Explainable AI for earlier detection of brain disease
I study how subtle patterns in cognition, movement, eye behavior, facial dynamics, and brain imaging can reveal Parkinson's and Alzheimer's disease before symptoms become obvious.
My work asks a simple question: can we detect neurodegenerative disease earlier by learning from subtle patterns in behavior, cognition, and the brain?
I combine brain imaging, online cognitive testing, eye tracking, movement analysis, facial dynamics, and other digital biomarkers with interpretable machine-learning methods. The goal is not only to classify disease, but to make the evidence behind a prediction understandable.
Alongside academic research, I have more than a decade of experience in software engineering and applied AI, which helps me turn research ideas into systems that are reliable, transparent, and practical.
Cognitive, motor, imaging, eye-movement, and facial-expression patterns may carry early signs of neurological change before symptoms are clinically obvious.
I build machine-learning systems that are not only accurate, but transparent enough for researchers and clinicians to inspect, challenge, and understand.
The aim is to connect measurable digital biomarkers with disease mechanisms in Parkinson's disease, Alzheimer's disease, and related disorders.
These papers explore one problem from different angles: how to turn subtle biomedical signals into interpretable evidence for earlier detection of neurodegenerative disease.
Reviews how online cognitive tests, eye tracking, movement data, and other digital biomarkers can support AI-based detection of early Parkinson's and Alzheimer's disease.
Uses T1 MRI scans and interpretable machine-learning features based on brain-region volumes and spatial relationships to distinguish healthy controls, prodromal cases, and Parkinson's disease.
Shows how eye-movement tasks can be analyzed with machine learning to better understand Parkinson's and Alzheimer's disease mechanisms.
Shows that response-time dynamics from online MoCA testing can improve machine-learning classification of Parkinson's disease versus healthy controls.
Uses an online Trail Making Test to show how timing-based measures can capture cognitive and motor changes linked to Parkinson's disease severity.
Compares healthy controls and Parkinson's patients with online neuropsychological tasks to identify subtle cognitive and behavioral changes.
Synthesizes machine-learning approaches for Parkinson's disease datasets, including diffusion tensor imaging, eye tracking, and online cognitive testing.
Argues that Alzheimer's prevention requires finding the beginning of neurodegeneration decades before symptoms become clinically visible.
Combines eye-tracking and neuropsychological testing to predict symptom progression in Parkinson's disease.
Compares rough-set models and other data-mining methods for predicting Parkinson's disease progression across patient groups.
Uses diffusion tensor imaging and data mining to estimate symptom development after deep brain stimulation.
Uses MRI-derived spatial biomarkers in a rule-based model for transparent Parkinson's disease severity classification.
Studies early dementia risk with explainable AI, cognitive test results, and genetic markers.
Studies marker-based gait stability with nonlinear dynamics as a possible screening signal for neurodegenerative disease.
Independently, I explore theoretical questions in physics and cosmology using numerical computation to test ideas against measurable structure. This includes work on cosmic time and the expansion of the universe, as well as gamma-ray bursts as cosmological probes.
Tests cosmological scale-factor assumptions with numerical analysis against supernovae, gamma-ray bursts, cosmic chronometers, and CMB-derived observations.
For publications, academic profiles, software work, and collaboration context, use the links below.