Applying deep learning models as a tool for Alzheimer's diagnosis: An analysis using (MRI)
7th International Conference for Physics and Advance Computation Sciences, ICPAS 2025, Baghdad, Irak, 18 - 19 Haziran 2025, cilt.3460, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 3460
- Doi Numarası: 10.1063/5.0328219
- Basıldığı Şehir: Baghdad
- Basıldığı Ülke: Irak
- Anahtar Kelimeler: Alzheimer's illness, Automated diagnosis and neural networks, Classification, Deep Learning Models, MRI
- Ankara Yıldırım Beyazıt Üniversitesi Adresli: Evet
Özet
Our study focuses on Alzheimer's disease (AD), a prevalent condition that impacts a significant proportion of the global population. Timely identification and accurate categorization of AD enable prompt interventions and personalized treatment. We provide an innovative and advanced automated deep learning method for early detection of AD to address this requirement. The OASIS magnetic resonance images (MRI) datasets were used from Kaggle. A total of 25,937 brain (MRI) from Alzheimer's patients and controls were used to build and validate the approach. This varied data set was carefully classified into slightly, non-demented, and very mildly demented categories. The study meticulously assesses the model's accuracy via data partitioning. Four deep learning models - InceptionV3, Mobile Net, DenseNet121 and VGG16 were tested for brain MRI image classification accuracy. The focus is on a multi-class classification task rather than binary classification, which enhances the clinical value of the model. The work includes enhanced preprocessing steps such as skull stripping, intensity normalization, and data augmentation, which helped improve the model's performance. The results showed that the VGG16 model achieved the highest accuracy of 81.32%, and performed exceptionally well in the second category, achieving an F1-score of 96.84%, highlighting the model's effectiveness in accurately distinguishing disease stages. We found that deep learning models can classify disease stages using brain MRI scans. This study contributes to the hunt for reliable AD early detection methods. We conclude that deep learning models have the ability to improve the accuracy and efficiency of diagnosis by learning from a real dataset, thus obtaining rapid treatments.