In an initial study, Canary analyzed speech recordings with AI and identified early stage Alzheimer’s with 96 percent accuracy.
Dave Arnold retired in his 60s and began spending time volunteering in local schools. But then he started misplacing items, forgetting appointments and losing his sense of direction. Eventually he was diagnosed with early stage Alzheimer’s.
“Hearing the diagnosis made me very emotional and tearful,” he said. “I immediately thought of all my mom had experienced.” His mother suffered with the condition for years before passing away. Over the last year, Arnold has worked for the Alzheimer’s Association as one of its early stage advisors, sharing his insights to help others in the initial stages of the disease.
Arnold was diagnosed sooner than many others. It's important to find out early, when interventions can make the most difference. One promising avenue is looking at how people talk. Research has shown that Alzheimer’s affects a part of the brain that controls speech, resulting in small changes before people show other signs of the disease.
Now, Canary Speech, a company based in Utah, is using AI to examine elements like the pitch of a person’s voice and their pauses. In an initial study, Canary analyzed speech recordings with AI and identified early stage Alzheimer’s with 96 percent accuracy.
Developing the AI model
Canary Speech’s CEO, Henry O’Connell, met cofounder Jeff Adams about 40 years before they started the company. Back when they first crossed paths, they were both living in Bethesda, Maryland; O’Connell was a research fellow at the National Institutes of Health studying rare neurological diseases, while Adams was working to decode spy messages. Later on, Adams would specialize in building mathematical models to analyze speech and sound as a team leader in developing Amazon's Alexa.
It wasn't until 2015 that they decided to make use of the fit between their backgrounds. ““We established Canary Speech in 2017 to build a product that could be used in multiple languages in clinical environments,” O'Connell says.
The need is growing. About 55 million people worldwide currently live with Alzheimer’s, a number that is expected to double by 2050. Some scientists think the disease results from a buildup of plaque in the brain. It causes mild memory loss at first and, over time, this issue get worse while other symptoms, such as disorientation and hallucinations, can develop. Treatment to manage the disease is more effective in the earlier stages, but detection is difficult since mild symptoms are often attributed to the normal aging process.
O’Connell and Adams specialize in the complex ways that Alzheimer’s effects how people speak. Using AI, their mathematical model analyzes 15 million data points every minute, focusing on certain features of speech such as pitch, pauses and elongation of words. It also pays attention to how the vibrations of vocal cords change in different stages of the disease.
To create their model, the team used a type of machine learning called deep neural nets, which looks at multiple layers of data - in this case, the multiple features of a person’s speech patterns.
“Deep neural nets allow us to look at much, much larger data sets built out of millions of elements,” O’Connell explained. “Through machine learning and AI, we’ve identified features that are very sensitive to an Alzheimer’s patient versus [people without the disease] and also very sensitive to mild cognitive impairment, early stage and moderate Alzheimer's.” Based on their learnings, Canary is able to classify the disease stage very quickly, O’Connell said.
“When we’re listening to sublanguage elements, we’re really analyzing the direct result of changes in the brain in the physical body,” O’Connell said. “The brain controls your vocal cords: how fast they vibrate, the expansion of them, the contraction.” These factors, along with where people put their tongues when talking, function subconsciously and result in subtle changes in the sounds of speech.
Further testing is needed
In an initial trial, Canary analyzed speech recordings from phone calls to a large U.S. health insurer. They looked at the audio recordings of 651 policyholders who had early stage Alzheimer’s and 1018 who did not have the condition, aiming for a representative sample of age, gender and race. They used this data to create their first diagnostic model and found that it was 96 percent accurate in identifying Alzheimer’s.
Christian Herff, an assistant professor of neuroscience at Maastricht University in the Netherlands, praised this approach while adding that further testing is needed to assess its effectiveness.
“I think the general idea of identifying increased risk for cognitive impairment based on speech characteristics is very feasible, particularly when change in a user’s voice is monitored, for example, by recording speech every year,” Herff said. He noted that this can only be a first indication, not a full diagnosis. The accuracy still needs to be validated in studies that follows individuals over a period of time, he said.
Toby Walsh, a professor of artificial intelligence at the University of New South Wales, also thinks Canary’s tool has potential but highlights that Canary could diagnose some people who don’t really have the disease. “This is an interesting and promising application of AI,” he said, “but these tools need to be used carefully. Imagine the anxiety of being misdiagnosed with Alzheimer’s.”
