Shots:

• Jay spoke about the role of AI-based EEG neurobiomarkers in the development of therapies to treat neurologic diseases
• Jay also talked about the use of the EEG analytics platform for dose determination clinical research
• The interview gives an understanding of Beacon Biosignals' vision to accelerate new treatment options for neurological and psychiatric diseases

Smriti: Throw some light on your EEG analytics platform. 

Jay Pathmanathan: Beacon Biosignals’ platform provides insights into brain disease and the effects of drugs that act on the brain in ways that previously have been impossible. Our machine learning platform for EEG enables and accelerates new treatments that transform the lives of patients with neurological, psychiatric, or sleep disorders. Novel machine-learning algorithms, massive clinical datasets, and advances in software engineering allow Beacon to empower biopharma companies with unparalleled tools for efficacy monitoring, patient stratification, and clinical trial endpoints from brain data. Our analytics platform scalably ingests EEG data from any standard acquisition format to identify novel neurobiomarkers from within that data and map them to novel treatment effects. Our approach and specific technology set us apart. 

Smriti: What role does EEG data play in the development of neurological therapies?

Jay Pathmanathan: EEG currently is the gold standard clinical diagnostic for epilepsy-related disorders and sleep disorders. As a result, it's widely accepted and utilized in clinical trials for those disorders. However, recording and analyzing the full breadth and depth of EEG data is challenging from a technical perspective, due to the complexity of setup, large volume of data, and limited availability of human experts. Therefore, when EEG is acquired, it tends to be in limited quantity and at sparse time points during a clinical trial. This means you may get an EEG on day one or day two, provide treatment for 100 days, and then run another EEG to assess the effects of the treatment. Having scalable machine-learning tools makes it feasible to analyze more EEG data than was previously possible, which overcomes the limitations of infrequent data collection. 

The clinical benefit of EEG data is it gives us an objective measure of brain activity at very short time scales; unlike traditional imaging, which looks at anatomical structure, EEG is time-series data that captures brain function at short time scales. Brain activity patterns (including changes in activity in sleep) may demonstrate changes in neurodegenerative disease, epilepsy, psychiatric disease state, or level of alertness that cannot be appreciated at the scale of MRI.

This allows new insights into diseases such as Alzheimer’s Disease, major depressive disorder, schizophrenia, and the effects of treatments. EEG data can be used to determine whether it’s safe to enroll a patient in a trial, to see if they have features of a specific disease and whether those features are being modulated by a treatment.

Smriti: Explain the mechanism of the AI-based EEG neurobiomarker.

Jay Pathmanathan: There are many features in EEGs that can appear over the course of very long EEG recordings. Previous approaches to extracting EEG biomarkers have required manual interrogation of data and visually counting events of interest in subsets of data. This simply isn’t possible for extremely large data sets, especially in Phase 3 clinical trials with hundreds of patients and thousands of hours of EEG data. 

Training machine-learning tools to identify and cluster events of interest allows us to extract the full depth and richness of the data from the entire recording. Once all events of interest have been flagged, you can derive quantitative metrics from them. You can capture every spike in activity and measure the amplitude of every event, something that is impossible for humans to do at scale. Machine-learning tools also enable replicability because an algorithm will perform identically every time it runs. Algorithms are easily adjusted, too, which means they can more accurately represent the consensus of trained experts.

Smriti: How is this platform efficient in dose-finding?

Jay Pathmanathan: Dose finding for drugs for clinical neurosciences therapies historically is difficult. Mapping of the dose-response curve in preclinical studies doesn't always inform the design of dosing strategies in humans – and particularly humans with brain disease. Now, though, we have tools that can reliably capture greater depth and breadth of information within an EEG and quantify changes in neural activity patterns after giving a new therapy. Having this type of quantitative precision enables us to provide a novel tool for examining drug effects at different dose levels on the brain, and provide a mapping of EEG features to PK/PD curves.

Smriti: Tell us about the role of EEG neurobiomarkers in psychiatric diseases.

Jay Pathmanathan: While a lot of EEG data is available for neurological diseases, the idea of integrating an understanding of neurophysiology for psychiatric disease clinical trials is new. As with neurological disorders, we’re trying to understand if there are features of a specific patient that might make them more likely to respond to therapy. 

Further complicating care plan design, psychiatric diseases are extremely heterogeneous and symptoms can fluctuate over months, days, or even hours. EEGs give you a more dynamic measure of neural activity to map the symptoms a patient is experiencing. EEG neurobiomarkers efficiently capture some of the heterogeneity of psychiatric diseases. This is a more objective way of seeing brain function in groups of patients that often cannot reliably report their feelings or outcomes. When you have a group of patients who are unable to give you dependable or consistent outputs, because of the shifting and heterogeneous nature of their disease, then you need a different way of obtaining data. EEG provides insight into the organ responsible for psychiatric disease. For example, most psychiatric diseases affect a variety of brain subsystems – with mood networks being the obvious (but also more difficult to assess). However, mood networks in the limbic system are intimately connected to limbic networks regulating sleep-wake cycles, and EEG is exquisitely sensitive to subtle changes in sleep metrics. These metrics provide insight into disease state and treatment effects, but show changes at time scales that are much shorter than the weeks to months it might take to see a change in clinical symptoms.

Smriti: List various other services provided by you.

Jay Pathmanathan: Our main service is providing unparalleled tools for efficacy monitoring, patient stratification, and clinical trial endpoints for brain data. In addition to EEG clinical trial services, we offer translational research, where we analyze existing EEG results for insights into the mechanism of action, PK/PD, and patient stratification to guide research and development efforts. Another service is qualified biomarker development. We develop and validate neurobiomarkers to improve cohort selection, de-confound standard clinical assessments, and identify novel endpoints.

Our efficacy and dose-finding service employ direct neurophysiologic insights to improve dose-finding and identify promising responses to properly power subsequent trials. We also have a continuous analysis service that aggregates and analyzes trial data in real-time to ensure consistency across trial sites, prioritize patient safety and reduce trial delays.

Smriti: Enumerate different software integrated into EEG neurobiomarker.

Jay Pathmanathan: Some people argue that EEG will not capture all the information you need to understand a disease. Historically that has been true and will continue to be true for some diseases. But our platform can integrate multisensory/mixed modality data, clinical metadata, and patient characteristics such as age, sex, and genetic information to closely tailor the EEG neurobiomarker to a specific patient. Brain activity is dynamic and changes across patient characteristics in genetics, so we can bring together this total breadth of information to precisely identify features that are descriptive of a particular disease or disorder for an individual.

Smriti: What other programs are Beacon working on and in which areas?

Jay Pathmanathan: In neurology, we are heavily invested in epilepsy, particularly pediatric epilepsy. However, we are working on Alzheimer’s Disease and other neurodegenerative conditions. In psychiatry, we’re working on sleep disorders and schizophrenia. Of course, we have an active R&D arm investigating new areas, but those are not ready for disclosure.

Source: Canva

About the Author:

Jay Pathmanathan is the Medical Director at Beacon Biosignals. He is a clinical neurophysiologist with 10+ years of experience in clinical epilepsy, including surgical evaluation of medication refractory seizures using invasive electrodes and technical EEG including EEG hardware and software. He earned a BS degree in Biomedical Engineering and Minor Economics from Johns Hopkins University and an MD, Ph.D. in Medicine And Neuroscience from The University Of Connecticut School of Medicine.
Related Post: Kindra Whitlatch, Product Applications Manager at ThermoFisher Scientific, Shares Insights on Drug Stability Testing