific data and the ability to process, analyze and quickly return processed information and recommended courses of action to the patient will give patients greater ability to manage conditions on their own,” the authors continued."

"For example, the report found that 5G will better support continuous monitoring and processing sensory devices, which will support the continuous monitoring of patients. The tech will “substantially increase the effectiveness of preventative care."

"Predictive analytics will also improve under 5G through the growth of continuous monitoring. While 5G’s ability to continuously monitor data will develop new data streams, it will also use distributed computing to power analytics and intelligent care."

An excellent summary of where we are today on the journey to transforming healthcare is provided by John Nosta who commented in relation to the adoption of AI in healthcare that "The assimilation of today's data stream (or should I say fire hose) is simply greater than human capabilities. From clinical trials to complex genomics, the human brain can just go so far and so fast to place this life-saving information in a clinical context. "

"Complexity will not go away. Data will continue its assault on our cognitive skills."

"A solution is to recognize that the role of the clinician is shifting to something that might appeal to Hippocrates and patients. It's the role of the compassionate clinician—empowered by technology's cognitive prowess—to provide the basis for a new medical paradigm."

"Technology, or at least technology in its future iteration, isn't the basis for today's physician dissatisfaction. It's the basis for smarter and richer clinical engagement. And the time has come for the clinical community to recognize that it's not "business as usual" but a fundamental shift in roles."

"Yet it still seems that many physicians have "dug in" and wish to own both the cognitive and the compassionate domains. Good luck drinking out of that fire hose."

"The day will come when patients ask, "What did the computer SAY?" And that's not far off. The next comment out of the patient's mouth is equally revealing: "What did the doctor DO?" It's this partnership (dare I say, of equals) that give us a sense of tomorrow's brilliant and compassionate physician."

James Gallagher reported in an article published in the BBC entitled "NHS to set up national Artificial Intelligence lab" that £250m will be spent on boosting the role of AI within the UK health service.

The article noted that AI had the potential to revolutionise medicine in relation to diagnosis, insights into diseases and the manner in which hospitals are run.

James Gallagher noted that medical imaging is leading the charge with clinical trials proving that AI is as capable; as leading doctors at detecting lung cancerskin cancer, and more than 50 eye conditions from scans.

The resulting benefit is to enable medical practitioners to prioritise the cases with the greatest urgency and eliminate those that are not requiring treatment.

James Gallagher also noted AI use cases that are able to predict ovarian cancer survival rates and assist in selecting the treatment that should be administered.

Furthermore UCL announced an AI application that identifies those patients who are more likely to miss an appointment with phone calls targeted to remind those patients to attend their appointment.

It is also imperative that hospitals and the healthcare system overall develops a collaborative approach to data sharing and training staff in data skills whilst ensuring that standards for privacy and consent for use of data are adhered to.

AI Winter?

There has been comment from some that we are on the verge of a new AI winter because AI and in particular Machine Learning techniques including Deep Learning have been failing to deliver. I don't believe that we are on the verge of an AI winter. Reviews and criticism can help enhance and improve technologies but equally there can be frustration when faced with a new technology that is complex and that technology has far reaching implications for a given sector.

Furthermore there are plenty of examples of highly successful individuals and companies who are backing existing AI technology.

Entrepreneurs such as Bill Gates who have made vast fortunes have predicted that AI will grow into a vast market, and billionaire Marc Cuban forecast "The world’s first trillionaire will be an Artificial Intelligence entrepreneur". Jeff Bezos stated that "AI is in a ‘golden age’ and solving problems that were once in the realm of sci-fi". Amazon has announced Amazon Go stores that deploy Deep Learning and where shoppers don't need to queue and pay. Alibaba has used AI technology to help it gross $31 billion of sales on Singles Day.

One of the most successful companies on the stock market over the last decade has been Google (reorganised with Alphabet as the parent in 2015).

Alphabet Share Price

Source for Image above: Yahoo Finance

Google was one of the leaders in adopting Deep Learning technology in the wake of AlexNet victory in ImageNet. Dave Gershgorn explains the transformation of Google and Silicon Valley that resulted in the period from 2012 onwards in an article entitled "The inside story of how AI got good enough to dominate Silicon Valley" observing "Thus, the current AI boom was born. Google hired the three researchers to seed a new, major projects using neural nets; the technology’s decision-making prowess soon put the words Deep Learning on the lips of every founder and Silicon Valley executive. Other tech companies like Facebook, Amazon, and Microsoft started positioning their businesses around the tech."

Google has been a strong pioneer in Deep Learning and believes that Machine Learning is its futureSpotify was an early pioneer of Machine Learning and experimented with Deep Learning as early as 2014.

Machine Learning techniques have also been successfully utilised by traditional businesses for example Brad Power authored an article in the Harvard Business Review entitled "How Harley-Davidson Used Artificial Intelligence to Increase New York Sales Leads by 2,930%."

In particular as the volumes of data are set to grow with the anticipated growth of the IoT that 5G and edge computing will drive which in turn will mean continued growth in the demand for Machine Learning. Various forecasts anticipate a substantial growth in connected devices over the next few years. Paul Bevan in "5g, IoT and Edge Computing" observed that "One of the main drivers for Industry 4.0 adoption will be IoT. From 8.4 billion connected devices in 2017 Gartner predicts that there will be 20 billion of these devices by 2020." Paul Bevan also noted that Mobile Health could deliver €99 billion a year, a saving of around 7% on EU health spending based on 2014 values.

