{Expert Insights} Dr. Cheng-Yu Chen (Distinguished Professor, School of Medicine, TMU) on the Role of AI in Solving Dementia Treatment Challenges

This number is projected to exceed 500,000 by 2030. Supported by the NSTC’s National Hub program, Dr. Cheng-Yu Chen and research team have successfully developed DeepBrain-Cognito – the world’s first AI model capable of predicting both brain age and dementia risk.

AI as a “Crystal Ball” for Dementia Risk Forecasting

One of the greatest challenges in dementia treatment is early detection. In many cases, patients are not diagnosed until symptoms such as forgetfulness or memory loss become noticeable, which is often too late for effective intervention. These symptoms are difficult to distinguish from natural aging. As a result, treatment usually begins only after a confirmed diagnosis, when medications may merely slow disease progression rather than reverse it.

To address this issue, Dr. Chen and his team developed an AI-driven tool that acts like a “crystal ball,” enabling the early prediction of dementia risk. The module allows health subjects and patients to prepare in advance, and also allow doctors to formulate timely treatment plans, and provides pharmaceutical companies with data to support the development of more effective drugs, ultimately improving treatment outcomes and reducing both social and healthcare burdens.

Three Core Features of the AI-based Dementia Prediction Module
The award-winning DeepBrain-Cognito module integrates datasets from dementia research in the United States, United Kingdom, and Japan, along with clinical data from four major Taiwanese medical centers: National Taiwan University (NTU), National Cheng Kung University (NCKU), Taipei Veterans General Hospital, and the TMU Healthcare System.

By incorporating 3D T1-weighted MR scans, age, gender, education level, socioeconomic status, and genetic data in a pipeline of so-called autoencoder and generative AI (GAN), the AI model generates reports that estimate brain age, assess dementia risk within the next two years, and evaluate signs of brain degeneration.

Dr. Chen highlights the three major features of the module that serve as a precise clinical decision support tool:

1. Brain Age Estimation

For individuals aged 50 and older, the AI system estimates brain age and compares it to chronological age. A brain that appears older than the person’s actual age may indicate accelerated neurodegeneration and a higher risk of developing dementia.

2. Two-Year Dementia Risk Forecast

For patients with Mild Cognitive Impairment (MCI), the model can estimate the risk of developing dementia within two years. Research studies shows that patients with severe amnesia type MCI have up to 30% chance of progressing to dementia within a year. Making accurate early risk assessment critical for early intervention.

3. Dementia Type Differentiation

The AI module identifies patterns of brain atrophy to help differentiate among the four major types of neurodegenerative dementia: Alzheimer’s disease, frontotemporal dementia (FTD), dementia with Lewy bodies (DLB), and vascular dementia (VaD). Each type involves distinct patterns of brain shrinkage. With AI, clinicians can recognize these patterns earlier and make more informed diagnostic decisions. For instance, Hollywood movie actor Bruce Willis retired in 2022 due to aphasia, but was not until 2023 that he was officially diagnosed with FTD —illustrating the difficulty of early and accurate dementia classification.

Currently undergoing regulatory validation, the DeepBrain-Cognito module  holds great promise in personalized cognitive care. Dr. Chen envisions the model being widely adopted in the near future, not only to help middle-aged and elderly in their brain health early but to reduce the burden on caregivers. Furthermore, the tool could support hospitals and pharmaceutical companies in discovering more effective dementia treatments on the prospects of preventive measures and significantly lowering the cost of long-term healthcare.