Finals | Pre/Post-medical Image Processing Flashcards
Let f (x, y) be a two-dimensional (2-D) cross- sectional image of a three-dimensional object.
2-D image reconstruction.
States that f (x, y) can be reconstructed from the cross-sectional one-dimensional projections.
Image reconstruction theorem
Mathematically, the image reconstruction theorem can be described with the help of —
Fourier transform (FT)
II. BACKGROUND REMOVAL Backgrounds can be removed in both (1) and (2) images. The idea is to (3) by discarding the (4) from the image.
- sectional (cross-section)
- projection
- reduce the size of the image file
- background
II. BACKGROUND REMOVAL
In digital images, unexposed areas appearing white on a display monitor is defined as (1) in this context.
(2) in this context means that the brightness of the background is converted from white to black
- background
- Background removal
III. NOISE REMOVAL
Quality of an image can be measured from:
A. Sharpness
B. Resolving Power
C. Noise Level
S RP Noise
Inherited from the design of the instrumentation
A. Sharpness
B. Resolving Power
Arises from photon fluctuations from the energy source and detector system used, and electronic noise accumulated through the imaging chain.
C. Noise Level
Noise Level arises from (1) from the energy source and detector system used, and (2) accumulated through the imaging chain.
- photon fluctuations
- electronic noise
Means that the image is less noisy.
High signal-to-noise ratio (SNR)
A common method to increase the SNR (i.e., reduce the noise in the image) is to —
obtain many images of the same object under the same conditions and average them
III. NOISE REMOVAL
Noise is present in all medical images. We can reduce the amount of noise by increasing (1) or increasing (2). (3) also offers ways to reduce noise, but sometimes at the cost of introducing artifacts.
- radiation dose (for X-ray methods)
- field strength or imaging time (MR)
- Image processing
Also offers ways to reduce noise, but sometimes at the cost of introducing artifacts.
Image processing
Noise Reduction Filters:
- Block Filtering/ Averaging
- Gaussian filter
- Median filtering
- Anisotropic diffusion filtering
- Bilateral filtering
- Nonlocal means filter
GAM AD BiNon
Simplest form of filtering
Block Filtering/ Averaging
Blurs the image and are classified by their sigma or width
Gaussian filter
Considers a patch of neighborhood pixels. They exhibit less blurring than averaging and gaussian filters
Median Filtering
Smoothens regions without strrong edges while preserving edges
Anistropic diffusion filtering
Recently developed method similar to the Gaussian filter
Bilateral filtering
Newer technique that examines a “patch” centered about each pixel.
Main concept: similar patches, when averaged, will wash out noise
Nonlocal Means Filter
Process applied to a graphics file to minimize its size without degrading image quality below an acceptable threshold.
IMAGE COMPRESSION
IV. IMAGE COMPRESSION
By reducing the file size, more (1) can be stored in a given amount of (2).
- images
- disk or memory space
IV. IMAGE COMPRESSION
Image compression attempts to (1) to minimize the (2) for a given image.
- reduce or eliminate the presence of redundancies
- storage size or transmission time requirements
IV. IMAGE COMPRESSION
In general, three types of redundancies can be identified that are relevant to compressed medical images:
a. Coding (R/IR)
b. Spatial, temporal, and bit depth (R/IR)
c. Psychovisual (IR)
IV. IMAGE COMPRESSION
Reversible (R): (1)
Irreversible (IR): (2)
- Lossless
- Lossy
Achieves higher compression levels
lossy compression algorithms
Pre-medical image process
I. IMAGE RECONSTRUCTION
II. BACKGROUND REMOVAL
III. NOISE REMOVAL
IV. IMAGE COMPRESSION
A processing method that enhances or removes a specific component in a signal or image.
I. FILTERING
I. FILTERING
The name could reflect what is (1), what is (2), or the (3) that is used.
- removed
- enhanced
- calculation
Subtopics under filtering
A. HISTOGRAM MANIPULATION
B. ENHANCEMENT
Standard window/level manipulation
HISTOGRAM MANIPULATION
Reducing the window width increases contrast, while wider windows reduce contrast.
HISTOGRAM MANIPULATION
This technique is useful for visual comparison across patients in MR exams.
HISTOGRAM MANIPULATION
Histograms may also be matched between images to produce similar (1) between images from different (2).
- contrast
- studies
Window width vs window level
Window width: Adjusting the contrast (x-axis)
Window level: Adjusting the density (y-axis)
Two commonly used enhancement in medical imaging are
- Unsharp masking
- Edge sharpening filters
A standard enhancement filter that selectively subtracts a blurred image from the original.
Unsharp masking
- (1): Pixels in the blurred image that differ from the original by an amount more than a user-specified threshold are considered to be ‘‘(2).’’ Any pixel under the mask is subtracted from the original, otherwise the pixel is (3).
- Unsharp masking
- masked
- unmodified
Selectively enhance edges in the image based on edge strength.
Edge sharpening filters
Process of becoming smaller
Contraction
Refers to any method that makes particular elements in the image more visible (in the context of image post-processing)
Enhancement
Color, when used, is mostly for (1), namely by a range of (2) converted to (3) in order to enhance the visual appearance of features within this range.
- enhancement
- gray levels
- pseudo colors
Process of aligning images with one another. This means that the same tissue sample exists at a given X, Y, Z location on all registered images.
REGISTRATION
Registration may be
- Intra-subject
- Inter-subject
- Intra-modality
- Inter-modality
Intra-subject
within the same subject
Inter-subject
across different subjects
Intra-modality
using images from one modality
Inter-modality
using images from multiple modalities
After relevant image features are extracted, (1) ensues to derive a usable output (e.g., likelihood of malignancy, differential diagnoses) from the analysis of the features.
- classification
A. CLASSIFICATION
After relevant image features are extracted, classification ensues to derive a usable output (e.g., (1), (2)) from the analysis of the features.
- likelihood of malignancy
- differential diagnoses
A. CLASSIFICATION
Depending on the clinical context for the particular patient who underwent the imaging study, the final output may include
- detection of an abnormality
- description/characterization of a known abnormality
- providing a diagnosis or list of differential diagnoses
- evaluation of disease progression
B. TEXTURING
* Texture analysis methods can be categorized into three main approaches:
a. statistical
b. transform
c. model-based
based on the description of image texture by unique statistical features
statistical
based on the transformation of the image using for example Fourier or wavelet approaches and characterizing texture in the transform domain
transform
mathematical model is developed that represents the texture.
model-based
Most essential medical imaging process as it extracts the region of interest (ROI) through a semiautomatic or automatic process.
Image segmentation
It divides an image into areas based on a specified description
Image segmentation
C. SEGMENTATION
It divides an image into areas based on a specified description, such as
- segmenting body organs/tissues in the medical applications for border detection
- tumor detection/ segmentation
- mass detection.
Post-medical Image Process
I. FILTERING
A. HISTOGRAM MANIPULATION:
B. ENHANCEMENT:
II. CONTRACTION AND ENHANCEMENTS
III. REGISTRATION
IV. CLASSIFICATION, TEXTURING, AND SEGMENTATION
