PowerPoint Presentation

Mini project(19AI701) – Final Review. ALZHEIMER’S DISEASE DETECTION USING DEEP LEARNING Submitted by KIRAN .J(212221240022) NIVETHA .M(212221240034) MEENA .S(212221240028) 2021-2025 Batch TEAM NO:04 DR.ASWINI .J Associate Professor, Department of AIML.

Agenda. 1. INTRODUCTION 2. PROBLEM STATEMENT 3. LITERATURE REVIEW SUMMARY 4. METHODOLOGY/FLOW 5. ALGORITHMS USED 6. IMPLEMENTATION 7. OUTPUT 8. CONCLUSION 9. REFERENCES.

[Audio] Introduction According to a generalized degeneration of the brain, Alzheimer's disease is an irrevocable brain disease leading to progressive mental decline and can strike at any age. Alzheimer's disease is uncertain when it is first developing. The project's goal is to identify people who might have Alzheimer's. Early diagnosis facilitates educating about dementia, reasonable expectation setting up, and future preparation for the person with it and guardians..

[Audio] Problem Statement Alzheimer's disease is a brain disease that gradually decreases thinking, memory, and the capacity to perform even the most basic tasks. Early detection of Alzheimer's disease opens up more clinical trial options for patients and gives caregivers more time to adapt to the functional, psychological, and personality changes that might characterize the disease and changing perspective from a caregiver position. The project aims to enhance the ability of the model to differentiate between people with Alzheimer's disease-dementia and those without the disease at the earlier stage..

[Audio] Summary of LR. Summary of LR.

Summary of LR.

Summary of LR.

[Audio] Methodology. Methodology. 2- Model Training I- Model Building (Cloud or Standalone) Data Repository Pre-processing Image slices 5- Model Evaluation Evaluate the Model Accuracy Deep Learning CNN (Densenet169, Resnet50) 4- Model Testing Test the Model HC Identifying the label of the person 3- Export Trained Model Trained Models Load the Trained Model AD.

[Audio] Purpose and Feature: Clearly outline the purpose of the Alzheimer's disease detection , emphasizing features like early detection, monitoring, and support for patients. This specify how deep learning will be integrated to improve diagnostic accuracy and efficiency. Data Collection: Gather a diverse and extensive dataset of brain imaging scans, including MRI and CT scans, from individuals with and without Alzheimer's disease. Collaborate with medical professionals and institutions to ensure the dataset's authenticity and relevance..

[Audio] Data Preprocessing: Implement preprocessing techniques to clean and standardize the imaging data. Consider techniques like normalization and augmentation to enhance the model's robustness. Model Development: Design a deep learning model suitable for Alzheimer's disease detection, considering architectures like convolutional neural networks (CNNs) for image analysis. Split the dataset into training, validation, and testing sets for model training and evaluation. Model Training : During Model Training,the system utilizes preprocessed MRI scan images to train a CNN and RNN. The training process,possibly by observing the training loss and accuracy..

[Audio] Validation and Testing: To keep from overfitting, validate the predictive model on the validation set and then test the final product on a different testing set to see how well it expands to new data. Continuous Improvement: Establish structures to support ongoing learning and development, takes account of frequent updates based on current data and research. You can develop to recognize Alzheimer's disease with the additional complexity of deep learning integration by using this updated methodology..

[Audio] Architecture Diagram. Architecture Diagram.

[Audio] Algorithms used Using deep learning to detect Alzheimer's disease calls for evaluating sequential data, mainly time-series data from brain imaging scans. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are frequently employed designs for this type of programs. Processing 3D imaging of the brain data, such as MRI examinations, which had been preprocessed by scaling and normalizing for consistency, is the initial stage in using the CNN model. After that, feature extraction is carried out using convolutional layers to identify spatial structure and pattern associated with Alzheimer's disease..

[Audio] Sequential data displaying changes over time is processed by an RNN model. Preprocessing of the input, because frequently depicts successive changes in activity in the brain, includes handling missing data and normalization. Ensemble methods integrating predictions from both CNN and RNN models are investigated for possible better performance in model integration and evaluation. To evaluate the resilience and generalization of the examples, cross-validation is used. Metrics including accuracy, recall, F1 score, accuracy, and the region of the ROC curve are used..

[Audio] Implementation from keras.models import Sequential from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, SimpleRNN, TimeDistributed def build_model(): model = Sequential() model.add(Conv2D(filters=16, kernel_size=(3, 3), strides=(1, 1), activation="relu", kernel_initializer='he_normal', input_shape=(128, 128, 3))) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Conv2D(filters=32, kernel_size=(3, 3), strides=(1, 1), activation="relu", kernel_initializer='he_normal')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Conv2D(filters=128, kernel_size=(3, 3), strides=(1, 1), activation="relu", kernel_initializer='he_normal')) model.add(MaxPooling2D(pool_size=(2, 2))).

[Audio] # Add TimeDistributed layer to make the output 3D model.add(TimeDistributed(Flatten())) # Add Simple RNN layer model.add(SimpleRNN(units=64, activation='relu')) model.add(Dense(128, activation="relu", kernel_initializer='he_normal')) model.add(Dense(64, activation="relu")) model.add(Dense(4, activation="softmax")) model.compile(optimizer='adam’,loss="sparse_categorical_crossentropy", metrics=['accuracy']) model.summary() return model model = build_model().

[Audio] Output. Output.

Output.

[Audio] Predicted Values. Predicted Values. Mild Demented Moderate Demented Non Demented accuracy macro avg weighted avg precision 0.97 0.90 0.89 0.95 0.93 0.92 recall 0.89 1.00 0.98 0.83 0.93 0.92 fl-score 0.93 0.95 0.93 0.89 0.92 0.92 0.91 support 76 9 317 238 640 640 640.

[Audio] Output. Output. aaaaa Predicted: Non Demented Predicted: Moderate Demented aaaaa aaaao Non _ aaaaa.

[Audio] Conclusion The project provides the proposed method utilizes Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) architecture for the early detection of Alzheimer's Disease from brain MRI images . The CNN and RNN model is more suitable for image processing, especially in image classification, and has shown higher accuracy in predicting Alzheimer's Disease affected-brain vs a normal aging brain . The use of advanced neuroimaging techniques, such as MRI, allows for the extraction of shape features and texture from the Hippocampus region, aiding in the detection of Alzheimer's Disease..

[Audio] References Reference: [1] Mendez M F 2012 Early-onset Alzheimer's disease: non amnestic subtypes and type 2 AD Archives of Medical [2] Ballard C, Gauthier S et al 2011 Alzheimer's disease [3] Dayan Peter, Abbott Laurence F et al 2001 Theoretical Neuroscience – Computational and Mathematical Modelling of Neural Systems (MIT press) [4] Hinton Geoffrey E 2011 Machine learning for neuroscience Neural Systems & Circuits.

[Audio] References Reference: [5] Meyera Sebastian and Muellera Karsten 2017 Predicting behavioral variant frontotemporal dementia with pattern classification in multi-center structural MRI data Neuro Image Clinical [6] Ahirwar Anamika 2013 Study of Techniques used for Medical Image Segmentation and Computation of Statistical Test for Region Classification of Brain MRI I.J.Information Technology and Computer Science [7] Bin Othman Mohd Fauzi, Abdullah Noramalina Bt et al 2011 2011 Fourth Int. Conf. on Modeling, Simulation and Applied Optimization (Kuala Lumpur, Malaysia) MRI Brain Classification using Support Vector Machines.