Sign in Subscribe
Computer science

Breakthrough in Dementia Detection: AI-Powered Ensemble Model Shows Promising Accuracy

Strategic Insights

— 1 min read

Researchers at the Department of Computer Science and Engineering have made significant strides in developing an early detection system for dementia, a debilitating condition that impairs cognitive abilities and causes memory loss. The AI-powered ensemble model incorporates machine learning and deep learning frameworks, leveraging publicly available datasets to identify dementia stages with high accuracy. This breakthrough has the potential to support clinical decisions and improve patient outcomes.

Key Takeaways:

  • The AI-powered ensemble model utilizes a comprehensive ensemble framework that combines machine learning (ML) and deep learning (DL) algorithms to classify dementia stages with high accuracy.
  • The model employs data preprocessing techniques, including handling missing values, normalization, and encoding, to prepare the dataset for analysis.
  • A total of seven features were selected using F-value and p-value analysis, and the dataset was divided into training (70%) and testing (30%) portions.
  • Four DL models (LSTM, CNNs, MLP, and ANNs) and 12 ML models (LR, RF, SVM, and others) were trained and evaluated using hyperparameter tuning and ensemble voting techniques.
  • The proposed model showcases a promising accuracy of 97.32% in early diagnosis and categorization of dementia, demonstrating its efficacy in clinical decision-making.
  • The model's transparency and interpretability are ensured through the application of SHAP and LIME techniques in ANN and LR.
  • A web-based solution was created to diagnose dementia in real-time, providing a valuable tool for healthcare professionals.

Statistics:

  • The AI-powered ensemble model achieved an accuracy of 97.32% in early diagnosis and categorization of dementia.
  • The proposed model utilizes a comprehensive ensemble framework that combines ML and DL algorithms.
  • A total of 12 ML models (LR, RF, SVM, and others) were trained and evaluated using hyperparameter tuning and ensemble voting techniques.
  • The dataset prepared for analysis includes handling missing values, normalization, and encoding.
  • A total of seven features were selected using F-value and p-value analysis.

Sources:

  • "Web-Based Early Dementia Detection Using Deep Learning, Ensemble Machine Learning, and Model Explainability Through LIME and SHAP" (IET Software, 2025)
  • Khandaker Mohammad Mohi Uddin, Department of Computer Science and Engineering
  • https://doi-org.sdpl.idm.oclc.org/10.1049/sfw2/5455082

Read more