Advancing Preclinical Alzheimer Research Through AI and Behavioral Analysis: Stephanie R. Miller, PhD

WATCH TIME: 2 minutes

"These new approaches allow us to assess the full repertoire of the animals’ spontaneous behavior."

Studies have shown that subtle behavioral changes can signal the earliest stages of Alzheimer disease (AD)—well before a formal diagnosis—but detecting these signs in a measurable, scientific way has long been a challenge, even in animal models. In a new preclinical study, researchers at Gladstone Institutes aimed to potentially address this gap by applying a machine learning tool called VAME (Variational Animal Motion Embedding) to video recordings of mice. This AI-powered platform was designed to identify nuanced patterns of spontaneous behavior that conventional testing methods typically miss.^1,2

Using VAME, the team analyzed 2 different mouse models of AD and reported an increased disorganized behavior as the disease progressed, including irregular activity transitions that may reflect memory and attention deficits. The platform also proved to be possibly useful in evaluating a potential treatment like by blocking the toxic effects of a blood protein known to contribute to AD, in which the researchers were able to prevent abnormal behaviors in the mice. Building on these findings, the team is planning to expand the use of the VAME tool across other neurological disease studies.

Recently, Stephanie R. Miller, PhD, staff scientist at Gladstone Institutes, shared these insights in a plenary session at the 2025 American Academy of Neurology (AAN) Annual Meeting, held April 5-9, in San Diego, California.³ She emphasized how AI-based tools like VAME can improve the detection of subtle behavioral shifts in animal models that more closely resemble human AD. In a conversation with NeurologyLive^® following the session, she explained that traditional behavioral assessments often overlook these nuances, underscoring the need for more comprehensive approaches. She also stressed the importance of cross-disciplinary collaboration to bring these innovations from the lab bench to the bedside.

Click here for more coverage of AAN 2025.

REFERENCES
1. Miller S. AI and Biomarkers for Early Disease Detection. Presented at: 2025 AAN Annual Meeting; April 5-9; San Diego, CA. Hot Topics Plenary Session.
2. Machine Learning Reveals Behaviors Linked with Early Alzheimer’s, Points to New Treatments. News Release. Gladstone Institutes. Published November 26, 2024. Accessed April 6, 2025. https://gladstone.org/news/machine-learning-reveals-behaviors-linked-early-alzheimers-points-new-treatments
3. Miller SR, Luxem K, Lauderdale K, et al. Machine learning reveals prominent spontaneous behavioral changes and treatment efficacy in humanized and transgenic Alzheimer's disease models. Cell Rep. 2024;43(11):114870. doi:10.1016/j.celrep.2024.114870

Editor’s note: Miller has disclosed that she has received personal compensation for serving as a Consultant for Cambrya, LLC. In addition, she noted that she has received research support from Gladstone Institutes and from National Institutes of Health.