Minor Project-2: Title

Minor Project-2
Title:
Alzheimer MRI Classification and detection using ResNet and
Presented by:
CNN
Roll number - R2142210380
Name-Jatin Duindi
Name-Sanskriti Bhardwaj Mentored By:
Roll number - R2142210837 Dr. Virendra Kadyan
Name-Vansh Kataria
Name-Amil Lal
Roll Number- R2142210096
Content
● Introduction
● Literature Review
● Objectives
● Methodology
● Pert Chart
● References
Introduction
● The quest for early detection of Alzheimer's disease (AD) has prompted extensive research into the
application of classification methods, leveraging various data modalities.
● Neuroimaging studies, such as those by Smith et al.[2], have employed magnetic resonance imaging
(MRI) and positron emission tomography (PET) data for distinguishing AD patients from healthy
controls.
● Concurrently, investigations by Chen et al.[2] have explored the potential of cognitive test scores and
biomarkers in AD detection, indicating a multifaceted approach to understanding the disease.
● Integrative studies, like the work of Wang et al.[5], have delved into the combination of genetic and
omics data, providing a holistic perspective on the pathology.
● The application of diverse machine learning algorithms, from traditional classifiers to deep learning
architectures, has shown promise, with Li et al.[1] comparing their performance and delineating their
respective strengths and limitations.
Technical Concepts (Algorithms) used
Statistical Analysis Steps:
Exploratory Data Analysis (EDA):

Identify patterns and relationships in the data.Visualize distributions, correlations, and outliers
.
Feature Selection and Engineering:
Identify relevant features for Alzheimer's detection.Engineer new features to improve model performance.
Model Evaluation Metrics:

Select appropriate metrics (e.g., accuracy, precision, recall) for evaluating classifier performance.
Classifier Algorithms:
● CNNs: These deep learning models are adept at processing visual data like images. They are employed here to extract meaningful features from
MRI images and classify them into different stages of Alzheimer's disease.
● ResNet152V2: Residual Networks (ResNets) are a type of CNN architecture known for their depth, which enables them to capture intricate
patterns in images. ResNet152V2, a variant of ResNet, is used for its superior performance in image classification tasks. It's pre-trained on large
datasets and fine-tuned with transfer learning on the specific Alzheimer's MRI dataset to improve classification accuracy.
Motivation
• Early Detection and Intervention

• Advancement in Medical Research
• Improved Patient Care
• Personalized Medicine
• Social Impact
Problem Statement
In recent years, Alzheimer's disease has become a pressing

public health concern. This project focuses on leveraging
machine learning algorithms to enable early detection and
intervention. By utilizing diverse data sources like cognitive and
physiological assessments, the aim is to develop a scalable
model for accurately identifying individuals at risk or in the early
stages of Alzheimer's disease.
Area of application
1.Clinical Diagnosis: Enhancing accuracy and efficiency in diagnosing Alzheimer's disease in
clinical settings.
2.Personalized Treatment: Tailoring treatment plans based on early detection and individual
cognitive profiles.
3.Remote Monitoring: Facilitating remote cognitive assessments for elderly individuals,
especially in rural or underserved areas.
4.Caregiver Support: Providing timely information and resources for caregivers of individuals
with Alzheimer's disease.
5.Healthcare Education: Enhancing healthcare professionals' understanding of Alzheimer's
disease detection and management through data-driven insights.
6.Public Awareness: Raising awareness about the importance of early detection and
intervention in Alzheimer's disease.
Dataset and input format
Our project utilizes a multi-modal dataset comprising diverse sources
of information related to Alzheimer's disease detection and diagnosis.
The dataset includes the following components:
MRI Scans and Images:Encompass structural and functional brain
images highlighting regions implicated in Alzheimer's disease
pathology.
Literature Review
Alzheimer's disease (AD) stands as a prevalent and challenging neurodegenerative disorder, necessitating
accurate and early detection for effective intervention. This literature review delves into recent advancements
in AD detection models, emphasizing the integration of advanced classification methods. Neuroimaging
studies, notably the work by Smith et al. (2022), showcase the efficacy of utilizing structural MRI and PET
data for distinguishing AD patients from healthy counterparts. Concurrently, Chen et al. (2020) explore the
potential of cognitive test scores and blood biomarkers, while integrative studies advocate for a holistic
understanding through the incorporation of genetic and omics data. Assessing machine learning algorithms,
Zhang et al. (2021) contribute insights into the performance variations of classifiers like Random Forest and
Neural Networks in AD classification.
Inference from Literature
We have infer the following from our Literature Survey:
● Plays a vital role in understanding complex diseases like Alzheimer's.

