Time Distributed Classification of Alzheimer’s Disease on MRI Scans

Dündar M. S., Yılmaz B.

European Biotechnology Congress, Ljubljana, Slovenya, 4 - 06 Ekim 2023, cilt.7, ss.28, (Özet Bildiri)

  • Yayın Türü: Bildiri / Özet Bildiri
  • Cilt numarası: 7
  • Doi Numarası: 10.2478/ebtj-2023-0018
  • Basıldığı Şehir: Ljubljana
  • Basıldığı Ülke: Slovenya
  • Sayfa Sayıları: ss.28
  • Erciyes Üniversitesi Adresli: Evet

Özet

This study introduces a novel approach to neuroimaging analysis for diagnosing and differentiating Alzheimer’s Disease (AD), Mild Cognitive Impairment (MCI), and Cognitively Normal (CN) individuals, leveraging advanced machine learning techniques. We employed a pre-trained 3D ResNet 101 Convolutional Neural Network (CNN) algorithm for the initial categorization of MRI scans, focusing on analyzing the entire scan volume to extract spatial features. Complementing this, Long Short-Term Memory (LSTM) networks were integrated to process sequences of 3 or 4 annual MRI scans per patient, thereby enhancing the temporal analysis of neurodegenerative progression.

The combination of 3D CNN and LSTM networks led to a substantial improvement in classification accuracy. Our method achieved a 96.7% accuracy rate in distinguishing AD from CN, and a 10% increase in accuracy for AD-MCI versus MCI-CN differentiation. However, challenges were encountered in the classification of MCI, largely due to the model's pre-training on CN and AD data, and the scarcity of annual MRI data for MCI patients.

The findings of this study underscore the effectiveness of combining CNNs with LSTMs for the time-differentiated analysis of neuroimaging data, offering significant insights into the early detection and progression of Alzheimer's Disease. This approach highlights the critical role of both spatial and temporal factors in neuroimaging, providing a promising direction for future research in medical imaging and the broader field of neurodegenerative disease management.