Researchers at Florida Atlantic University have developed an advanced deep learning model that utilizes electroencephalography (EEG) to differentiate between types of dementia, specifically Alzheimer’s Disease (AD) and Frontotemporal Dementia (FTD). This innovative approach provides a promising alternative to traditional diagnostic methods, which can be costly and time-consuming.
Dementia encompasses a range of disorders that progressively impair cognitive functions such as memory, thinking, and daily activities. By 2025, an estimated 7.2 million Americans aged 65 and older are projected to be living with AD, while FTD, the second most common cause of early-onset dementia, typically affects individuals between the ages of 40 and 60. Distinguishing between these two forms is essential for effective treatment and care, as their symptoms often overlap.
Revolutionizing Diagnosis with EEG and AI
Current diagnostic practices for AD often include MRI and PET scans, which, while effective, require specialized equipment and considerable resources. EEG provides a more accessible, non-invasive method for measuring brain activity through sensors that capture various frequency bands. Despite its affordability, the interpretation of EEG signals has been challenging due to noise and individual variability.
Researchers have tackled these issues by creating a model that enhances the accuracy of EEG analysis by focusing on both frequency and time-based brain activity patterns. The findings, published in the journal Biomedical Signal Processing and Control, highlight that slow delta brain waves serve as significant biomarkers for both AD and FTD, primarily observed in the frontal and central brain regions. The model achieved an impressive accuracy rate of over 90% in distinguishing patients with dementia from cognitively normal individuals.
The study also revealed that brain activity disruption in AD is more extensive, affecting additional areas and frequency bands, which contributes to its easier detection compared to FTD. The researchers noted that while both conditions share similar symptoms, their neural patterns differ significantly.
Advancements in Machine Learning Techniques
Utilizing feature selection techniques, the researchers significantly improved the model’s specificity, enhancing its ability to correctly identify healthy individuals from 26% to 65%. Their two-stage design, which first detects healthy subjects before distinguishing between AD and FTD, achieved an accuracy of 84%, positioning it among the leading EEG-based diagnostic methods to date.
The model employs a combination of convolutional neural networks and attention-based long short-term memory (LSTM) networks, which not only identify the type of dementia but also assess its severity. This dual capability allows for a more comprehensive understanding of each patient’s condition. The application of Grad-CAM (Gradient-weighted Class Activation Mapping) helps visualize which brain signals influenced the model’s decisions, providing valuable insights for clinicians.
Tuan Vo, the study’s first author and a doctoral student at the FAU Department of Electrical Engineering and Computer Science, stated, “What makes our study novel is how we used deep learning to extract both spatial and temporal information from EEG signals. By doing this, we can detect subtle brainwave patterns linked to Alzheimer’s and frontotemporal dementia that would otherwise go unnoticed.”
The implications of these findings extend beyond simply identifying dementia types; they also offer a new perspective on how brain activity evolves across various regions and frequencies. This understanding can aid clinicians in making more informed diagnoses and treatment decisions.
The research indicates that AD is generally more severe, affecting a wider array of brain areas and resulting in lower cognitive scores, while FTD’s impact is more localized. “Our findings show that Alzheimer’s disease disrupts brain activity more broadly, especially in the frontal, parietal, and temporal regions, while frontotemporal dementia mainly affects the frontal and central areas,” explained co-author Hanqi Zhuang, Ph.D., associate dean and professor at FAU.
The potential of this deep learning model to streamline dementia diagnostics carries significant promise for improving patient care. With millions affected by dementia globally, these advancements could facilitate earlier detection and more personalized treatment strategies.
Stella Batalama, Ph.D., dean of the College of Engineering and Computer Science, remarked, “This work demonstrates how merging engineering, AI, and neuroscience can transform how we confront major health challenges. Breakthroughs like this open the door to earlier detection, more personalized care, and interventions that can truly improve lives.”
The study’s co-authors include Ali K. Ibrahim, Ph.D., an assistant professor of teaching, and Chiron Bang, a doctoral student, both from the FAU Department of Electrical Engineering and Computer Science.
As the medical community continues to explore the capabilities of AI in healthcare, this research sets the stage for more efficient and accurate dementia diagnosis, ultimately benefiting countless individuals and their families.
For further reading, the detailed study is available in the journal Biomedical Signal Processing and Control, with a publication date of March 2026.
