Medical Imaging Flashcards

Flashcards generated from Medical Imaging lecture notes.

Image in Image Processing

A function that maps image coordinates x, y to intensity values.

f(x,y) for gray-value images

A scalar function.

f(x,y) for color images

A vector consisting of the color channels.

Histograms of Images

Information about the distribution of the intensity values of an image.

A histogram h(i)

Consists of several bins that contain single intensity values or ranges of intensities.

For each bin i

The number of occurrences ni of the corresponding intensity values in the image are counted.

CDF

Cumulative distribution function

For visual inspection, it is often beneficial:

To change the contrast of an image.

Window and Level

Semi-automatic adjustment of the display using window and level functions.

In CT, the gray values have known physical properties and allow interpretation of the material. This effect is displayed in Fig. 3.2. The image on the left hand side displays the Hounsfield unit (HU)

From -1000 to 1000.

This range covers the materials

Air, soft tissue, contrast agent, and bone.

Image Contrast is

Represented by a difference in pixel values

Digital Image Processing

Contrast is changed by changing pixel values.

A Look Up Table

Gives Replacement Pixel Values

LUTS are

Selected to Give an Image Specific Contrast

Contrast Processing

A LUT that looks like High Contrast Film

Contrast Processing

A LUT that Produces Contrast appropriate for Chest Imaging

Inverted Brightness Scale

A LUT that Inverts the Brightness Scale

Digital Image Windowing

Windowing separates the Pixel Value Range into 3 Segments

Effect of Window Selection

Windows to Enhance Contrast in Different Segments of the Pixel Value Range

Blurred (Unsharp) Mask Subtraction

Visibility of Detail Enhanced by the Blurred Mask Subtraction Process

Gamma Correction

A is a normalization constant to keep the resulting intensities within a valid range, gamma is the gamma correction parameter. If gamma < 1: The image becomes brighter. If gamma > 1: The image becomes darker. If gamma = 1: The image remains unchanged.

Histogram Equalization

Aim to distribute pixel intensity values more evenly across an image, making details more visible.

Edge detection

Used to identify boundaries within an image by detecting areas of rapid intensity change.

Image Filtering

Filters can be applied to images in order to process them, e. g., for noise reduction.

Average / Mean / Box Filter

Averaging filter of size

Gaussian filter

A better choice for blurring an image than the averaging filter is the Gaussian filter.

Median Filter

A non-linear image filtering technique primarily used for noise reduction, particularly in images affected by salt-and-pepper noise. Unlike linear filters, such as the mean filter, which average pixel values, the median filter preserves edges while effectively reducing noise.

Morphological operators

Tools in image processing, primarily used for analyzing binary and grayscale images.

Image Segmentation

The process of partitioning an image into multiple meaningful regions to simplify analysis and processing.

Thresholding

The simplest method, where pixel intensity is compared to a predefined threshold.

Edge-Based Segmentation

Detects boundaries between objects using edge detection.

Region-Based Segmentation

Region Growing: Starts from seed points and expands based on similarity in pixel intensity. Watershed Algorithm: Treats intensity values as topographical elevation and partitions the image into catchment basins.

Deep Learning-Based Segmentation

U-Net, Mask R-CNN, FCN (Fully Convolutional Networks): Used for medical and object segmentation. Requires large datasets and computational power.

Noise Types

Caused by the statistical variation in the number of X-ray photons reaching the detector. Follows a Poisson distribution, meaning noise is higher in low-intensity regions.

Electronic Noise

Introduced by the electronic components of the X-ray detector . Includes thermal noise, amplifier noise, and ADC (Analog-to-Digital Converter) noise.

Scattering Noise

X-rays scatter within the body, leading to blurred edges and reduced contrast in the image.

Motion Artifacts

Caused by patient movement or scanner vibrations during exposure.

Beam Hardening Artifacts

Occurs when lower-energy X-ray photons are preferentially absorbed, making the remaining beam “harder” (higher energy). Results in non-uniform attenuation and streaking artifacts, especially around dense objects like bones or implants.

Overexposure

Occurs when too many X-ray photons reach the detector, resulting in an image that is too bright (high signal intensity), potentially leading to loss of detail in low-density structures.

Underexposure

Occurs when too few X-ray photons reach the detector, leading to a dark and noisy image with insufficient information for accurate diagnosis.

Balancing Exposure in X-ray Imaging

Exposure Index (EI): Modern digital X-ray systems provide an EI value to guide proper exposure. Histogram Analysis: Digital systems analyze pixel intensity histograms to detect over/underexposure. Dose Optimization: Use the “As Low As Reasonably Achievable” (ALARA) principle to ensure the lowest radiation dose while maintaining.