AI and digital tech open up new frontiers in neurology research
Researching treatments for cognitive disorders can often seem like an uphill battle, but new digital tools and AI technologies are changing the game in clinical trials. Brian Mur phy at Cumulus Neuroscience explains the exciting ways in which new tech can outstrip humans in the detection and measurement of brain disorders
Developing impactful treatments for cognitive and neurodegenerative disorders such as dementia or amyotrophic lateral sclerosis (ALS) remains one of the biopharmaceutical industry’s most difficult challenges.
Alzheimer’s disease drug candidates, for example, have one of the highest failure rates of any disease area at 99.6%. But often issues lie not only in the compounds themselves, but in the difficulties faced in measuring their impact.
Much of the recent discussion around emerging therapies for Alzheimer’s concerns the balance between surrogate neurobiological endpoints (e.g., amyloid plaque clearance) and measures that reflect patient well-being and quality of life. This highlights the need for better ways to measure the effectiveness of drugs like this. Without accurate measurements of disease progression, it is hard to know how, and whether an intervention actually is, affecting a patient’s cognition.
Therefore, gathering frequent, real-world data is incredibly important to neurology researchers.
Additionally, many neurodegenerative disorders begin to develop long before symptoms show, which makes it difficult to detect diseases early, and as a result many experimental drugs may be administered to patients too late in their progression to have a significant impact.
There have been several calls from within the industry for greater use of digital technology in clinical trials for cognitive disorders to solve these challenges.
For example, The Digital Medicine Society (DiMe) recently announced the launch of a consortium with several major pharma companies, including Biogen, Eisai, Eli Lilly, Merck & Co and digital biomarker providers, to find new ways of measuring the effectiveness of Alzheimer’s drugs.
Although the life sciences industry is no longer a stranger to sweeping digital transformation, neurology has mostly lagged behind other disease areas in seeing its benefits. For decades, the typical assessments for dementia have centred on clinical interview and ‘paper and pen’ tests, providing the physician only a single snapshot of a person’s symptoms on what could be a good or bad day, and in an artificial environment that is not reflective of the patient’s daily life. This makes interpretation of the impact of any new drug difficult.
The potential for a paradigm shift by integrating digital technology is huge, and the industry is starting to take notice. Often the signs of cognitive decline are far subtler than what can be measured by traditional techniques, so researchers are now turning to AI technology to analyse complex data sets, and harnessing at-home digital tools to improve frequency of measurement and gather real-world data.
There have been some extraordinary strides in digital measurement technology over the last decade, allowing the industry to begin to understand the impact of CNS disorders better than ever before. Some programmes are now able to passively monitor a patient’s everyday smartphone usage, looking for changes in domains such as speed of typing or amount of social media activity that might be indirect signs of changes in cognition.
However, more active and direct digital measures of cognition are now also becoming prominent. With improving speech recognition technology, for instance, researchers can now assess subtle changes in word choice, grammar and the acoustic properties of speech that may indicate neurodegeneration. A patient who uses more pronouns, fewer nouns and more high-frequency words in their speech may be showing signs of semantic impairment, with emptier, vaguer and more non-specific speech. Some of these changes are incredibly difficult for humans to distinguish, such as acoustic abnormalities like slightly longer pauses between words and sentences, and this is one area in which AI can play an important role.
Likewise, research has suggested that many neurodegenerative diseases are associated with impaired recognition of emotion. Software can now assess how easily patients can identify emotions by presenting them with several facial expressions at different levels of intensity; a simple but effective way of gaining an objective measurement of how their perception of social cues is changing.
Digital tools like these also come with more practical benefits. Many of them can be installed on a patient’s own tablet device, avoiding the need for lab assessment, and allowing for data collection at home with much lower trial burden. In fact, there has been a slew of new cognitive assessment tasks developed in recent years that are specifically designed to be used in a real-world setting.
Meanwhile, electroencephalogram (EEG) technology, which has proven benefit in measuring cognitive decline more specifically, has now developed to a point where EEG headsets can be used at home, yielding the same quality data as seen in the clinic. This substantially reduces the burden on the patient, again allowing high quality and high frequency data collection without the need to visit a lab.
Such headsets can also now be combined and integrated with measurements of sleep patterns, which is important since research suggests a link between sleep disturbance and brain disorders like dementia.
For the first time, gathering frequent, longitudinal, real-world data – something of a holy grail for neurology researchers – is becoming achievable. This longitudinal data can then be analysed by AI algorithms, which look for the extremely subtle changes in cognition over time that may signal cognitive decline; changes that can be easy for human researchers to miss.
The future of CNS clinical trials
Easier, more frequent, and more accurate data collection will ultimately result in clinical trials that are less burdensome for patients and have faster development cycles.
This will improve our chances of finding treatments that have a significant impact on cognitive disorders by allowing for better and more timely assessment of drug candidates earlier in development.
It seems Big Pharma agrees, with several companies already supporting these technologies and utilising them in their own
CNS studies. Of course, greater industry buy-in will always be needed to help new digital approaches reach their full potential, but it is not unrealistic to expect that in the near future we may see the wide adoption of a gold standard approach to the digital assessment of cognitive disorders.
We are learning more about these devastating diseases every day, and despite the slew of high-profile trial failures in the past, the industry has never been in a better position to find new and effective treatments.
Digital measurement tools and machine learning technology will only continue to evolve, further improving the robustness of research. Already there are clinical studies testing digital technology in diseases including Alzheimer’s, frontotemporal dementia (FTD) and ALS. With a digital-driven approach, we are hopefully on the precipice of an exciting new era for neurological research, with the adoption of new standards in the conduct of CNS clinical trials, which will bring more new treatments to patients faster.
Brian Murphyis Chief Scientific Officer of Cumulus Neuroscience. Brian is a computational neuroscientist and an expert in brain-reading technology, using machine learning methods to decode cognitive states from recordings of brain activity and other data. He co-founded Cumulus Neuroscience to provide the world’s most advanced, at-home, medical grade, flexible data collection and analytics platform to solve some of the biggest challenges in monitoring functional brain health and brain disorders. Previously, his academic career included positions as senior researcher at Carnegie Mellon University’s Department of Machine Learning, the world’s leading lab for applying machine learning to brain activity, and as post-doc at the Centre for Mind/Brain Sciences, University of Trento, Italy, one of Europe’s leading cognitive neuroscience centres. Prior to academia, Brian worked in and co-founded web start-ups in Germany and China.