Brain age prediction in people with Parkinson’s disease using minimally processed images

213 

Abstract Submission 

Franziska Albrecht¹, Anna Inguanzo¹, Caroline Dartora¹, Erika Franzén¹, Eric Westman¹

¹Karolinska Institutet, Stockholm, Sweden

Franziska Albrecht  
Karolinska Institutet
Stockholm, Sweden

Anna Inguanzo  
Karolinska Institutet
Stockholm, Sweden

Caroline Dartora  
Karolinska Institutet
Stockholm, Sweden

Erika Franzén  
Karolinska Institutet
Stockholm, Sweden

Eric Westman  
Karolinska Institutet
Stockholm, Sweden

The emerging field of brain age prediction has garnered significant attention in recent years due to its potential to revolutionize healthcare, personalized medicine, and our understanding of aging. Brain age prediction aims to estimate an individual's brain age based on various factors and measurements, e.g., magnetic resonance imaging (MRI). The early detection of misalignment of brain age with the chronological age may allow for timely intervention and treatment, potentially improving patient outcomes and quality of life. Furthermore, brain age prediction could be used to assess the effectiveness of interventions aimed at improving brain health. This knowledge may pave the way for the development of innovative preventive strategies and therapeutic interventions that target the root causes of advanced brain aging.

Recently, we developed and validated a convolutional neural network for brain age prediction utilizing only minimally processed T1-weighted structural MRI (Dartora et al. 2023). Multi-cohort data of 15289 cognitively healthy participants was included, using only MNI-space registered images. This model predicted brain age in people with Parkinson's disease (PD) in two cohorts: baseline data of 84 people with PD taking part in an intervention (EXPANd trial (Franzén et al. 2019)) and 341 people with de novo PD from the Parkinson's Progression Markers Initiative (PPMI (Marek et al. 2018)). The brain age gap was calculated as predicted brain age - chronological age. Thus, positive values mean an older-looking brain, as the brain looks older than expected, and negative values relate to a younger-looking brain. We aimed to replicate the pattern of the only other PD brain age study that has been published so far (Eickhoff et al. 2021). Spearman correlations between the brain age gap and measures of cognition, disease severity, and other clinical measures were run using RStudio (2022.07.0+548). In exploratory analyses, we corrected the brain age gap for chronological age by dividing the gap by the chronological age.

The EXPANd cohort had a mean age of 70.65 (Standard deviation, SD 5.84) and a predicted brain age of 72.14 (SD 4.98). Predicted brain age correlated significantly with chronological age (r=0.72, p<0.001). The brain age gap was negatively correlated with disease duration, i.e., the older-looking the brain, the longer the disease duration (r=-0.27 p=0.015)(Figure 1). There were no further significant correlations with either the brain age gap or the corrected one.
The PPMI cohort had a mean chronological age of 61.58 (SD 9.59) and a brain age of 67.17 (SD 7.29). Predicted brain age correlated significantly with chronological age (r=0.69, p<0.001). We found no other significant correlations with clinical or other demographic data in our PPMI cohort.

We identified a brain age gap in both cohorts. However, the gap was larger in the de novo PPMI cohort. This might be a bit surprising since one might expect the more severe people with Parkinson's disease (i.e., the EXPANd cohort) to have a larger brain age gap. We can only speculate that medications could reduce the brain age gap and thus lead to the EXPANd cohort having younger-looking brains. Further, we found that only disease duration in the interventional cohort was related to the brain age gap. Nevertheless, our study is an important step toward the assessment of the clinical applicability and usability of brain age prediction in people with Parkinson's disease. Brain age may yield the potential to identify those individuals who might need more intensive treatment. Thus, brain age prediction holds immense promise as a powerful tool in the field of healthcare and individualized treatment. With its ability to assess brain health, guide interventions, and deepen our understanding of aging, it has the potential to revolutionize how we approach brain health and improve patient outcomes.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Aging

Classification and Predictive Modeling ²

Anatomy and Functional Systems

Aging

Degenerative Disease

Machine Learning

Motor

Movement Disorder

MRI

Multivariate

Neurological

STRUCTURAL MRI