“The application process for the AITC program was characterized by its thoughtful approach, allowing applicants to comprehensively articulate the gaps in science/technology and innovative aspects of their research proposals. This thoughtful application process proved to be immensely beneficial not only for securing AITC funding but also for enhancing the quality and clarity of future grant applications, particularly those aimed at larger funding opportunities. Once the grant was awarded, we gained access to a wealth of invaluable resources. The AITC boasts a team of experts with diverse backgrounds and expertise, fostering a transdisciplinary environment that can address a wide range of project-related needs. These experts are well-equipped to provide support across various dimensions of a project, be it clinical, technical, or business/start-up-oriented. We gained significant knowledge around the details to translating a research idea into product.”
– Chintan Pandya,
Grant Year 1 Awardee
Alzheimer’s Disease and Related Dementia
The goal of the Alzheimer’s Disease and Related Dementia Pilot is to use a multidisciplinary, collaborative approach to identify, develop, refine, and disseminate promising technologies that have a high potential to improve the health and well-being of older Americans living with ADRD and/or their caregivers, with an emphasis on technologies that can mitigate current disparities in the access and delivery of dementia care. Pilot leadership’s duties include:
Year 1
An AI algorithm to screen for and detect probable dementia cases in the emergency department.
This project will produce a software product intended to aid in the detection of early mild cognitive impairment. Accomplishing this will assist clinicians on the front lines of early stages of dementia investigation, management, and treatment.
WellSaid.ai’s solution to address the need for early detection of cognitive impairment leverages simple-to-use, inexpensive, in-home consumer technologies like the Amazon Alexa platform, but equipped with scientifically validated, clinical-grade cognitive assessments modeled from gold-standard assessment batteries. This solution allows for the screening, identification, and clinical categorization of cognitive impairment and dementia well before an individual typically seeks medical attention and allows for the large-scale expansion of dementia screening and detection at an earlier stage.
WellSaid.ai’s pilot goal is to develop, test, and validate machine learning model(s) that accurately predict the cognitive status of older adults in their own home. The pilot has three main objectives: 1) Retrospectively analyze our existing datasets to guide model development; 2) Prospectively collect comprehensive cognitive assessment data, including raw audio data from 160 patient-caregiver dyads from our in-home assessment technology; and 3) Develop predictive models using natural language processing and supervised learning algorithms that can identify and predict cognitive impairment.
Care Daily, previously known as People Power, will develop and research an AI caregiver service called “Arti” that receives data from Apple watches and sends useful alerts, recommendations, and encouraging messages to caregivers watching over people with health problems such as Alzheimer’s disease. “Arti” should be able to alert on falls, predict falls before they occur, detect wandering, and alert caregivers of the GPS location of a missing person with Alzheimer’s, monitor their sleep quality trends and offer recommendations for improving sleep quality, monitor their fitness trends, recommend ways to improve caregiver health, and provide other senior care microservices.
The Sovrinti in-home system combines device utilization and real-time location information to monitor seniors and their caregivers in the home and assisted living environments. The system’s rich sensor data allows for the continuous identification of activities of daily living and and instrumental activities of daily living.
For this project, Sovrinti will utilize AI techniques with de-identified data from an ongoing National Institute on Aging Phase II study to preemptively identify adverse effects in seniors with cognitive challenges. This project will utilize the hundreds of months of available sensor data and associated monthly health status reports to develop and validate the use of the Sovrinti system to predict rising risks of specific adverse events associated with senior care. These adverse events include urinary tract infections, COVID-19 events, falls, hospitalizations, and more.
This approach combines an anomaly detection component with a trained embedding for each home or care recipient to create a risk prediction model; the risk prediction model will be used to establish alerts for caregiver intervention. In addition to model development, validation will establish confidence parameters for alerts and quantify confidence level as a function of days prior to an adverse event, with a goal of high confidence alerts occurring 2–5 days in advance of an adverse event.
