Moving with the times

When the British surgeon James Parkinson first described ‘shaking palsy’ in 1817, he did so by looking at the way patients’ bodies moved.

Two centuries on, most specialists seeking to diagnose Parkinson’s disease still do the same. They rely on physical symptoms to tell them what’s happening in people’s brains: operating largely in the dark.

This lack of understanding contributes hugely to the growing health burden of Parkinson’s, with over 10 million people living with the disease and incidence doubling every 25 years, according to the World Health Organization (WHO).

However, the immense computational capacity of the cloud and the accelerating capabilities of machine learning (ML) and AI are offering new hope.

By transforming our understanding of the brain and how Parkinson’s impacts it, they are able to speed diagnosis, develop new treatments and better empower patients themselves.

Parkinson’s is a progressive disease caused by the loss of dopamine-producing neurons in the brain.

This condition worsens over time. Because the brain relies on dopamine for motor control, it leads to physical symptoms like stiffness, decreased arm movement, reduced blinking or facial expressions, and involuntary shaking or tremors when the body is at rest.

It can also cause less obvious symptoms such as low blood pressure, cognitive impairment, depression, anxiety, hallucinations and delusion. Research shows that people with Parkinson’s are more likely to develop some forms of dementia, broadening the impact of the disease further still.

Because researchers don’t know what causes patients’ dopamine-producing neurons to start shutting down, they are unable to treat the root cause. Instead, most treatments have focused on replacing the lost dopamine.

This can temporarily restore motor function, but can’t prevent the progression of the disease. It also makes misdiagnosis a serious issue, as treatments that boost dopamine and help with Parkinson’s symptoms can worsen those of similar neurological conditions like dementia or essential tremor.

Finding a genuine cure for Parkinson’s involves collecting and analysing a vast number of different types of data, using a much deeper and more granular understanding of the brain to enable new forms of treatment.

Up to 15% of Parkinson’s cases can currently be linked to deletions or mutations in people’s genes. The more DNA data that researchers have to work with, the more of these links they may discover, revealing genetic markers that can help warn of susceptibility to the condition. These enable earlier diagnosis and can signpost the way to treatments.

The California-based company Ultima Genomics has developed software, algorithms and trained its AI models on Amazon Web Services (AWS) for its next-generation DNA sequencer.

This scalable architecture reduces the cost of sequencing an entire human genome from roughly $1,000 to just $100. This can help broaden genetic understanding of the disease and enable the development of gene therapy treatments that can edit DNA to prevent it.

The immense diversity of Parkinson’s symptoms and experiences mean that patients themselves have an invaluable role in advancing medical understanding.

The Michael J. Fox Foundation for Parkinson’s Research (MJFF) is dedicated to finding a cure for Parkinson’s disease, ensuring the development of improved therapies for those living with Parkinson’s today.

As part of a research initiative to evaluate the use of wearable technology to measure and track Parkinson’s symptoms, MJFF partnered with AWS and Intel and is utilising their big data analytics platform to run a number of research projects.

This platform uses scalable big data and Internet of Things (IoT) technologies to collect, process and store large streams of de-identified data from the smartphones and wearable devices of study participants.

Research data is hosted on AWS and is made available to Parkinson’s researchers around the world via Intel’s platform. Through analysis, this data may reveal new insights about living with Parkinson’s and accelerate progress towards a cure.

Proteins aren’t the only potential biomarker for Parkinson’s being investigated with the help of cloud data analysis and AI. Icometrix is using AI imaging solutions to monitor changes in brain tissue volume and explore how these correlate with the advance of the disease.

Rebuilding its deep learning (inference) pipeline has enabled Icometrix to drive big improvements in accuracy while reducing computation time.

This new model decreased the measurement error in brain atrophy and white matter lesion detection by more than 30%, while cutting processing time by 40%. This means that patients and neurologists have faster access to the results, reducing the time a patients spend waiting between the MRI being taken and seeing their neurologist.

Icometrix develops, trains and deploys deep learning algorithms that help clinicians better understand each individual’s unique clinical profile and disease progression in Parkinson’s, including the neurological underpinnings of its motor and cognitive components.

Connecting changes in the brain to changes in people’s experience will represent a huge advance in understanding Parkinson’s. However, a vast amount of what takes place within the brain remains invisible – even to MRI scans.

Mapping changes in the 200 billion cells the brain contains is one of the objectives of the Brain Knowledge Platform: a major new initiative led by the Allen Institute, which is building the world’s largest open-source database of brain cell data.

Combining high-performance computing services alongside AI and machine learning (ML) technologies, the Brain Knowledge Platform decodes the characteristics of different brain cell types and monitors what happens to them as neurological diseases progress.

The Brain Knowledge Platform is beginning to aggregate information about the properties of vulnerable cell populations in Alzheimer’s disease: what they look like; how they function and what the consequence of their loss may be in disease. These cells now become targets for therapies to prevent their degeneration, guiding new treatments.

The platform will become an open registry of neurological data, available to doctors and researchers worldwide. For example, it could enable physicians to better diagnose diseases like Parkinson’s and open the door to new therapies to prevent changes that lead to the loss of dopamine-producing neurons, tackling the root cause of the disease.

Precise mapping of a patient’s brain can enable a wider range of treatments, beyond pharmaceutical approaches. Deep brain stimulation (DBS) electrically stimulates carefully selected areas of the brain to treat neurological movement disorders.

AI and the cloud can help make this treatment option accessible to wider numbers of patients, making treatment more precise and less invasive while reducing side effects. This includes using AI to adjust stimulation therapy to each patient’s brain activity.

Rolling back the burden of Parkinson’s, and improving the lives of those living with the condition, involves approaching the challenge from a number of different directions simultaneously. Greater understanding enables earlier diagnosis and a wider range of treatments that significantly enhance quality of life.

Wider awareness sweeps away stigma and grows interest in technologies that can better support patients. Collective action through clinical trials and research projects increases patients’ sense of agency while bringing a cure closer.

In all of these areas, immense progress is being made through the efforts of Parkinson’s patients, their families, caregivers and medical practitioners. Every one of these groups is discovering they can do even more through the cloud and AI.

Dr Rowland Illing is Chief Medical Officer at AWS