Statistical Analysis and Information Entropy in Medical Imaging

Comprehensive practice flashcards covering statistical analysis, information entropy, coding types, image transformations, noise removal filters, and post-processing techniques in medical imaging.

Entropy

Also known as Shannon’s Entropy (1948) or Information Entropy, it describes the amount of randomness or uncertainty in a signal or image and represents the quantitative digital value of the image.

Fixed Length Coding

The use of the same number of bits to represent all pixels, allowing a program to read an image file without needing special symbols to represent the end of each pixel’s data.

Variable Length Coding

A coding method that uses a variable number of bits to represent pixel values, providing short code-words for frequent characters and long code-words for infrequent characters based on frequency of occurrence.

Coding Redundancy

The use of more bits than are needed to convey a given amount of information, which can lead to bloated source code and reduced reliability.

Encoded

A compressed image, referring to an analog image already connected to a digital image that a computer can read.

Decoded

A compressed image that has been reconstructed into its digital form so it can be viewed on a monitor.

Huffman Coding

An entropy encoding algorithm developed in 1952 used for lossless data compression that replaces each pixel value with a special code on a one-to-one basis.

Arithmetic Coding

A compression method developed in 1963 that replaces each pixel value with one code by assigning each symbol an interval; it is asymptotically better than Huffman coding in terms of complexity.

Fourier Transform

The transformation pair $f(x, y)$ and $F(u, v)$ that enables the conversion of a 2D image from the spatial domain to the frequency domain and vice versa.

Frequency Domain

A representation of an image describing how fast the brightness changes from one pixel to another.

Rigid Registration

A transformation involving only reflection, translations, and rotations, suitable for registering images of rigid objects like bone without changing their shape.

Affine Transformation

A transformation model that preserves the parallelism of lines but not their lengths or angles, extending rigid transformation with scaling and shearing.

Nonlinear Transformation

A transformation allowing the mapping of straight lines to curves, which can be free-form or guided by physical models like tissue elasticity.

Image Reconstruction Theorem

A theorem stating that a 2D cross-sectional image $f(x, y)$ can be reconstructed from 1D projections, often described mathematically using the Fourier transform.

Background Removal

The process of reducing image file size by converting the brightness of unexposed areas (background) from white to black.

Signal to Noise Ratio (SNR)

A measure where a high ratio indicates an image is less noisy; it can be increased by averaging many images of the same object or increasing radiation dose/field strength.

Block Filtering

Also known as averaging or mean filtering, this simplest form of filtering replaces each pixel with the average value of its neighboring pixels.

Gaussian Filter

A filter that blurs an image by giving more importance to nearby pixels than distant ones, classified by their width.

Median Filtering

A filter that chooses the middle value of pixels rather than averaging them, making it the best method for salt and pepper appearance noise.

Anisotropic Diffusion Filtering

A noise reduction method designed to remove noise while preserving the edges within the image.

Non Local Means Filter

A newer filter technique that searches the entire image for patches centered above each pixel rather than looking only at nearby pixels.

Image Compression

The process of minimizing graphics file size to save storage space or transmission time without degrading quality below an acceptable threshold.

Histogram manipulation

A post-processing method where reducing the window width increases contrast, while wider windows reduce contrast.

Unsharp Masking

A standard enhancement filter that selectively subtracts a blurred image from the original to highlight specific pixels.

Registration

The process of aligning images so that the same tissue sample exists at a given $X, Y, Z$ location across all registered images.

Intrasubject Registration

The alignment of images taken from the same patient, such as different slices or intervals in a procedure.

Segmentation

An essential medical imaging process that divides an image into areas based on a description to extract a region of interest, such as body organs.

Class 1 Display Monitors

High-quality display monitors specifically designated as workstations used for diagnosis.

Workstation

A fully equipped computer terminal that can access images, manipulate image quality, and permanently save changes to the PACS.

Display Station

A computer terminal limited to the display of stored images with no ability to permanently manipulate, change, or print the images.

Interpolation

The mathematical process of assigning a value to a dead pixel based on recorded values of adjacent pixels, or estimating a value between known values.

Extrapolation

The process of estimating a value that lies beyond the range of known values.

Volume Rendering

Methods used in computer graphics to create a 2D projection from discretely sampled 3D data, such as CAT or MRI scans.

Surgical Theater Precision Virtual Reality

A specific VR technology used to create 360-degree patient models for planning neuro and cardiac surgeries.