نهمین کنگره نقشه برداری مغز

نمایش مقاله

Title:

AI-Powered Early Detection of Neurological Disorders: A Multimodal Approach

Subject:

Artificial Intelligence and Brain Mapping

Presenter Author:
Name:	Mahmood HafshejaniNogoorani

Author:
Name 1:	Mahmood HafshejaniNogoorani - Corresponding - Presenter
Affiliation 1:	PHD Student in Computer Engineering-Software Engineering, Islamic Azad University ,Qazvin Branch, Qazvin, Iran

Name 2:	Fatemeh Nouri Norouzi
Affiliation 2:	PHD in Sports Psychology, University of Tehran, Tehran, Iran

Name 3:
Affiliation 3:

Name 4:
Affiliation 4:

Name 5:
Affiliation 5:

Name 6:
Affiliation 6:

Name 7:
Affiliation 7:

Name 8:
Affiliation 8:

Name 9:
Affiliation 9:

Name 10:
Affiliation 10:

Abstract

Original Research

introduction :

Neurological disorders affect over 1 billion people worldwide, with Alzheimer's disease, Parkinson's disease, and epilepsy representing significant global health burdens. Early detection is crucial for effective intervention, yet current diagnostic methods often identify pathologies only after irreversible neural damage has occurred. The integration of artificial intelligence with multimodal neurodata offers unprecedented opportunities for pre-symptomatic diagnosis. This research addresses three key gaps: (1) limited sensitivity of unimodal approaches, (2) inability to detect subclinical pathological changes, and (3) inter-individual variability in disease progression. By leveraging recent advances in deep learning and multi-source data fusion, we develop a comprehensive framework for early neurological disorder detection with potential clinical translation

Methods:

Data Collection and Preprocessing · Cohorts: 2,500 subjects from ADNI, PPMI, and Temple University Hospital datasets (2015-2023) · Modalities: · Structural MRI (3T) with cortical thickness measurement · Resting-state fMRI for functional connectivity · High-density EEG (256-channel) · Clinical biomarkers (CSF tau, α-synuclein levels) · Preprocessing: FSL for MRI normalization, EEGLAB for signal filtering, ComBat for harmonization AI Architecture 1. Alzheimer's Detection: 3D DenseNet with attention mechanisms 2. Parkinson's Classification: Hybrid CNN-BiLSTM with feature fusion 3. Epilepsy Prediction: Temporal Convolutional Network (TCN) with self-supervised learning Validation Framework · 5-fold cross-validation with stratified sampling · External validation on independent cohort (n=300) · Comparison with radiologists' assessments · Statistical analysis: ROC curves, precision-recall, Cohen's kappa

Results:

Table 1. Performance Comparison of AI Models Disorder Model Accuracy Sensitivity Specificity AUC Alzheimer's 3D DenseNet 96.3% 95.1% 97.2% 0.98 Parkinson's CNN-BiLSTM 94.7% 93.8% 95.4% 0.96 Epilepsy TCN 92.8% 91.5% 93.9% 0.95 Key findings: 1. Alzheimer's model detected hippocampal atrophy 5 years pre-diagnosis (p=0.003) 2. Parkinson's model identified vocal changes 3 years before motor symptoms 3. Epilepsy predictor achieved 30-minute seizure forecasting window 4. Multimodal integration improved accuracy by 18.7% versus single-modality approaches

Conclusion:

This study demonstrates AI's transformative potential in early neurological disorder detection. The proposed multimodal framework significantly outperforms conventional diagnostic methods, enabling timely interventions that could modify disease trajectories. Future work should focus on: (1) validating in diverse populations, (2) integrating wearable sensor data, and (3) addressing ethical considerations in predictive analytics. AI-driven early detection paradigms represent a paradigm shift toward preventive neurology.

Keywords:

Early Diagnosis, Neurological Disorders, Artificial Intelligence, Deep Learning, Multimodal Data Integration