Utilizing routinely acquired clinical neuroimaging and electronic health record data to advance precision medicine in dementia care.
Authors
Affiliations (5)
Affiliations (5)
- Richman Family Precision Medicine Center of Excellence in Alzheimer's Disease, Johns Hopkins University School of Medicine, Baltimore, MD, USA.
- Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, MD, USA.
- Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD, USA.
- Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine and Johns Hopkins Bayview, Baltimore, MD, USA.
- Department of Computer Science, Whiting School of Engineering, Johns Hopkins University, Baltimore, MD, USA.
Abstract
Alzheimer's disease (AD) exhibits significant clinical variability in symptom onset, progression rates, neuropsychiatric symptoms and treatment responses. This variability reflects a range of underlying biological, genetic and environmental factors. This review summarizes recent advances in leveraging real-world electronic health records (EHRs) and clinical brain MRI to enhance precision medicine in dementia care. Traditional MRI research has identified consistent subtypes of atrophy associated with AD. However, these models often struggle to apply to routine clinical imaging, which can vary widely in contrast, resolution and acquisition protocols. Recent technological developments now allow for reliable measurement of gray matter, white matter, brainstem and cerebellar structures from routine clinical scans, effectively overcoming long-standing limitations of conventional neuroimaging methods. Additionally, efforts in EHR analysis, including the use of natural language processing on unstructured clinical notes, have enabled large-scale extraction of cognitive scores, neuropsychiatric symptoms and treatment responses. By integrating structured EHR data with detailed imaging markers, researchers have enabled predictive modeling of cognitive decline and treatment responses, though generalizability across settings remains a challenge. Federated learning frameworks offer a privacy-preserving approach to collaboratively develop models across multiple institutions. Together, these strategies outline a practical, data-driven approach to individualized diagnosis, prognosis and treatment planning for dementia.