alzheimer project content ppt(1)(2)(1).pptx

Aa

1. An Effective Deep Learning Based Proposition for Alzheimer’s Disease Stages Classification from Functional Brain Changes in Magnetic Resonance Images Presented by: R.Meiyarasi (620822405014) GNANAMANI COLLEGE OF TECHNOLOGY DATE: GUIDE NAME Mrs. S.ARULARASI Professor/CSE Department of CSE

2. Abstract AlzheimerNet: Proposed fine-tuned CNN classifier for Alzheimer's disease detection. Disease Stages: SMC, MCI (Early and Late), AD, and NC. Data Source: MRI scan dataset from ADNI database. Data Preparation: CLAHE image enhancement, data augmentation for dataset balance. Dataset: 60,000 images across 6 classes. Conclusion: InceptionV3 shows highest accuracy, potential for accurate Alzheimer's disease classification.

3. Introduction Timely Diagnosis Challenge Narrow Window for Intervention Global Burden Advancements in Medical Imaging Machine Learning Integration Promise of Improved Accuracy Streamlined Patient Care Enhanced Treatment Outcomes Potential for Addressing Diagnostic Challenges Improving Patient Management

4. Existing system Multiple studies have explored the use of convolutional neural networks (CNNs) in classifying Alzheimer's disease (AD) stages using magnetic resonance imaging (MRI) data. Various pre-trained models including SqueezeNet, ResNet18, AlexNet, VGG16, DenseNet, InceptionV3, GoogLeNet, and MobileNet have been utilized for AD classification. One study achieved a validation accuracy of 82.53% using pre-trained SqueezeNet on MRI images from the OASIS database.

5. ADVANTAGES OF PROPOSED SYSTEM Early Detection and Diagnosis Patient Monitoring and Progression Tracking Research and Clinical Trials Personalized Treatment Strategies Public Health Initiatives

6. Methodology

7. Block Diagram

8. Data Acquisition

9. Data Pre-processing

10. Architecture Required VGG16 MobileNetV2 AlexNet ResNet50 InceptionV3

11. VGG16 A deep convolutional neural network architecture. Proposed by the Visual Geometry Group at the University of Oxford. It consists of 16 layers, including convolutional layers with small 3x3 filters

12. MobileNetV2 Lightweight convolutional neural network architecture designed for embedded devices. It is optimized for efficiency and low latency Maintains high accuracy. MobileNetV2 employs depth wise separable convolutions.

13. AlexNet Pioneering deep convolutional neural network architecture. It consists of eight layers, including five convolutional layers and three fully connected layers. Used for dropout regularization to prevent overfitting.

14. ResNet50 Deep convolutional neural network architecture that belongs to the ResNet (Residual Network) family. It consists of 50 layers and incorporates residual connections. Used for image classification and feature extraction tasks.

15. InceptionV3 Computational Efficiency: InceptionV3 employs dimension reduction, regularization, factorized convolutions, and parallelized computations. Enables faster training and inference times, suitable for resource-constrained environments. Improved Performance: Incorporates factorized convolutional layers, label smoothing, and auxiliary classifiers.

16. Results

17. Performance Comparison of All Utilized Models Models Mean Accuracy Mean Precision Mean Recall Mean Specificity F1-score Training accuracy % Training loss% Validation accuracy Validation loss % Test Accuracy % Test loss % VGG16 78.85 79.97 78.85 95.77 79.41 79.72 0.27 78.69 0.34 78.85 0.35 MobileNetV2 86.85 87.40 86.85 97.37 87.13 88.75 0.21 86.77 0.28 86.85 0.29 AlexNet 78.86 79.58 78.86 95.77 79.22 78.91 0.41 78.12 0.30 78.86 0.28 ResNet50 80.48 80.87 80.48 96.10 80.67 81.53 0.25 80.81 0.31 80.48 0.29 InceptionV3 92.32 92.56 92.32 98.46 92.44 91.62 0.20 91.97 0.23 92.32 0.18

18. Conclusion Inception v3: Designed to classify Alzheimer's disease stages using MRI scan data and image enhancement techniques. Model Performance Evaluation: Compared five pre-trained models and one fine-tuned model. Inception v3 achieved notable accuracies: Training (91.62%), Validation (91.97%), Testing (92.32%). Image Processing Impact: Image enhancement techniques improve data quality, enhancing classification model accuracy. Transfer Learning Significance: Transfer learning facilitates efficient utilization of small and large image datasets in computer vision tasks. Future Directions: Proposed hybrid model implementation on ADNI fMRI and PET datasets for refined Alzheimer's disease diagnosis. Focus on improving patient care and outcomes through advanced understanding and detection of Alzheimer's stages.

19. THANK YOU