This project will apply a deep learning framework to develop a prediction model that can be commercialized as a service offering personalized cognitive impairment risk prediction. Toward precision medicine and care, a deep-learning-based brain MRI analysis will be applied to provide an objective measure with which to quantify the risk for the development of cognitive impairment and identify modifiable factors that could accelerate or slow the pathological processes for each individual. This study will provide clues for disease prevention and therapeutic interventions.
Sequoia is developing a wearable technology that makes acoustic stimulation accessible to older adults in the comfort of their own homes. By adapting this technology to the unique needs of older adults and packaging the complex sensors, circuitry, and algorithms required for acoustic stimulation into a comfortable and easy-to-use headband, we are translating this laboratory technique into an accessible therapy. This developmental work will entail an exploratory usability study to thoroughly understand essential design features. The resulting information will be used to focus our prototyping efforts as we continue our algorithm development. To validate the results of this work, we propose a small clinical feasibility evaluation of our prototype device.
Lucidity is a mobile technology platform for home use that reports cognitive changes in an individual with Alzheimer’s disease to their caregiver. The system will also provide a framework for quantifying cognitive fluctuations and lucid intervals in individuals with Alzheimer’s.
An AI-based facial expression analysis software will be integrated with an autonomous navigating robot to visit nursing home residents with advanced AD/ADRD and detect their agitation. This can open opportunities to deliver non-pharmacologic interventions and to improve quality of care.
This project will apply AI analytics and electronic health records to develop/disseminate prediction models to identify persons with ADRD who would benefit from palliative care. The tools developed will support resource planning, quality and equity assessment within health delivery systems, health plans, and communities.
This is a virtual reality world for people with dementia that delivers therapies and activities of daily living engagement. This project will build with a AI-based dialogue management system to make it more interactive and to allow caregivers more downtime.
Year 2
To advance the use of digital technologies in diagnosing early stages of Alzheimer’s disease and mild cognitive impairment, we propose a research project using voice recordings, neuropsychological tests, neuroimaging, and neuropathology data. By developing interpretable deep learning models, we aim to identify cognitive abnormalities in speech that align with brain atrophy patterns and neuropathology grades. The project’s success will provide a validation framework for researchers and expand the application of voice technologies, making them accessible to diverse populations.
EZ-Aware is an easily accessible digital health platform that uses active and passive monitoring of cognitive and behavioral functions in older adults’ everyday environments. Current cognitive screening tools to detect mild cognitive impairment (MCI) are time-consuming, are not used routinely, do not assess functional abilities in daily life, and are not sensitive to the earliest signs of neurodegenerative disease. To overcome these current barriers and limitations of cognitive screening in older adults, the goal of this pilot project is to design a functional version of EZ-Aware that will be piloted for feasibility among cognitively healthy individuals and those with MCI. All study data will be used to form a digital twin model, or a digital representation of each individual participant in the study.
Emerging data suggest COVID-19 increases the risk for or potentiates the onset of Alzheimer’s disease in older adults. Those living in underserved urban and rural communities face distinct challenges in accessing dementia care. There is an urgent need to identify novel methods to evaluate the impact of COVID-19 on the cognitive health and dementia risk of older, at-risk adults, and to do so in a manner that facilitates access for those in low-resource settings.
The DANATM Cognitive Assessment Tool is an FDA-cleared mobile application developed and validated to detect subtle cognitive dysfunction. The study team has successfully partnered with DANA developers to track cognitive trajectories following COVID-19. In this prospective cohort study, older adults will complete the DANA battery on multiple occasions following the contraction of COVID-19. By applying machine learning techniques to the in-home, app-based measurement of cognition, this project will develop a more accurate, reliable, and accessible method of detecting and predicting cognitive decline and Alzheimer’s risk following a COVID-19 diagnosis.
The EyeControl platform is designed to support patients in the ICU to improve communication between patients, families, and medical teams and to support the management of delirium. This project will evaluate the efficacy of the EyeControl platform specifically in older adults admitted to the Johns Hopkins ICU and pilot new device features for delirium management.
