The Future of Artificial Intelligence & Machine Learning in Healthcare Conference took place in London on April 27th.
At the end of April leaders and innovators from the fields of artificial intelligence and healthcare gathered in Old Street—London’s digital tech hub, affectionately known as “silicon roundabout.” The event was hosted by IMT in partnership with Aviva Ventures and brought together participants from academia, industry, and London’s substantial start-up scene.
The day began with a keynote address from Rabin Yaghoubi, Chief Commercial Officer at Babylon Health. Rabin’s team at Babylon created an AI-powered virtual physician app that boasts over 750,000 users worldwide. The keynote introduced a recurring theme of the day: that the global demand for access to healthcare professionals is exponentially outstripping our ability to train them. To combat this problem Babylon’s app provides an AI platform for diagnosis with the possibility of connecting to a doctor for confirmation and a prescription. The work of the Babylon team showed the power of leveraging big data to predict health problems and intervene early, keeping people healthy enough to avoid a visit to a real doctor. Rabin also emphasized that when connecting human interaction with machine learning, the quality and sophistication of any training set must be carefully considered to account for geography, context, and human factors.
Ground-breaking research on facial recognition and feature analysis was presented by Maja Pantic, from Imperial College. The research is underpinned by sophisticated machine learning models that are beginning to extract features we often consider to be readable by humans alone, such as gender, mood, thoughts, and personality. Maja’s work has a particular focus on aging, dementia, and depression; her striking findings suggest that it is the quality of facial expressions, such as smiling and eye movements, that often hold the key to a diagnosis.
Jordi Guitart, from Atem-NT, predicted a future in which clinicians will be training and working hand-in-hand with machine learning systems, thus moving diagnosis “from perception to cognition.” Jordi’s presentation addressed a common misconception of the futurist vision in AI: it is important to remember that machine learning systems are no more accurate than humans, but they do learn significantly faster from their mistakes. The presentation went on to showcase impressive work by the Atem-NT team that used AI-based feature recognition to accurately identify metastases in low-quality imaging that could reduce the time and cost burdens for hospitals by an order of magnitude.
Rupert Whitehead, from Google developer relations, pointed out the increasing number of Google products that are powered by machine learning. Of particular interest was the discussion of Google’s open source TensorFlow machine learning library that is being used in-house at Google and which is also available for anyone to install on systems of different scales, from smart phones to data centers!
Spyros Kotoulas, detailed the work of IBM Research against the startling statistic that in the US and Europe only 10% of the population accounts for 90% of the complexity and cost of healthcare delivery. Spyros spoke of “building software to help people help people” and using the gains in computational power to understand more dimensions of health, including behavior, environment, and genetics. In one of IBM’s social care projects, patient empowerment and decision-making are placed center stage by using AI software to remove mundane tasks for care professionals, allowing more time to prioritize face-to-face engagement.
Bringing a clinical perspective, Indra Joshi, NHS England, spoke about some of the complexities of integrating decision-making software into the clinical realm, and stressing that the power of such systems lies not in replacing clinicians, but in improving clinicians’ ability to do their jobs. A number of example initiatives from within the NHS illustrated how smarter systems are provisioning care more appropriately, improving the outcomes of the very sickest patients, and giving peace of mind to patients receiving palliative care in the community. Indira also cautioned the audience to ask the tough questions that remained unanswered, including: “who should have governance over the data of more than 65 million people in the NHS system?”
In the afternoon session, Billy Boyle, CEO of Owlstone Medical and who was interviewed by Medgadget last year, presented inspiring results of early cancer detection using the company’s innovative breathalyzer system. Capitalizing on the human lung’s fantastic ability to exchange molecules from the blood to breath, the Owlstone medical team has supposedly developed a diagnostic platform that requires only one minute of breathing into a mask to sample the biochemistry of all the blood in the body. This system is coupled with detection algorithms that can offer real-time readouts of blood chemistry and the possibility of monitoring disease progression entirely non-invasively. The Owlstone team hopes to use the technology to improve stage-one cancer diagnoses—in particular, hard-to-diagnose lung and colorectal cancers—and increase the survival rate from a painfully low number that has effectively not changed in decades.
Gareth Stokes and Bonella Ramsay, from DLA Piper, guided the audience though the complex and ever-changing waters of legal issues relating to artificial intelligence and healthcare. Together they offered valuable advice for founders and tech-entrepreneurs on data-sharing collaborations with large healthcare providers, as well as insight into data protection and security. The talk sparked a lively discussion, especially regarding the burden of responsibility for software-generated mistakes and the complexities of IP protection in a system perceptually adapting and altering itself in response to data.
Nuno Godinho, from GE Healthcare, posed the interesting question of how connected machines in the Internet of Things (IoT) will impact healthcare, given the disruption to business and commerce seen from connecting 1 billion individuals over the internet. A number of GE’s projects that blur the line between the physical and digital were discussed, including “digital twins”—entirely virtual instances of real-world objects like jet engines, which may be monitored and tested with both measured data and hypothetical risks to better understand performance. Nuno also predicted an impending clash between clinical and consumer health data, exemplified by the possible health insurance implications of activity trackers, but cautioned the audience that in ignoring such issues we risk obsolescence.
Pearse Keane, of the Moorfields Eye Hospital, spoke about reinventing the eye exam for the 21st century by integrating machine learning into optical coherence tomography (OCT). Pearce showcased new OCT hardware he has co-developed, which allows for increased diagnostic accuracy with more automation and reduced waiting times in the eye exam. The Moorfields team is also collaborating with Google DeepMind, using a dataset of 1.2 million patient images to help automate early detection and exponentially improve outcomes for diabetic retinopathy and age-related macular degeneration patients.
The event was brought to a close by Shafi Ahmed, from Medical Realities, who discussed the so-called Fourth Industrial Revolution and its implications for medicine. He spoke of how the traditional role of the physician—talking to a patient whilst holding their hand—has not changed in 5,000 years, but that disruptive technology has the power to radically change this. Shafi predicted a future “not FaceTiming with a clinician on a screen, but V-Timing by virtually porting your hologram anywhere in the world.” Shafi echoed a common sentiment of the day by insisting we must confront the challenge of disruptive technology in healthcare due to its immense potential power…or in Shafi’s own words: “innovation that is disruptive enough to have put me in Cosmopolitan magazine!”
During the conference, an unparalleled level of technical and policy expertise from speakers and participants alike was on display; lively discussion that spanned disciplines and perspectives continued throughout the day, united by a fervent desire to shape the future of healthcare.
Useful AI and Machine Learning Articles
- The Business of Longevity, a Deep Dive Report Released
- Inside Google’s Strategy to Transform Healthcare Using AI
- Big Data’s Big Impact on Healthcare
- What Does the Future Hold for Medical Robotics?
- Understanding the Medical Diagnosis Processes, to build an AI Solution
- A Day at The Future of AI and ML in Healthcare Conference
- Why AI is Going to Shock Us All