Without brain data, we won’t improve outcomes for patients who have neurological diseases

brain x-ray image


brain x-ray image

As the recent controversy around new Alzheimer’s disease therapies has highlighted, our limited insights into the brain have led to difficulty characterizing disease pathology, flawed clinical trial design and diagnoses with insufficient therapeutic utility—and this is true for many neurological diseases and disorders. In the hope of better outcomes, the precision neurology movement is going to need a trajectory akin to that of the advances in patient and disease specificity within oncology, where data from patients is used to develop and deliver highly precise therapies.

Of course, challenges in the brain are different, but we’ve started to carve a parallel track by harnessing multimodal data from new and existing tech, ranging from medical imaging technology to digital biomarkers to real-world data. To continue this process, we must raze data silos, interconnect disparate volumes of new data in creative ways, and train algorithms to parse it all, in order to paint a more complete picture for both precision medicine development and care.

Diagnosing the brain

Currently, many illnesses of the brain are diagnosed more by symptoms than etiology. There’s no blood test for depression, for example, or a single biomarker for Alzheimer’s. Parkinson’s disease is functionally diagnosed by medication trial-and-error. These diagnostic challenges also have implications for disease progression and therapeutic development. Various parkinsonian syndromes have overlapping symptoms but are caused by different proteins aggregating in distinct parts of the brain, resulting in different rates of progression and complexity in clinical trial design.

The good news is that a wide variety of potential data sources are available or in development that can serve as biomarkers for different aspects of neurological disease. Wearable devices allow for real-time self-reporting and movement detection, while implanted devices are providing a look at the brain from the inside.

Data scientists are training algorithms to detect signs of Alzheimer’s, autism spectrum disorder (ASD), Parkinson’s and depression using tools that analyze voice, odor, GPS or behavior. Our challenge now is to validate and integrate data sets from each source, with the expectation that together they will provide the context needed to impact diagnosis and treatment.

The data dilemma

This new frontier requires accessing and handling sensitive data. That means spending a lot of time carefully navigating the practical, ethical and legal implications of the work.

See also  End of Medicaid continuous coverage represents opportunity to boost patient engagement

The utility of brain data is limited by how it is processed and shared. Delivering clear signals to users—be they clinicians or patients directly—hinges on the reliability of the analytical process, which in part requires a degree of transparency.

It’s also important to consider the legal limitations around how to share health data and with whom, and we have a moral obligation to do so with appropriate context, particularly when showing patients their own data. The push and pull of transparency versus privacy requires evolving regulatory guidance to ensure we can all do the most collective good.

Bridging the synapse

If data analytics can help predict which patients are likely to develop brain-related illness, it will open the door to a range of insights into causes of disease, therapeutic targets and precision patient-matching.

Diseases like Alzheimer’s and Parkinson’s have long confounded drugmakers for myriad reasons, not the least of which is that neurodegeneration may begin years, if not decades, before symptoms occur. Therapeutics are developed to target biological hallmarks of disease, but often appear too late to impact the course of disease. The first amyloid-targeting therapy was finally approved last year, for example, but its benefit may be limited to patients in early stages of diagnosed disease with mild cognitive impairment.

Right now, β amyloid detection via a PET scan is the gold standard for Alzheimer’s diagnosis—though cerebrospinal fluid is sometimes used, and commercial blood tests could be next, once FDA approved. Deep learning algorithms have been shown capable of detecting mild cognitive impairment from functional MRI brain scans.

Most likely, early detection will rely on a combination of approaches, and data from consumer tech can play a role. Digital biomarkers, based on algorithms gleaned from web browsers, cellphones or GPS usage data, can already identify symptoms from behavior. Clinicians, who have long hoped that wearable technologies could deliver useful data between clinical visits, are seeing progress. Consumer devices like smart watches can detect movements typical of the motor disease symptoms, dangerous falls and physiological metrics like stress responses or sleep patterns that often contribute to symptom spikes.

See also  Healthcare workforces need sustainable systems

Patients can also use consumer devices to manually track their medication schedules and dietary patterns, and record unique or worsening symptoms that can’t be detected automatically.

That level of patient feedback helps guide diagnosis and therapy optimization, but is the tip of the iceberg. Implantable electrical stimulation devices are increasingly tapped for symptom management in a number of neurological conditions, including Parkinson’s, obsessive compulsive disorder (OCD) and depression.

Some of the newest deep-brain stimulation (DBS) devices are adaptive, incorporating sensors that capture feedback from brain signaling to modulate impulses. The feedback data is also transmitted externally, used to improve algorithms that dictate the timing and pattern of electrical pulses.

By linking directly recorded brain data with behavioral data, we can further optimize treatment, personalize medication regimens and respond nimbly to symptom progression. A recent example is adaptive DBS in patients with severe OCD, a disease with no reliable biomarkers today.

In a Brown University study, adaptive DBS data was combined with computer-recorded facial expressions and body movement in the clinic, alongside self-reported symptom intensity and biometrics from wearables. Researchers were able to develop machine learning algorithms that identified potential OCD biomarkers, to be confirmed in larger studies.

Forward thinking

Distinguishing mental illnesses, neurodegenerative diseases and other central nervous system disorders based on symptoms harkens back to the days when cancers were exclusively identified by the organ where they were discovered.

Could aggregate brain data biomarkers distinguish between diseases we lump together as “parkinsonism” today, but distinguish as Parkinson’s, multiple systems atrophy, progressive supranuclear palsy or corticobasal degeneration in autopsy? Can the umbrella of physiologically similar anxiety disorders, broken into subtypes by symptoms like social anxiety disorder and agoraphobia, be characterized in a way that guides therapeutic development? Everything we do to bring precision medicine to neuroscience shows us this is possible.

A precision neuromedicine movement, powered by AI to integrate data from direct brain measurement, wearable technology and patient experience, will unlock a new generation of therapies across the spectrum of neurological illness.

Photo: Jolygon, Getty Images


Source link