This project will develop methods for accurate and non-intrusive assessment of neuropsychiatric symptoms (NPS) at home over time, document the sequence of occurrences of NPS, predict adverse events, and provide interpretable data for accurate diagnosis. We will develop a clinician dashboard summarizing patients’ trajectories of NPS-related behaviors and correlate these measures with the onset of dementia. The goal of this project is to build the infrastructure for a large-scale clinical trial.
Year 3
Alzheimer’s Disease and related dementias (ADRD) afflict over 6 million Americans, with prevalence projected to triple by 2050 as the population ages. The demand both for skilled caregivers and for upskilling family caregivers is at an all-time high. The mission of YayaGuide by CareYaya is to expand access to quality care by empowering caregivers of those living with dementia.
Caring for someone with dementia poses myriad challenges for caregivers. However, current training options are often inaccessible, costly, and fail to adapt to individual learning needs. Studies show caregiver education programs improve competencies in managing behaviors, communication, activities of daily living. However, accessibility barriers persist. In-person workshops are costly and inconvenient. Online modules see low adoption due to lack of personalization. Training on optimal strategies can greatly empower both frontline staff and family caregivers.
YayaGuide addresses this critical unmet need by harnessing the power of AI and micro-learning to provide personalized, engaging training at caregivers’ fingertips – essentially, creating the “Duolingo of dementia care training”. Coupled with conversational agents and machine learning for personalization, micro-learning can make training highly accessible for time-strapped caregivers. It also promotes knowledge retention versus traditional learning formats.
For CareYaya’s rapidly growing caregiver workforce of over 10,000 college students across the country engaged in dementia care, YayaGuide empowers learning suitable for the GenZ population in line with existing popular apps in that demographic like Duolingo for language learning.
The introduction of AI and micro-learning into caregiver training marks a significant innovation. Leveraging advanced natural language processing and machine learning algorithms, YayaGuide will guide caregivers through interactive training modules, tailoring content to their unique competencies and learning pace. From communication techniques to assisting with daily activities, this comprehensive curriculum aims to equip caregivers with the skills and confidence to provide high-quality, person-centered care.
Together, we have the opportunity to transform the landscape of caregiver training, improve quality of life for individuals with ADRD, and set a new standard for technology-enabled, compassionate care.
In this project, we will use new Alzheimer’s Disease (AD) multimodal digital biomarkers developed with artificial intelligence. We will utilize data that are obtained from eye movement, voice and speech, handwriting, and gait & posture from people in the early stages of AD while they are multitasking and will compare them with other people to observe if the proposed biomarkers indicative of the presence of AD. The use of multimodal digital biomarkers is quite unique, as to our knowledge, no study reports results using all of these diverse and complementary physiological measures at once. The main goal of this study is to compare these multimodal biomarkers obtained in two different manners: with High Precision Equipment (baseline) and Portable Integrated Equipment proposed in this project. Whereas the first one has already been employed in our previous analyses and can be accurate and precise, the second would be faster, more cost-effective, and portable, allowing its integration into clinical practice more efficiently and reaching more locations.
Dementia poses complex communication challenges that affect diagnosis, treatment, and health. Communication increasingly occurs outside scheduled healthcare visits, with care partners relaying their concerns to busy healthcare clinicians. At the same time, care partners often go unidentified by health care systems. Recent policies prioritize routine and systematic documentation of care partner identity as a well-established gap in dementia care. In response to this need, we will develop and test “Smart Rep”, Smart Automation to REspond to Patient portal messages. Smart Rep will apply natural language processing (NLP) to process electronic messages regarding patients with dementia, the majority of which come from care partners. Smart Rep will facilitate identification of care partners, and detect and respond to care partner concerns. We focus on concerns of mentation, mobility, and medications, which draw from the 4 M’s of an Age-Friendly Health System. Smart Rep would ensure care partners receive timely resources needed to manage care, and reduce healthcare clinicians’ workload. We envision Smart Rep as a feature that any practice could “plug-in” to their system for receiving electronic patient messages. This pilot study sets the stage for proactive and coordinated support of care partners. We have an existing network to move Smart Rep from prototype development to real-world implementation and evaluation following the pilot period. PI’s – Kelly Gleaseon, PhD, RN, Johns Hopkins School of Nursing
Many severe adverse health outcomes can be avoided through adherence to medications like beta
blockers. While common wearable devices detect biophysical signals affected by beta blockers and other
medications, current analytical techniques are not suitable for tracking adherence. This project will
develop a set of artificial intelligence approaches that can track individual-specific adherence to beta
blocker medications in older adults, and can be generalized to a wide variety of medical problems.
