IEEE Conference Paper

NEUROVISION-AI: Alzheimer's Disease Detection Using Multimodal Deep Learning

Deep LearningMedical AIHealthcareNeurology
99%
Accuracy

Validation accuracy achieved by our hybrid CNN-RNN model

4-Stage
Classification

Detects Normal, Mild, Moderate, and Severe stages

CNN+RNN
Hybrid Architecture

Combines spatial (MRI) and temporal (cognitive) analysis

0.99 AUC
Performance

Area under the curve for robust disease detection

Early and accurate detection of Alzheimer's Disease (AD) is critical for effective intervention, a task where traditional methods often fall short. This paper presents a multimodal deep learning framework to address this diagnostic challenge.

Problem Statement

Traditional diagnostic protocols often fail to provide definitive diagnosis until the disease has advanced, limiting intervention effectiveness.

Solution

Hybrid CNN-RNN architecture combining spatial MRI features with temporal cognitive patterns for comprehensive AD staging.

Dr. R. Priyadarshini

Principal Investigator

Siddartha Institute of Science and Technology

A Chandhana

Research Scholar

Dept. of CSM, SIST

Vidyala Karthik

Research Scholar

Dept. of CSM, SIST

Avula Hemanth Kumar Reddy

Research Scholar

Dept. of CSM, SIST

Tholeti Keerthana Reddy

Research Scholar

Dept. of CSM, SIST

Kasanna Gari Guru Venkat Sai

Research Scholar

Dept. of CSM, SIST

ModelInput ModalityAccuracyF1-ScoreAUC
CNN (Baseline)MRI Only94%0.920.95
RNN (Baseline)Cognitive Data Only90%0.890.91
Hybrid CNN-RNN (Proposed)MRI + Cognitive99%0.980.99

Key Achievements

  • 99% validation accuracy on four-stage classification
  • 5-9% improvement over single-modality baselines
  • Minimal overfitting with aligned training/validation curves

Clinical Impact

  • Enables early intervention during Mild Cognitive Impairment (MCI) stage
  • Reduces diagnostic delay by detecting subtle patterns
  • Supports personalized treatment planning through staging

Try Our Live Implementation

Experience the power of our research through our interactive assessment tools. Test the same models described in this paper.