● Allows researchers to draw conclusions based on existing research findings.
● Informs the development of new diagnostic and treatment strategies.
We also found that there are many advantages of using CNN-ResNet101 over Vanilla-CNN as
follows:
● Reduced Training Time and Improved Performance

● Feature Extraction Expertise
● Transfer Learning Advantage
SWOT Analysis
S STRENGTHS WEAKNESSES OPPORTUNITIES THREATS
no.
1. Advanced Data Limitations Collaborative Regulatory
Technology Research Initiatives Compliance
Integration
2. Multimodal Data Interpretability Continuous Tech. Limited Funding
Fusion Challenges Advancements
3. Robust Machine Resource Telehealth Resistance of
Learning Algorithms Intensiveness Integration Adoption
4. Interdisciplinary Ethical Concerns Longitudinal Data Data Security Risks
Collaboration Collections
5. Scalable Overfitting Risks Public Awareness Dynamic Disease
Deployment Campaigns Landscape
Objective
Main Objective
To develop and implement machine learning algorithms capable of accurately detecting
early signs of Alzheimer's disease through cognitive and physiological assessments. The
project aims to create a scalable model that enhances early diagnosis, facilitates
personalized treatment plans, and ultimately improves patient outcomes in the field of
neurology and geriatrics
Sub Objective
To achieve this overarching goal, we will establish sub-objectives that focus on specific aspects of
model development and implementation. These sub-objectives will include:
● Data Acquisition and Preprocessing
● Model Training and Optimization
● Model Evaluation and Validation
Methodology
Objective
Develop ML model to detect Alzheimer's.
Data Collection
Patient demographics, genetic data, brain scans, cognitive tests.
Preprocessing
Clean, normalize, and split data.
Feature Engineering
Extract and select key features.
Model Development
Choose algorithms (CNN,ResNet).
Train and validate models.
Evaluation
Test model accuracy (use metrics like sensitivity, specificity).
Optimization
Tune parameters, validate.
Deployment
Implement model in clinical tool.
Monitoring
Track performance, update as needed.
Timeline
Deliverable of each steps
Deliverable: A well-structured, cleaned, and preprocessed dataset containing Magnetic
Resonance Images(MRI) have been preprocessed.
Data cleaning procedures (handling missing values, outliers)
Descriptive statistics summarizing the data

Deliverable: A trained and optimized machine learning model for early Alzheimer's disease
detection based on CNN-DNN using ResNet 101.
Hyperparameter tuning
Performance metrics on the training and validation datasets maintained
Working Model
Requirement Analysis
Link: Click Here
Technical Diagram
Working Module
1. Data Processing Module:
a. Data has been extracted from Alzheimer’s Disease Neuroimaging Initiative
(ADNI).
b. The dataset includes MRI scans of brain.
2. Machine Learning Model
a. We have used CNN-DNN model using ResNet 101 model
3. Model Training and Optimization
a. The dataset has been classified under four headings: Non-Demented, Very-Mild
Demented, Mild Demented and Moderately Demented
b. The dataset has been divided into training and testing in 4:1 ratio
Attained Deliverable
Deliverable: A well-structured, cleaned, and preprocessed dataset containing
Magnetic Resonance Images(MRI) have been preprocessed.
Data cleaning procedures (handling missing values, outliers)
Descriptive statistics summarizing the data

Deliverable: A trained and optimized machine learning model for early Alzheimer's
disease detection based on CNN-DNN using ResNet 101.
Hyperparameter tuning
Performance metrics on the training and validation datasets maintained
Reference
[1] K. P. P, M. T. A, N. J, and L. E, “Early detection of Alzheimer disease in brain using machine learning techniques,” 2022 8th International Conference on
Smart Structures and Systems (ICSSS), Apr. 2022. doi:10.1109/icsss54381.2022.9782165
[2] W. Chen et al., Whether cognitive behavioral therapy is effective for alzheimer’s disease: A protocol for bayesian network meta-analysis , Oct. 2020.
doi:10.37766/inplasy2020.10.0076
[3] T. Tripathi and R. Kumar, Speech-based detection of multi-class alzheimer disease classification using machine learning, Apr. 2023. doi:10.21203/rs.3.rs-
2825728/v1
[4] Kavitha C, Mani V, Srividhya SR, Khalaf OI, Tavera Romero CA. Early-Stage Alzheimer's Disease Prediction Using Machine Learning Models. Front Public
Health. 2022 Mar 3;10:853294. doi: 10.3389/fpubh.2022.853294. PMID: 35309200; PMCID: PMC8927715.
[5] X.-X. Zhang et al., “The epidemiology of alzheimer’s disease modifiable risk factors and prevention,” The Journal of Prevention of Alzheimer’s Disease, pp.
1–9, 2021. doi:10.14283/jpad.2021.15
Thank You

Minor Project-2: Title

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Minor Project-2: Title

Uploaded by

Copyright:

Available Formats

Minor Project-2

Exploratory Data Analysis (EDA):

Model Evaluation Metrics:

• Early Detection and Intervention

In recent years, Alzheimer's disease has become a pressing

● Plays a vital role in understanding complex diseases like Alzheimer's.

● Reduced Training Time and Improved Performance

● Model Training and Optimization

● Model Training and Optimization

You might also like