A different report from IoT Analytics observed that "The number of connected devices that are in use worldwide now exceeds 17 billion, with the number of IoT devices at 7 billion (that number does not include smartphones, tablets, laptops or fixed line phones)."

Source for Image above: IOT Analytics

IDC forecast that "The Growth in Connected IoT Devices Is Expected to Generate 79.4ZB of Data in 2025". The report also noted that "IDC projects that the amount of data created by these connected IoT devices will see a compound annual growth rate (CAGR) of 28.7% over the 2018-2025 forecast period. Most of the data is being generated by video surveillance applications, but other categories such as industrial and medical will increasingly generate more data over time."

Using Machine Learning to make sense of the data will be essential. Furthermore, I believe that the focus for those who are responsible for running the healthcare system should be to ensure that we move to a data driven system with data sharing.

Data Driven Healthcare

Data is the fuel for Machine Learning. And the availability and quality of the data is essential for enabling Machine Learning to deliver.

It is essential for regulators and those who are responsible for running hospitals and other facilities across the healthcare system to adapt a data driven approach.

An Open Access Government article entitled "Does the future of the NHS rest on its staff learning to ‘speak data’?" noted that "Giving the NHS’s 1.7 million workers proper data analytics tools will have an enormous impact, and so far we have seen only the tip of the iceberg. As more NHS staff, at every level and across all departments, have access to data analytics tools, patient care will transform for the better and the results will become clearer, demonstrating the influence that data can have. Combining the power of data with staff who have the correct analysis tools to view and extract value from the data, presents a very exciting and transformative future for the organisation."

Heather Landi noted in relation to the USA "Fewer than 4 in 10 health systems can successfully share data with other hospitals, survey finds" that "The most crucial elements needed to drive interoperability in health care, according to survey respondents, are commitment by senior leadership, financial incentives or penalties that encourage organizations to share data with one another and with individual patients and advances in tools and technologies.

Dolores Green stated in "Data-sharing can transform care, so let's get connected" that "We’re finally getting it—or at least, we’re getting closer. There’s a growing understanding that the key to helping physicians, hospitals and health plans improve care quality, reduce cost and enhance the patient experience is freeing data. Putting it to work. Getting it out of those silos!"

"Data exchange affords us insight that’s life-saving, cost-effective and that allows providers to make an impact. Utilizing alerts from our data network, we know exactly when our ACO patients have gone to the ER in any hospital in the Inland Empire. Our ACO has seen improved care for patients and reduced healthcare costs. "

"Specifically: We’ve seen a 39.4% reduction in patients not seen within seven days of discharge; ER visits have decreased 3.1% from 2017; ER visits leading to hospitalisation fell by 5.3% from 2017; and per-beneficiary, per-year spend has also decreased dramatically in a year, helping patients live healthier lives more affordably."

"Statistics show that if a high-risk patient is seen by a physician within seven days of discharge from the hospital, the number of ER visits for that patient drops significantly. That sounds simple, but the coordination needed to make this happen is anything but. Technology can be a huge asset, beginning with provider notification that the patient was hospitalised in the first place. "

Enabling and enhancing the adoption of data sharing will be fundamental to improving the performance and cost efficiency of the healthcare system with resulting benefits of better outcomes for the patient. I believe that it will become a major issue for politicians and regulators in the early 2020s in the USA and UK along with other countries in the word and that the next few years will result in a transition towards digital and data driven healthcare system with AI increasingly deployed across the system. Furthermore, the impact of 5G and Edge Computing across the healthcare system in the 2020s will have a massive impact in turn creating greater demand for Machine Learning to make sense of the data generated.

Furthermore, there are medical practitioners who believe that AI will enhance the healthcare system. For example Angela Chen interviewed Eric Topol a senior cardiologist in "A doctor explains how Artificial Intelligence could improve the patient-doctor bond" who stated that"AI can see things that humans can’t. Deep learning trains machines to see things far better than what a human will ever see, and we’re starting to realize all of these things that we never would have guessed before. There are so many examples now. You can determine potassium in your blood on your watch without any blood. You can analyze the retina to see whether it’s male or female with high accuracy. You can analyze a colonoscopy, and machine vision will pick up polyps that are missed by GI doctors. The list goes on and on. "

"The missing piece, of course, is the careful, rigorous, prospective studies with the validation and replication. We have the promise now. We’ve seen enough data and it’s as exciting as anything I’ve seen in my four decades in medicine, but we also need to take it from excitement and hyperbole to the level of reality and unequivocal proof."

My personal contribution for Positive Twitter Day is to post this article with the belief that AI alongside the 5G revolution will play a major role in helping solve the healthcare crisis and result in an improved outcome for the patient.


Original. Reposted with permission.

Bio: Imtiaz Adam is the founder of Deep Learn Strategies Limited (DLS), a Sloan Fellow from the London Business School, and a former Executive Director of Morgan Stanley. Imtiaz studied Computer Science at the postgraduate level with a specialised research project in AI and Deep Learning.