Early-stage Alzheimer’s Disease and Alzheimer’s Disease Related Dementias (AD/ADRDs) are currently characterized using traditional neuropsychological evaluations, which dichotomize cognitive impairment in two diagnostic categories: “mild neurocognitive disorder” (major ND) and “major neurocognitive disorder” (minor ND). This project aims to leverage digital tools, including cognitive and speech artificial intelligence technologies (cognitive and speech AI), as well as actigraphy monitoring, in addition to “gold standard’ neuropsychological evaluations, to better stratify and characterize degrees of cognitive impairment. Our prior and ongoing research suggests that we can better assess degrees of cognitive impairment by combining traditional neuropsychological evaluations with digital tools. In this project, we will: 1) Assess the degree of cognitive impairment (none, mild ND, major ND/AD/ADRD) using gold standard neuropsychological assessments in patients living in an assisted living facility; 2) Conduct repeated cognitive ecological momentary assessment (cognitive EMA) and speech AI (patented technology with IBM) among these patients; 3) Utilize actigraphy monitors to track steps, resting heart rate, and heart rate variability. We will repeat the assessments over time for each participant to determine how fluctuations in cognition, speech, and activity dynamically interact in an individuals’ everyday environment (i.e., outside the clinic setting). We will also characterize differences between patients in mild versus major ND groups, to determine which technologies best discriminate between diagnostic categories compared to traditional measures. By developing psychometrically robust and ecologically valid batteries, we can contribute to the dissemination of accessible, remote batteries that are able to rapidly characterize degrees of cognitive impairment at scale.
Healthy Aging
The goal of the Healthy Aging Pilot is to utilize a multidisciplinary collaborative approach to identify, develop, refine, and disseminate promising technologies that have high potential to improve the health and well-being of older Americans and/or their caregivers, with an emphasis on technologies that can mitigate current disparities in the access and delivery of health care in rural and urban areas across the US.
Year 2
Physical function, or the ability to perform daily activities, is important to screen among geriatric patients, since it can determine one’s independence and is linked to frailty, disability, and death. However, time and space constraints, lack of trained staff, and limited reimbursement mechanisms make it difficult for providers to screen for physical function. Additionally, patients are unable to accurately and consistently measure their own physical function; therefore, their treatment plans will be affected.
However, wearable devices can collect mobility, balance, and gait data and can incorporate software to analyze physical function accurately without the need for supervision. We developed a Geriatric Function Assessment (GFA) system that effectively evaluates physical function among older adults. The GFA system will be incorporated into a wearable device that will perform motion and visual monitoring as patients navigate among common clinical areas during clinic visits. We aim to optimize the GFA system’s design and calibration, generate deep neural network models, and perform a feasibility design among 21 older adults with multiple chronic conditions in a clinic-based setting to maximize the user experience. We envision that the development of this system will provide a solution to both clinicians and geriatric patients by producing accurate physical function data in an efficient manner.
Falls are associated with immobility, mortality, and decreased independence, leading to social isolation and reduced quality of life. One of four individuals over 65 years old sustains a fall each year. While predictors of falls have been identified—including fall history, age, and cognition deficits—these data fail to provide insight as to how to treat or track fall risk. As the aging population increases, the economic burden for hospitalization and supervision for those at risk for falls rises. A critical step toward improving our ability to treat fall risk is to identify technology to efficiently collect clinically targetable and fall-relevant outcomes both in the clinic and the home.
