Resting-State fMRI:

Large Data Analysis

in Neuroimaging
TABLE OF
CONTENTS
0 INTRODUC
TION
0 ANALYSIS

1 3
You can describe the You can describe the
topic of the section topic of the section
here here

0 PRESENTA
TION
0 CONCLUSI
ON

2 4
You can describe the You can describe the
topic of the section topic of the section
here here
 Principles of Functional Magnetic
Resonance Imaging (fMRI)
Mechanism:
Definition:
• Measures Cerebral Blood
fMRI is a brain imaging Flow and Neuronal
technique used to Activity.
detect brain activity by • Tracks functionally
measuring changes in dependent levels of blood
blood flow. oxygen, termed Blood
Oxygen Level
Dependent (BOLD)
signal.
Key Components:

• BOLD Signal: Depends on Cerebral Blood Flow, Blood Volume, and

Metabolic Rate of Oxygen.
 Principles of Functional Magnetic
Resonance Imaging (fMRI)
Voxel Concept:
• A voxel is a cubical •Neurons within a voxel
region in the brain become active during
where BOLD signals are brain activity, causing
measured. variations in the BOLD
signal over time.
Neural Activity & BOLD Signal:

• During activation, the BOLD signal peaks 4–6 seconds after onset, decreases,
and peaks again at 8–12 seconds.
• Returns to baseline after 20 seconds if no additional neural activity occurs.
• This time-dependent response is called the Hemodynamic Response Function
(HRF).
 Resting State fMRI (rsfMRI) for
Neuroimaging

rsfMRI is used to study functional connectivity between

different brain regions by analyzing spontaneous brain activity
during rest.
Mechanism
Variations are linked to changes
Detects low-frequency
in deoxygenated hemoglobin
variations in the BOLD signal
and blood oxygenation,
(approximately 0.01–0.1 Hz).
driven by Cerebral Blood
Flow (CBF).
 Resting State fMRI (rsfMRI) for
Neuroimaging
Methods of Analysis

1 2
Whole Brain Region-of-Interest
Investigation (ROI) Analysis
• Examines • Focuses on specific
connectivity across predefined regions.
the entire brain. • Provides detailed
• Provides a global analysis of selected
view of brain areas.
networks.
 The Measurement of Fully
Connected Network and
Construction
Measurement of Fully Connectedof Default Mode
Region Construction of DMN
Network
• Brain Division: The brain is divided into 90 (DMN)
• Seed Region: DMN is constructed using a
regions. seed region located in the bilateral medial,
• Time Series Analysis: The average time inferior temporal gyrus, and precuneus
series of all voxels in each region is region of the brain.
calculated.
• Correlation Measurement: Correlation • Threshold Setting: The threshold for DMN is
between every pair of regions is measured. determined using the standard deviation of
• Correlation Matrix: the subnetwork.
• A 90×90 correlation matrix is
constructed.
• The average correlation matrix is
calculated to represent the fully
connected network.
 Brain Connectivity

Anatomical Functional
Connectivity Connectivity
Definition: Definition:
Refers to the physical connections Refers to the temporal correlation
between neural units (e.g., neurons, of activity between different brain
brain regions). regions, indicating coordinated
functionality.
Imaging Tools:
Static images, such as those from Imaging Tools:
MRI or diffusion tensor imaging Electroencephalogram (EEG).
(DTI), help neurologists identify the Spectroscopy.
underlying structure of the brain. Resting-State fMRI (rsfMRI).
 Better Image Availability
Large
data

Features Challenges

Huge sample High The sheer volume of Representation of

sizes dimensionality data creates this large-scale data
(more variables, significant requires advanced
complex patterns). challenges in data computational tools
storage, processing, and methods.
and analysis.
 Better Image Availability

The large sample size

Big data enables the
improves our
discovery of hidden
understanding of
patterns within
heterogeneity in brain
neuroimaging images.
activity and diseases,
leading to better insights
and treatments.
 Need of Resting-State MRI IDEA

01 02 03
Cerebral Energetics Signal to Noise Ratio Multi-Purpose Data
(SNR) Sets
The brain consumes 20% of
the body’s energy at rest, rsfMRI has higher SNR rsfMRI can study
supporting neuronal compared to task-based multiple cortical systems
signaling. fMRI. with a single data set.
 Need of Resting-State MRI IDEA