As with many other AI tools, privacy and bias are additional issues to monitor closely, Walsh said.
Other languages
A related issue is that not everyone is fluent in English. Mahnaz Arvaneh, a senior lecturer in automatic control and systems engineering at the University of Sheffield, said this could be a blind spot.
“The system may not be very accurate for those who have English as their second language as their speaking patterns would be different, and any issue might be because of language deficiency rather than cognitive issues,” Arvaneh said.
The team is expanding to multiple languages starting with Japanese and Spanish. The elements of the model that make up the algorithm are very similar, but they need to be validated and retrained in a different language, which will require access to more data.
Recently, Canary analyzed the phone calls of 233 Japanese patients who had mild cognitive impairment and 704 healthy people. Using an English model they were able to identify the Japanese patients who had mild cognitive impairment with 78 percent accuracy. They also developed a model in Japanese that was 45 percent accurate, and they’re continuing to train it with more data.
The future
Canary is using their model to look at other diseases like Huntington’s and Parkinson’s. They’re also collaborating with pharmaceuticals to validate potential therapies for Alzheimer’s. By looking at speech patterns over time, Canary can get an indication of how well these drugs are working.
Dave Arnold and his wife dance at his nephew’s wedding in Rochester, New York, ten years ago, before his Alzheimer's diagnosis.
Dave Arnold
Ultimately, they want to integrate their tool into everyday life. “We want it to be used in a smartphone, or a teleconference call so that individuals could be examined in their home,” O’Connell said. “We could follow them over time and work with clinical teams and hospitals to improve the evaluation of patients and contribute towards an accurate diagnosis.”
Arnold, the patient with early stage Alzheimer’s, sees great promise. “The process of getting a diagnosis is already filled with so much anxiety,” he said. “Anything that can be done to make it easier and less stressful would be a good thing, as long as it’s proven accurate.”
The U.S. population is becoming more diverse, but clinical trials don't reflect that, experts say. Some are focusing on recruiting minorities to participate in research.
While racial and ethnic minority groups represent almost half of the U.S. population, the lack of diversity in clinical trials poses serious challenges. Limited awareness and access impede equitable representation, which is necessary to prove the safety and effectiveness of medical interventions across different groups.
A Yale University study on clinical trial diversity published last year in BMJ Medicine found that while 81 percent of trials testing the new cancer drugs approved by the U.S. Food and Drug Administration between 2012 and 2017 included women, only 23 percent included older adults and 5 percent fairly included racial and ethnic minorities. “It’s both a public health and social justice issue,” said Jennifer E. Miller, an associate professor of medicine at Yale School of Medicine. “We need to know how medicines and vaccines work for all clinically distinct groups, not just healthy young White males.” A recent JAMA Oncology editorial stresses out the need for legislation that would require diversity action plans for certain types of trials.
Ensuring meaningful representation of racial and ethnic minorities in clinical trials for regulated medical products is fundamental to public health.--FDA Commissioner Robert M. Califf.
But change is on the horizon. Last April, the FDA issued a new draft guidance encouraging industry to find ways to revamp recruitment into clinical trials. The announcement, which expanded on previous efforts, called for including more participants from underrepresented racial and ethnic segments of the population.
“The U.S. population has become increasingly diverse, and ensuring meaningful representation of racial and ethnic minorities in clinical trials for regulated medical products is fundamental to public health,” FDA commissioner Robert M. Califf, a physician, said in a statement. “Going forward, achieving greater diversity will be a key focus throughout the FDA to facilitate the development of better treatments and better ways to fight diseases that often disproportionately impact diverse communities. This guidance also further demonstrates how we support the Administration’s Cancer Moonshot goal of addressing inequities in cancer care, helping to ensure that every community in America has access to cutting-edge cancer diagnostics, therapeutics and clinical trials.”
Lola Fashoyin-Aje, associate director for Science and Policy to Address Disparities in the Oncology Center of Excellence at the FDA, said that the agency “has long held the view that clinical trial participants should reflect the clinical and demographic characteristics of the patients who will ultimately receive the drug once approved.” However, “numerous studies over many decades” have measured the extent of underrepresentation. One FDA analysis found that the proportion of White patients enrolled in U.S. clinical trials (88 percent) is much higher than their numbers in country's population. Meanwhile, the enrollment of African American and Native Hawaiian/American Indian and Alaskan Native patients is below their national numbers.
