IMAGE ANALYSIS
Contents
PACS (Picture Archiving and Communication System)...................................................1
Components...................................................................................................................1
Digital Image.....................................................................................................................2
Sampling: Nyquist.........................................................................................................2
Fourier...............................................................................................................................2
Encoding and Storage........................................................................................................2
Parametric Images.............................................................................................................3
DICOM (Digital Imaging and Communications in Medicine).........................................3
Image Processing...............................................................................................................3
Identifying Features...........................................................................................................3
Registration........................................................................................................................3
Image Registration.........................................................................................................3
Rigid Body Registration................................................................................................4
Spatial Normalization........................................................................................................4
ROI (Region of Interest)....................................................................................................4
Segmentation.....................................................................................................................4
VBM (Voxel-Based Morphometry)...................................................................................4
PACS (Picture Archiving and
Communication System)
Definition: A medical imaging technology that provides economical storage,
retrieval, management, distribution, and presentation of medical images.
Components
Image Acquisition: Images are acquired from various modalities (e.g.,
X-ray, CT, MRI).
Storage: Digital storage replaces physical film archives, reducing
storage space and enhancing retrieval efficiency.
Transmission: Images are transmitted over a network to various
locations, allowing remote access and review.
Display: Images are displayed on high-resolution monitors for diagnosis
and treatment planning.
Advantages:
Efficiency: Streamlines workflow from image acquisition to reporting.
Accessibility: Enables access to images from multiple locations.
Integration: Integrates with other hospital information systems (HIS,
RIS).
Digital Image
Sampling: Nyquist
Nyquist-Shannon Theorem: If a function contains no frequencies higher than ν
Hertz, it can be completely determined by samples spaced 1/(2ν) seconds apart.
Sampling Frequency: Must be at least twice the highest frequency present in
the analogue signal (Nyquist criterion).
Aliasing: Occurs when the sampling frequency is less than twice the highest
frequency, leading to incorrect representation of the signal.
Fourier
Fourier Analysis: Decomposes a signal into its constituent frequencies.
Continuous Fourier Transform (CFT): Converts a function from time/space
domain to frequency domain.
Formula: 𝑓(𝑣)=∫−∞∞𝑓(𝑥)𝑒−2𝜋𝑗𝑣𝑥𝑑𝑥f(v)=∫−∞∞f(x)e−2πjvxdx
Discrete Fourier Transform (DFT): Used for discrete signals, representing the
amplitude and phase of sinusoids in the signal.
Formula: 𝑋𝑘=∑𝑛=0𝑁−1𝑥𝑛𝑒−𝑗(2𝜋/𝑁)𝑘𝑛Xk=∑n=0N−1xn
e−j(2π/N)kn
Encoding and Storage
Image Encoding: Assigns numerical values to image samples, representing
intensity at each point.
Grey-scale Representation: Uses varying bits (e.g., 8-bit, 4-bit) to represent
different shades of grey.
Storage Formats: Bitmap, JPEG, PNG, each with specific compression and
quality characteristics.
Parametric Images
Definition: Images that present one or more biologically relevant parameters.
Example: T2 map in MRI obtained by fitting voxel intensities to an exponential
decay function.
DICOM (Digital Imaging and
Communications in Medicine)
Definition: A standard for handling, storing, printing, and transmitting
information in medical imaging.
Components:
File Format: Includes a header with metadata and a pixel data section.
Network Protocol: Ensures interoperability between different imaging
devices and systems.
Purpose: Facilitates the integration of medical imaging devices from multiple
manufacturers.
Image Processing
Quality Metrics: Spatial resolution, contrast, noise.
Quantitative Analysis:
Detection Task: Identifying abnormalities.
Quantification Task: Measuring signal concentration in a voxel.
Identifying Features
Rose Model: A probabilistic model for low-contrast object detection.
Formula: 𝑘=signalnoise=𝐶⋅Φ⋅𝐴k=noisesignal=C⋅Φ⋅A
𝐶C: Contrast, ΦΦ: Photon density, 𝐴A: Area.
Registration
Image Registration
Definition: Aligning image volumes from different sources to a common
coordinate space.
Applications: Comparing images from different modalities (e.g., CT, MRI).
Rigid Body Registration
Definition: Uses translations and rotations to align images.
Applications: Co-registering images of the same subject from different
modalities.
Spatial Normalization
Definition: Non-rigid transformation to align images to a standardized space.
Applications: Brain imaging, comparing data across groups.
ROI (Region of Interest)
Definition: Specific area within an image for detailed analysis.
Applications: Measuring signal intensity in defined regions, such as organs or
lesions.
Segmentation
Overview
Definition: Process of defining ROIs in images.
Challenges: Can be subjective and vary between observers.
Approaches
Manual Segmentation: Hand-drawn ROIs, labor-intensive.
Automated Segmentation: Edge detection, clustering, and other algorithms.
VBM (Voxel-Based Morphometry)
Definition: A technique for comparing brain anatomy across individuals.
Steps:
Spatial Registration: Align images to a common space.
Segmentation: Classify voxels into different tissue types.
Smoothing: Apply Gaussian filters to images.
Statistical Analysis: Perform voxel-wise comparisons to identify
differences.
