Neuroscience technology company Brooklyn Health is using artificial intelligence to target a fundamental problem in neurology and psychiatry: the flawed approach to mental health outcomes measurement.
The startup aims to modernize mental health measurement and scoring in central nervous system drug development, an area of CNS research that faces limitations in objectivity and standardization.
“Clinical interviews, the standard for symptom assessment, are fundamentally unreliable and imprecise," said Anzar Abbas, Ph.D., a neuroscientist and founder of Brooklyn Health, in an interview.
Brooklyn's platform uses AI and digital phenotyping methods it developed to evaluate the quality and scoring of clinical interviews in real time.
"Measurement is a pretty core problem in neurology and psychiatry, because it's just so tough to quantify what somebody has and how badly their symptoms are manifesting," Abbas said.
"We rely on these subjective clinical interviews, but they can be flawed for a multitude of reasons. What Brooklyn Health is focused on is solving that measurement problem in mental health by talking through the underlying science of, how we can more objectively phenotype patient behavior and quantify their illness, but, also, how we can build scalable tools around that foundational scientific work?" Abbas said.
Brooklyn's current focus is on drug development and CNS clinical trials to improve outcome measurement—essentially, how well a drug is working. But the company has ambitions beyond clinical trials to support outcome measurement across all forms of behavioral health delivery, including in-clinic psychiatric care and virtual mental health platforms.
"Our tech is really a step function forward from the status quo today, which relies on a lot of manual review and manual work to using tech to more objectively quantify patient health and functioning and have that be the measures that inform whether or not a drug is working," he said.
CNS clinical trials rely on clinical interviews as the primary outcome measure for treatment efficacy. The interviews allow for scoring of symptoms through observation and are necessary to measure change in response to treatment. But these clinical interviews are difficult to standardize, and scoring is subjective and susceptible to biases. This leads to unreliable outcome measures and is associated with placebo response, both of which directly contribute to the high failure rate of CNS clinical trials and cost the industry billions of dollars annually, according to the company.
Pharmaceutical sponsors will record these clinical interviews for a second round of clinician reviews to ensure quality and score accuracy.
Brooklyn launched its electronic clinical outcome assessment (eCOA) solution, called Willis, which automates these legacy services and provides an AI-powered review of clinical interview quality and score accuracy. Clinicians get real-time feedback on interview administration, and pharmaceutical sponsors have visibility into data quality at scale, according to the company.
"As soon as the recording comes in, the interview gets transcribed and it goes through our models. The models will do that secondary review and make sure that the interview was administered correctly and that the scores were applied correctly as well," Abbas said, adding, "This is something that's already done manually, but it's not scalable, and it doesn't really provide the level of reliability that pharmaceutical sponsors are looking for."
Brooklyn recently secured $6.5 million in seed funding to expand its team, enhance Willis’ capabilities and accelerate commercial deployments with pharmaceutical companies.
HealthX Ventures led the round, with participation from Metrodora Ventures, Story Ventures, RiverPark Ventures, Laconia Capital, Everywhere Ventures, Hypothesis Studio, Blue Falcon Capital and other investors.
With Willis, Brooklyn is modernizing the technology stack for drug development, moving it from services to software, Abbas noted.
The platform includes an intuitive user experience, native clinician training, real-time flagging of concerning events, robust data analytics and easier communication between clinical sites and pharmaceutical sponsors all in a secure and scalable cloud architecture, according to the company.
"The big picture is that pharma just doesn't have a platform to run their trials on. Everything is happening over email. Everything's happening over phone calls or Zoom calls. What this allows the field to do is have a unified place where all that information lives. It all gets updated in real time," Abbas asserts.
Brooklyn has secured partnerships with major pharmaceutical companies including Bristol Myers Squibb and Boehringer Ingelheim.
"We've been able to validate a lot of this underlying science in partnership with those companies. That's been a big part of our success. We were trusted early on by these bigger names, and it has really helped us with building scientific and clinical credibility in the field," Abbas noted.
Its technology has been used by researchers at Harvard University, Yale University, UCLA, Columbia University and other organizations, the company said.
Brooklyn's eCOA platform, Willis, utilizes digital phenotyping to collect and process real-time behavioral data.
"Digital phenotyping is a field that is relatively novel, and the core motivator is to more objectively measure the behavioral indicators of illness," Abbas said.
As an example, an individual with schizophrenia may have a symptom called "flat affect," characterized by a reduced or absent emotional expression. "That's something that a clinician knows to look for. It's well documented in psychology literature that that's a symptom, but it's very subjective to quantify. When you use computer vision to look at hundreds of facial landmarks and how much they're displaced from frame to frame, you can do it a lot more objectively," he said.
In a recent analysis Brooklyn conducted for one of its pharma customers, it was demonstrated that using Willis to score clinical interviews in a psychiatry study improved separation of drug from placebo by 34% as measured by effect size, highlighting the potential impact of the technology on trial outcomes.
“Brooklyn Health is directly addressing what has ailed CNS drug development for decades: endpoint quality and placebo response,” said Mark Bakken, founder and managing partner at HealthX Ventures, in a statement.
Central to Brooklyn’s approach is OpenWillis, an open-source Python library for digital phenotyping, serving as the foundation of its measurement technology. Brooklyn made its core methods available to the scientific community while building proprietary tools to scale them commercially.
RELATED
Clario acquires WCG's neuroscience-focused clinical outcome assessment business
OpenWillis provides researchers with a simple toolkit for quantifying facial emotions, voice and speech characteristics, motor functioning and other behavioral indicators of mental health. This approach helps foster a community-driven approach to validation of novel methods in digital phenotyping, Abbas said.
Working with pharma companies in clinical trials for drug development represents a commercially viable path for Brooklyn's technology, Abbas noted. "There's a very dire and tangible need for this in drug development. If you ask pharma companies, 'Why do trials fail?' they'll give you one reason only, which is endpoint quality," he said.
But solving the measurement problem in mental health extends beyond clinical trials and research, Abbas asserts.
"The goal is very much to enable clinicians to provide more effective and targeted care to their patients. How do you get there is a really tough question," he said. "If we can lower the barrier for drug development, if we reduce the risk that these trials usually come with, then pharma can invest in more targeted treatments that apply to specific portions of populations, rather than optimizing for as broad of a population as possible. When that happens, you have more targeted treatments available in the market. When you use digital phenotyping to develop those treatments, you have a lot more nuanced understanding of what kinds of phenotypes each drug is actually helping with."
More targeted treatments will lead to more precise treatment plans for patients, he noted.
"Other practices of medicine have moved on to much more precise treatment plans, such as in oncology, cardiology and immunology. We don't have that in neurology and psychiatry, because we don't have the drugs that are specific enough," he said.
Brooklyn's AI-based technology and approach to clinical assessments also could help advance research into neurological disorders, according to Abbas.
"In the field of psychiatry and neurology, our current classification of these brain disorders is outdated, and we want to be able to, in a more nuanced way, understand all the different ways that something can be wrong with the brain and have a tool kit to address that more precisely," he noted.
The company's name is a nod to where the company is based—Brooklyn Heights—but also a tribute to the startup's employees, Abbas said.
"We want to celebrate the employees that are in the weeds. For me, it's a constant reminder of that mission," he said.