The FDA’s guidance is accelerating researchers’ efforts to be more inclusive of diverse groups in clinical trials, said Joyce Sackey, a clinical professor of medicine and associate dean at Stanford School of Medicine. Underrepresentation is “a huge issue,” she noted. Sackey is focusing on this in her role as the inaugural chief equity, diversity and inclusion officer at Stanford Medicine, which encompasses the medical school and two hospitals.
Until the early 1990s, Sackey pointed out, clinical trials were based on research that mainly included men, as investigators were concerned that women could become pregnant, which would affect the results. This has led to some unfortunate consequences, such as indications and dosages for drugs that cause more side effects in women due to biological differences. “We’ve made some progress in including women, but we have a long way to go in including people of different ethnic and racial groups,” she said.
A new mural outside Guadalupe Centers Middle School in Kansas City, Missouri, advocates for increasing diversity in clinical trials. Kim Jones, 51-year-old African American breast cancer survivor, is second on the left.
Artwork by Vania Soto. Photo by Megan Peters.
Among racial and ethnic minorities, distrust of clinical trials is deeply rooted in a history of medical racism. A prime example is the Tuskegee Study, a syphilis research experiment that started in 1932 and spanned 40 years, involving hundreds of Black men with low incomes without their informed consent. They were lured with inducements of free meals, health care and burial stipends to participate in the study undertaken by the U.S. Public Health Service and the Tuskegee Institute in Alabama.
By 1947, scientists had figured out that they could provide penicillin to help patients with syphilis, but leaders of the Tuskegee research failed to offer penicillin to their participants throughout the rest of the study, which lasted until 1972.
Opeyemi Olabisi, an assistant professor of medicine at Duke University Medical Center, aims to increase the participation of African Americans in clinical research. As a nephrologist and researcher, he is the principal investigator of a clinical trial focusing on the high rate of kidney disease fueled by two genetic variants of the apolipoprotein L1 (APOL1) gene in people of recent African ancestry. Individuals of this background are four times more likely to develop kidney failure than European Americans, with these two variants accounting for much of the excess risk, Olabisi noted.
The trial is part of an initiative, CARE and JUSTICE for APOL1-Mediated Kidney Disease, through which Olabisi hopes to diversify study participants. “We seek ways to engage African Americans by meeting folks in the community, providing accessible information and addressing structural hindrances that prevent them from participating in clinical trials,” Olabisi said. The researchers go to churches and community organizations to enroll people who do not visit academic medical centers, which typically lead clinical trials. Since last fall, the initiative has screened more than 250 African Americans in North Carolina for the genetic variants, he said.
Other key efforts are underway. “Breaking down barriers, including addressing access, awareness, discrimination and racism, and workforce diversity, are pivotal to increasing clinical trial participation in racial and ethnic minority groups,” said Joshua J. Joseph, assistant professor of medicine at the Ohio State University Wexner Medical Center. Along with the university’s colleges of medicine and nursing, researchers at the medical center partnered with the African American Male Wellness Agency, Genentech and Pfizer to host webinars soliciting solutions from almost 450 community members, civic representatives, health care providers, government organizations and biotechnology professionals in 25 states and five countries.
Their findings, published in February in the journal PLOS One, suggested that including incentives or compensation as part of the research budget at the institutional level may help resolve some issues that hinder racial and ethnic minorities from participating in clinical trials. Compared to other groups, more Blacks and Hispanics have jobs in service, production and transportation, the authors note. It can be difficult to get paid leave in these sectors, so employees often can’t join clinical trials during regular business hours. If more leaders of trials offer money for participating, that could make a difference.
Obstacles include geographic access, language and other communications issues, limited awareness of research options, cost and lack of trust.
Christopher Corsico, senior vice president of development at GSK, formerly GlaxoSmithKline, said the pharmaceutical company conducted a 17-year retrospective study on U.S. clinical trial diversity. “We are using epidemiology and patients most impacted by a particular disease as the foundation for all our enrollment guidance, including study diversity plans,” Corsico said. “We are also sharing our results and ideas across the pharmaceutical industry.”
Judy Sewards, vice president and head of clinical trial experience at Pfizer’s headquarters in New York, said the company has committed to achieving racially and ethnically diverse participation at or above U.S. census or disease prevalence levels (as appropriate) in all trials. “Today, barriers to clinical trial participation persist,” Sewards said. She noted that these obstacles include geographic access, language and other communications issues, limited awareness of research options, cost and lack of trust. “Addressing these challenges takes a village. All stakeholders must come together and work collaboratively to increase diversity in clinical trials.”
