Deep learning for risk-based stratification of cognitively impaired individuals
Michael F. Romano,
Xiao Zhou,
Akshara R. Balachandra,
Michalina F. Jadick,
Shangran Qiu,
Diya A. Nijhawan,
Prajakta S. Joshi,
Shariq Mohammad,
Peter H. Lee,
Maximilian J. Smith,
Aaron B. Paul,
Asim Z. Mian,
Juan E. Small,
Sang P. Chin,
Rhoda Au,
Vijaya B. Kolachalama
Affiliations
Michael F. Romano
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, USA
Xiao Zhou
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; Department of Computer Science, Boston University, Boston, MA, USA
Akshara R. Balachandra
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA
Michalina F. Jadick
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
Shangran Qiu
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
Diya A. Nijhawan
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
Prajakta S. Joshi
Department of Anatomy and Neurobiology, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; Department of General Dentistry, Boston University School of Dental Medicine, Boston, MA, USA; The Framingham Heart Study, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
Shariq Mohammad
Department of Biostatistics, Boston University School of Public Health, Boston, MA, USA
Peter H. Lee
Department of Radiology, Lahey Hospital & Medical Center, Burlington, MA, USA
Maximilian J. Smith
Department of Radiology, Lahey Hospital & Medical Center, Burlington, MA, USA
Aaron B. Paul
Department of Radiology, Massachusetts General Hospital, Boston, MA, USA
Asim Z. Mian
Department of Radiology, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
Juan E. Small
Department of Radiology, Lahey Hospital & Medical Center, Burlington, MA, USA
Sang P. Chin
Department of Computer Science, Boston University, Boston, MA, USA; Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA; Center of Mathematical Sciences & Applications, Harvard University, Cambridge, MA, USA
Rhoda Au
Department of Anatomy and Neurobiology, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; The Framingham Heart Study, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; Boston University Alzheimer’s Disease Research Center, Boston, MA, USA; Department of Epidemiology, Boston University School of Public Health, Boston, MA, USA; Department of Neurology, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
Vijaya B. Kolachalama
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA; Department of Computer Science, Boston University, Boston, MA, USA; Boston University Alzheimer’s Disease Research Center, Boston, MA, USA; Faculty of Computing & Data Sciences, Boston University, Boston, MA, USA; Corresponding author
Summary: Quantifying the risk of progression to Alzheimer’s disease (AD) could help identify persons who could benefit from early interventions. We used data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI, n = 544, discovery cohort) and the National Alzheimer’s Coordinating Center (NACC, n = 508, validation cohort), subdividing individuals with mild cognitive impairment (MCI) into risk groups based on cerebrospinal fluid amyloid-β levels and identifying differential gray matter patterns. We then created models that fused neural networks with survival analysis, trained using non-parcellated T1-weighted brain MRIs from ADNI data, to predict the trajectories of MCI to AD conversion within the NACC cohort (integrated Brier score: 0.192 [discovery], and 0.108 [validation]). Using modern interpretability techniques, we verified that regions important for model prediction are classically associated with AD. We confirmed AD diagnosis labels using postmortem data. We conclude that our framework provides a strategy for risk-based stratification of individuals with MCI and for identifying regions key for disease prognosis.