To achieve this goal, a novel, portable, and inexpensive shoe insert has been developed to measure standing and dynamic balance outcomes. The smart insole provides balance and gait parameters in real time, providing the opportunity to collect highly quantitative fall predictors in the home or the clinic and thereby facilitating the design of specific interventions to prevent future falls. The purpose of this project is to validate this insole technology and relate it to reported falls. Walking assessments will be conducted on 50 individuals with and without these insoles and on a subsample of individuals wearing insoles in the community. Focus groups will provide input from health professionals and insole users. Collectively, this project will allow us to 1) validate insole outcomes to in-lab gold standards, 2) validate the clinical utility of outcomes by relating them to falls, and 3) validate the clinical usability of these outcomes by stakeholders.
This pilot will capture patient balance data using a force plate at the point of care in a geriatric outpatient clinic to provide insight into what medical conditions and social determinants of health influence balance and the underlying risk of injuries for older adults. A large registry of balance data is key to using machine learning to aid in clinical assessment though a real-time dashboard. This will allow for the identification of appropriate treatment options, as well as the implementation of fall prevention strategies to create solutions and shift the focus from treatment to prevention.
This project will leverage our lightweight and affordable exosuits in concert with learning-based personalized control to promote physical activities in everyday settings and explore the potential to ultimately decrease the risk of cognitive decline which leads to dementia in older adults.
Year 3
This study aims to pilot AI-based analytics and optimization techniques to improve the efficiency of scheduling caregiver in-home visits using a new neighborhood model of staffing. The project will use historical data to optimize the sequencing of visits and tasks within a geographic zone on a given day.
In the United States there are 37 million people with hearing loss but only 8 million wear hearing aids. Use of
hearing aids by older adults slows the decline of thinking and memory abilities by 48%, making their low rate of
adoption a big public health concern. The most common reason for not wearing hearing aids is difficulty
participating in conversations in noisy social environments: family reunions, birthday parties, weddings, etc.
Imagine a proud mom at her son’s wedding frustrated by loud noise instead of enjoying and reconnecting with
her family & friends. Or an uncle struggling to catch up with their niece at the family thanksgiving party.
AudioFocus is reimagining the hearing aid from the ground-up to solve this noise problem using AI. Their
patented proximity-based speech enhancement algorithm enhances conversations in front of patients while
suppressing noise from other conversations around them. To validate the benefits of their technology over
existing AI hearing aids, they are running pilot studies with Dr. Reed from John Hopkins University, Dr.
Fitzgerald from Stanford, and Dr. Hu from the University of the Pacific. AudioFocus’ team includes hearing aid
science & engineering experts Dr. Shariq Mobin (PhD UC Berkeley, Google Brain) & Dr. Reza Kassayan
(EarLens, Sonitus Medical).
Dr. Shariq Mobin, PhD, Chief Scientific Officer of AudioFocus
Professor Jiong “Joe” Hu, PhD, AuD, Vice Chair of Audiology at University of the Pacific
Falls are a major concern for older adults in the United States, leading to injuries, decreased mobility, and hefty medical costs. With the aging population increasing, effective interventions are crucial. While physical therapy can help prevent falls, issues like limited access and engagement hinder its impact, especially in rural areas. Brightway Physical Therapy (PT) offers a promising solution by using innovative technology to monitor at-home exercises via a smartphone application and computer vision.
In this project we will enhance Brightway PT’s capabilities to track exercises accurately and establish automated assessments based on standard fall risk measurements. Importantly, we will engage with stakeholders, including physical therapists and older adults in rural areas, in designing the exercises and assessments. By gathering feedback and co-designing features like multimedia content and gamification, we aim to make the platform more appealing and user-friendly. Finally, we will test the initial design with older adults who have a history of falls to evaluate its usability and gather feedback for improvement. This will involve observing participants as they complete tasks and conducting interviews to assess acceptability and identify barriers.
This pilot project seeks to tailor and validate Brightway PT for fall prevention in older adults, focusing on real-world usability and flexibility. Ultimately, our goal is to develop a functional platform for remote fall risk assessment and exercise delivery, paving the way for future evaluations on adherence, effectiveness, and impact on falls through feasibility testing and clinical trials.