04
05
Expanded Patient Populations Reliability
rsfMRI is beneficial for patients who addressing reproducibility and
cannot perform tasks (e.g., those with intersubject variability in rsfMRI
cognitive dysfunction or physical ensures that the findings are
impairments). reliable
Mild Cognitive Impairment (MCI) and Alzheimer’s
(AD)

● rsfMRI imaging is used to find the intrinsic

brain connectivity which is considered a
promising biomarker for AD and MCI
diagnosis

● In comparison with health controls, there is

a decreased connectivity at baseline in
posterior (memory) default mode areas and
increased in frontal (higher cognitive
functions) region for AD patients.
Fronto-Temporal Dementia (FTD)

In bvFTD (behavioral variant FTD), there is a reduced

connectivity in important regions like:

• Anterior Cingulate Cortex: Involved in

motivation, emotion regulation, and decision-
making.

• Amygdala: Responsible for processing emotions,

especially fear.

• Frontal Lobe: Involved in executive functions

such as planning, reasoning, and impulse control.

• Thalamus and Ventral Striatum: Involved in

regulating movement and mood.

• Insula: Plays a role in emotional processing and

 Multiple Sclerosis (MS) & Amyotrophic
Lateral Sclerosis (ALS)

Multiple Sclerosis (MS) Amyotrophic Lateral

Sclerosis (ALS)
• In MS, DMN dysfunction is a
prominent issue, showing changes Decreased connectivity observed in
in brain activity and structural the frontal cortex and precuneus
connectivity. (important regions for cognitive and
motor function).
• Structural connectivity changes in
MS are associated with damage to
white matter, the part of the brain
that connects different regions.
 Depression

rsfMRI in Depression:

• Functional neuroimaging helps assess brain activity

in regions involved in mood regulation.

• Key regions impacted by depression include:

Anterior Cingulate
Medial Prefrontal Cortex
Posterior Cingulate
Thalamus
Precuneus
Schizophrenia

Functional Connectivity Issues:

Schizophrenia patients show reduced functional

connectivity between key brain areas, particularly:
• Medial Frontal Cortex: Involved in
decision-making, attention, and
executive functions.
• Precuneus: Associated with self-
awareness, memory, and spatial
orientation.
Resting-State Functional Imaging of Neonatal
Brain Image
● Infancy is a crucial period for brain
development. During this time, the brain
undergoes rapid growth and functional
maturation.
● rsfMRI (Resting-State Functional MRI) is a
method used to monitor how different
parts of the brain connect and
communicate while the baby is at rest.
⮚ It helps to understand how the
brain regions work together
during this important stage of
development.

⮚ It plays a key role in detecting

brain injuries or abnormalities
early
Pre_processing
step0 0 0
1
Denoising 2
Resampling with 3
Skull Stripping
Algorithm Bilinear Interpolation

0 0
4
Connected
Component
Image
Normalization
6
Filters
Analysis
rsFMRI of infant
brain.
Axial

Coronal

Saggital
 Machine Learning & Deep Learning in
rsfMRI Data Analysis
Role of Machine Learning & Deep Learning:
Vital in analyzing rsfMRI data to identify neurological
disorders.
Unsupervised
• DiscoverLearning
spatial patterns and latent
factors in resting-state functional MRI
images.
• Helps identify reproducible patterns
across different subjects and brain regions.
Supervised
Learning:
• Used to classify and analyze
various brain disorders.
• Effective for identifying hereditary
conditions and other neurological
abnormalities.
 Supervised Learning
Methods
Support Vector Machines(SVM)
• linear SVM,
0 • Kernelized SVM.
1

0
2
Linear Models:
• Ridge regression,
• LASSO regression,
• Elastic-Net regression,
• Logistic regression.
 Supervised Learning
Methods
Decision Trees:
• Random forest,
• Gradient tree , 0
4

0
3
Deep Learning
• Fully connected network,
• Convolutional network.
 Unsupervised Learning
MethodsLatent Models
⮚ Decomposition methods,
0 ⮚ Markov model.
1

0
Clustering 2
⮚ K-means, Spectral cltering,
⮚ Hierarchical clustering,
⮚ Gaussian model.
Unsupervised Learning
Methods
Non-linear Models:
⮚ Local linear ,

⮚ Autoencoders.

0
3
 THAN
KS!
Team members :
Omnia Mohammed
Salma Magdii
Aliaa Tarek
Supervised by:
Dr.Adel Fathy