It takes a village indeed. Hope Krebill, executive director of the Masonic Cancer Alliance, the outreach network of the University of Kansas Cancer Center in Kansas City, which commissioned the mural, understood that well. So her team actively worked with their metaphorical “village.” “We partnered with the community to understand their concerns, knowledge and attitudes toward clinical trials and research,” said Krebill. “With that information, we created a clinical trials video and a social media campaign, and finally, the mural to encourage people to consider clinical trials as an option for care.”
Besides its encouraging imagery, the mural will also be informational. It will include a QR code that viewers can scan to find relevant clinical trials in their location, said Vania Soto, a Mexican artist who completed the rendition in late February. “I’m so honored to paint people that are survivors and are living proof that clinical trials worked for them,” she said.
Jones, the cancer survivor depicted in the mural, hopes the image will prompt people to feel more open to partaking in clinical trials. “Hopefully, it will encourage people to inquire about what they can do — how they can participate,” she said.
Drones are just one of several new technologies that are rising to the challenge of more frequent wildfires.
Robots have been a game-changer. Inspired by Paris, the Los Angeles Fire Department (LAFD) was the first in the U.S. to deploy a firefighting robot in 2021, the Thermite Robotics System 3 – RS3, for short.
RS3 is a 3,500-pound turbine on a crawler—the size of a Smart car—with a 36.8 horsepower engine that can go for 20 hours without refueling. It can plow through hazardous terrain, move cars from its path, and pull an 8,000-pound object from a fire.
All that while spurting 2,500 gallons of water per minute with a rear exhaust fan clearing the smoke. At a recent trade show, RS3 was billed as equivalent to 10 firefighters. The Los Angeles Times referred to it as “a droid on steroids.”
Robots such as the Thermite RS3 can plow through hazardous terrain and pull an 8,000-pound object from a fire.
Los Angeles Fire Department
The advantage of the robot is obvious. Operated remotely from a distance, it greatly reduces an emergency responder’s exposure to danger, says Wade White, assistant chief of the LAFD. The robot can be sent into airplane fires, nuclear reactors, hazardous areas with carcinogens (think East Palestine, Ohio), or buildings where a roof collapse is imminent.
Advances for firefighters are taking many other forms as well. Fibers have been developed that make the firefighter’s coat lighter and more protective from carcinogens. New wearable devices track firefighters’ biometrics in real time so commanders can monitor their heat stress and exertion levels. A sensor patch is in development which takes readings every four seconds to detect dangerous gases such as methane and carbon dioxide. A sonic fire extinguisher is being explored that uses low frequency soundwaves to remove oxygen from air molecules without unhealthy chemical compounds.
The demand for this technology is only increasing, especially with the recent rise in wildfires. In 2021, fires were responsible for 3,800 deaths and 14,700 injuries of civilians in this country. Last year, 68,988 wildfires burned down 7.6 million acres. Whether the next generation of firefighting can address these new challenges could depend on special cameras, robots of the aerial variety, AI and smart systems.
Fighting fire with cameras
Another key innovation for firefighters is a thermal imaging camera (TIC) that improves visibility through smoke. “At a fire, you might not see your hand in front of your face,” says White. “Using the TIC screen, you can find the door to get out safely or see a victim in the corner.” Since these cameras were introduced in the 1990s, the price has come down enough (from $10,000 or more to about $700) that every LAFD firefighter on duty has been carrying one since 2019, says White.
TICs are about the size of a cell phone. The camera can sense movement and body heat so it is ideal as a search tool for people trapped in buildings. If a firefighter has not moved in 30 seconds, the motion detector picks that up, too, and broadcasts a distress signal and directional information to others.
To enable firefighters to operate the camera hands-free, the newest TICs can attach inside a helmet. The firefighter sees the images inside their mask.
TICs also can be mounted on drones to get a bird’s-eye, 360 degree view of a disaster or scout for hot spots through the smoke. In addition, the camera can take photos to aid arson investigations or help determine the cause of a fire.
More help From above
Firefighters prefer the term “unmanned aerial systems” (UAS) to drones to differentiate them from military use.
