fairfolio
Saved Projects

revoAD: Deep Learning for Alzheimer’s & Parkinson’s Diagnosis

AJAS · 2024 Biomedical and Health Sciences (inferred)

Benjamin Lu · North Carolina Academy of Science

Overview

Alzheimer’s disease (AD) and Parkinson's disease (PD), progressive neurodegenerative conditions, are characterized by neural apoptosis from amyloid beta and α-synuclein oligomeric proliferation overwhelming neural homeostasis. As prevalence rises exponentially, early detection and intervention grow imperative, though systemic obstacles obstruct access. Investigated is advanced machine learning leveraging subtle speech and handwriting (n=18,942) changes to enable affordable screening. Multilayer long short-term memory networks and 3D convolutional neural networks aimed to sensitively discern prodromal indications of dysfunction. Repeated cross-validation, class weights, and tissue-inspired topologies optimized sensitivity across cohort heterogeneity (7 ethnicities; 4 income strata). Terminal validation achieved 96.2% diagnostic precision distinguishing pathology from normal cognition. Rapid mobile application implementation, revoAD, delivers instantaneous screening using embedded AI discerning subtle writing/vocal dysfunction from patient samples (n=50) with 97.6% accuracy, 10-fold efficiency gains over conventional diagnosis. Initial results validate multimodal deep learning’s potential to enable scalable detection, catalyzing a shift toward presymptomatic precision medicine for unrelenting neurocognitive disorders. Expanding access for marginalized groups is urged, as contemporary systemic barriers obstruct status quo diagnosis. Immediate next phases encompass pragmatic clinical trials and rapid product deployment to mitigate alarming prevalence increases. While promising for earlier intervention, continued methodological vetting across expanded heterogeneous populations remains vital to substantiate external validity claims and qualify appropriate clinical integration benefiting multiplied constituencies through enhanced access.

Competition history

  • AJAS 2024 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

Source: AAAS Annual Meeting (Confex) / American Junior Academy of Science

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google