Advanced convolutional neural network with attention mechanism for Alzheimer's disease classification using MRI.

Journal: Computers in biology and medicine
PMID:

Abstract

This paper introduces a novel convolutional neural network model with an attention mechanism to advance Alzheimer disease (AD) classification using Magnetic Resonance Imaging (MRI). The model architecture is meticulously crafted to enhance feature extraction and selectively focus on crucial regions within brain images, thereby improving diagnostic accuracy. A unique component, the MRI Segmentation Block (MSB), is introduced to manage the skull stripping task effectively, highlighting the model ability to learn from complex, multilayered information. We have incorporated a detailed experimental evaluation of the MSB, demonstrating its superior performance in cranial debridement tasks compared to existing methods. The experiments involved a range of MRI scans, assessing the MSB's accuracy through metrics like the Dice Coefficient and Jaccard Index against ground truth annotations by expert radiologists. The results substantiate the MSB's effectiveness, setting a new benchmark for precision in medical imaging diagnostics. The proposed method integrates densely connected neural networks with a connection-wise attention model to extract multiscale features from MRI scans. Furthermore, the attention mechanism is fine-tuned to emphasize salient features significantly associated with various stages of Alzheimer's disease, thereby setting a new benchmark for precision in medical imaging diagnostics. Extensive experiments on the Alzheimer Disease Neuroimaging Initiative (ADNI) dataset demonstrate the superiority of our method over traditional and contemporary approaches, with our model achieving high accuracy and computational efficiency. This makes it suitable for clinical applications where resources are limited. This study represents a significant advancement in the diagnostic process for AD, with potential implications for improving patient outcomes in clinical settings.

Authors

  • Shakhnoza Muksimova
    Department of IT Convergence Engineering, Gachon University, Sujeong-Gu, Seongnam-Si 461-701, Republic of Korea.
  • Sabina Umirzakova
    Department of Computer Engineering, Gachon University, Sujeong-gu, Seongnam-si, 461-701, Gyeonggi-do, South Korea.
  • Nargiza Iskhakova
    Department of Systematic and Practical Programming, Tashkent University of Information Technologies Named After Muhammad Al-Khwarizmi, Tashkent, 100200, Uzbekistan.
  • Aziz Khaitov
    Republican Scientific and Methodological Center for the Development of Education of the Republic of Uzbekistan, Uzbekistan.
  • Young Im Cho
    Department of Computer Engineering, Gachon University, Sujeong-Gu, Seongnam-si 461701, Republic of Korea.