AMRC Digital Health Bulletin
9th January 2019
Looking for some digital health reading following the festive break? We’ve compiled some of our highlights from the past few weeks to keep you up-to-date with what’s been going on besides the launch of the NHS Long Term Plan.
New in health tech
- Data has been published on a pair of socks that may be able to detect early warning signs of neuropathy in diabetes via continuous temperature monitoring, with patients rating the technology reasonably highly (19th December).
- IBM Research have announced a fingernail sensor aimed at providing an indication of patients’ grip strength, which could have potential applications across a range of indications including Parkinson’s, schizophrenia, cardiovascular health and geriatric wellbeing (21st December).
- Researchers have developed a wearable that measures pulse oximetry to detect when oxygen levels drop low enough to indicate a potential opioid overdose. Following 10 seconds of sustained low levels, the device sounds an alarm accompanied by flashing red lights, and sends out a text message alert with the wearer’s location (26th December).
- Smartwatches are becoming ever more prolific, providing medically-pertinent data about our bodies in an instant. Some researchers are taking body sensors a step further, however. Researchers at Iota Biosciences have created a device so tiny that it can be fixed to a single nerve or muscle fibre, mainly because it uses ultrasound waves which have a very short wavelength, in the scale of microns. The device receives ultrasound waves and converts these into electricity, creating a voltage that runs through the device. This current is altered by the electrical activity of the tissue it is monitoring, which is in turn detected due to the resultant changes in the ultrasonic waves that bounce back from the device. The sensors could in theory live permanently in the body, either singly or in combination, and could purportedly do ‘everything from monitoring heart tissue to controlling a prosthesis’ (27th December).
- Cancer Research UK are collaborating with Owlstone Medical to run a clinical trial for technology testing patients’ breath to see whether a unique pattern of volatile organic compounds can be detected that are associated with cancerous cells as opposed to typical cellular metabolic processes, potentially paving the way for earlier diagnoses (2nd January).
Ideas from abroad
- This data on use of patient portals in the US highlights well how digital health solutions can increase health inequalities on two levels: whether individuals are offered it or not, and if they are, whether they then take it up (3rd January).
Leveraging data & AI
- 6 new projects applying data science techniques to key research questions relating to cardiovascular health have received over £0.5 million in funding from the Alan Turing Institute and British Heart Foundation (18th December).
- A new programme has invited applications from researchers looking to enhance their medical research by developing a purpose-built image analysis algorithm using the AI platform for free (19th December).
- This article is not health-related but discusses an interesting demonstration of how difficult it can be to ensure that AI provides not only the desired solution but also does so in the desired way (31st December).
- There have been a few articles around the application of AI to brain images in the past few weeks. One being a machine learning tool that can quickly quantify brain atrophy, allowing large image sets to receive visual atrophy ratings, which has applications to dementia both in clinic and in research. Unlike other solutions, the algorithm has broader applicability to various scan types, notably cheaper and more commonly-used CT images, due to its use of voxel intensity values (7th January). Another example is an AI system that was trained to detect acute intracranial haemorrhages. Interestingly, the researchers attempted to design the system to overcome the ‘black box’ problem. The system was trained on the classic, distinguishing features of the five haemorrhage subtypes so that it was then able to show images it determined to be similar to the CT scan being analysed to ‘explain’ its decisions (21st December). Both research groups have highlighted plans to make their tools more widely available for clinicians and/or researchers. Finally, this article highlighted research that had developed an algorithm that could potentially detect Alzheimer’s disease from PET scans 6 years before a clinical diagnosis would be made (2nd January).
- A number of research groups have made promising steps in producing intelligible computer-generated speech based on signals detected by electrodes placed directly on the brain’s surface during surgery, a process that requires the system to be tuned uniquely to each individual. It is hoped that these systems could be further developed for patients unable to speak (though neuronal signals for internal speech, as in these patients, are not identical to those for out loud speech) (2nd January).
- A recently published paper describes the machine-learning algorithms behind a smartphone app that, based on facial images, highlights a range of rare genetic disorders that the patient might have. As is often the case with machine learning, however, biases in the input datasets have resulted in differential performance in across ethnicities. There are also concerns of data silos occurring due to commodification (7th January).
Government schemes, NHS & policy
- The European Commission’s High-Level Expert Group published draft ethics guidelines for trustworthy, which include a proposal of ten ‘Requirements of Trustworthy AI’ on page 14. The draft is open for consultation until January 18th (18th December).
- The Department for Digital, Culture, Media and Sport announced the three projects that received funding from the Digital Inclusion Innovation Fund, covering ‘smart homes’ to help older people improve their digital skills, an app specifically designed for those with Down’s Syndrome and research into the use of technology in palliative care (6th January).