A UAS carrying a camera can provide aerial scene monitoring and topography maps to help fire captains deploy resources more efficiently. At night, floodlights from the drone can illuminate the landscape for firefighters. They can drop off payloads of blankets, parachutes, life preservers or radio devices for stranded people to communicate, too. And like the robot, the UAS reduces risks for ground crews and helicopter pilots by limiting their contact with toxic fumes, hazardous chemicals, and explosive materials.
“The nice thing about drones is that they perform multiple missions at once,” says Sean Triplett, team lead of fire and aviation management, tools and technology at the Forest Service.
Experts predict we’ll see swarms of drones dropping water and fire retardant on burning buildings and forests in the near future.
The UAS is especially helpful during wildfires because it can track fires, get ahead of wind currents and warn firefighters of wind shifts in real time. The U.S. Forest Service also uses long endurance, solar-powered drones that can fly for up to 30 days at a time to detect early signs of wildfire. Wildfires are no longer seasonal in California – they are a year-long threat, notes Thanh Nguyen, fire captain at the Orange County Fire Authority.
In March, Nguyen’s crew deployed a drone to scope out a huge landslide following torrential rains in San Clemente, CA. Emergency responders used photos and videos from the drone to survey the evacuated area, enabling them to stay clear of ground on the hillside that was still sliding.
Improvements in drone batteries are enabling them to fly for longer with heavier payloads. Experts predict we’ll see swarms of drones dropping water and fire retardant on burning buildings and forests in the near future.
AI to the rescue
The biggest peril for a firefighter is often what they don’t see coming. Flashovers are a leading cause of firefighter deaths, for example. They occur when flammable materials in an enclosed area ignite almost instantaneously. Or dangerous backdrafts can happen when a firefighter opens a window or door; the air rushing in can ignite a fire without warning.
The Fire Fighting Technology Group at the National Institute of Standards and Technology (NIST) is developing tools and systems to predict these potentially lethal events with computer models and artificial intelligence.
Partnering with other institutions, NIST researchers developed the Flashover Prediction Neural Network (FlashNet) after looking at common house layouts and running sets of scenarios through a machine-learning model. In the lab, FlashNet was able to predict a flashover 30 seconds before it happened with 92.1% success. When ready for release, the technology will be bundled with sensors that are already installed in buildings, says Anthony Putorti, leader of the NIST group.
The NIST team also examined data from hundreds of backdrafts as a basis for a machine-learning model to predict them. In testing chambers the model predicted them correctly 70.8% of the time; accuracy increased to 82.4% when measures of backdrafts were taken in more positions at different heights in the chambers. Developers are working on how to integrate the AI into a small handheld device that can probe the air of a room through cracks around a door or through a created opening, Putorti says. This way, the air can be analyzed with the device to alert firefighters of any significant backdraft risk.
Early wildfire detection technologies based on AI are in the works, too. The Forest Service predicts the acreage burned each year during wildfires will more than triple in the next 80 years. By gathering information on historic fires, weather patterns, and topography, says White, AI can help firefighters manage wildfires before they grow out of control and create effective evacuation plans based on population data and fire patterns.
The future is connectivity
We are in our infancy with “smart firefighting,” says Casey Grant, executive director emeritus of the Fire Protection Research Foundation. Grant foresees a new era of cyber-physical systems for firefighters—a massive integration of wireless networks, advanced sensors, 3D simulations, and cloud services. To enhance teamwork, the system will connect all branches of emergency responders—fire, emergency medical services, law enforcement.
FirstNet (First Responder Network Authority) now provides a nationwide high-speed broadband network with 5G capabilities for first responders through a terrestrial cell network. Battling wildfires, however, the Forest Service needed an alternative because they don’t always have access to a power source. In 2022, they contracted with Aerostar for a high altitude balloon (60,000 feet up) that can extend cell phone power and LTE. “It puts a bubble of connectivity over the fire to hook in the internet,” Triplett explains.
A high altitude balloon, 60,000 feet high, can extend cell phone power and LTE, putting a "bubble" of internet connectivity over fires.
Courtesy of USDA Forest Service
Advances in harvesting, processing and delivering data will improve safety and decision-making for firefighters, Grant sums up. Smart systems may eventually calculate fire flow paths and make recommendations about the best ways to navigate specific fire conditions. NIST’s plan to combine FlashNet with sensors is one example.
The biggest challenge is developing firefighting technology that can work across multiple channels—federal, state, local and tribal systems as well as for fire, police and other emergency services— in any location, says Triplett. “When there’s a wildfire, there are no political boundaries,” he says. “All hands are on deck.”