Medical Image Computing Flashcards
What are the main steps in medical image formation?
Generate Energy - Energy transfer (focus, collimate,…) - Tissue Interaction - Energy collection - Image formation, reconstruction
Explain the main steps for medical image formation based on a modality of your choice.
US: Ultrasonic waves generated, penetrate the body, interact with the tissue: reflection, device collects the thrown back waves, generates US image
X-Ray: Electromagnetic waves, guiding waves with magnetic field and shutters through patient, interaction with tissue: Different tissue types have different absorption coefficients (Beer-Law), collect pass through radiation with detector (e.g. flat panel detection), format image by processing sensor dates
What are the characteristics of the electromagnetic energy spectrum? Which bands are used for medical imaging?
Two perpendicular waves ( oscillating electric field and magnetic field)
Bands: 1. Radion (MRI 50 MHz), 2. Visible (Optical Imaging 600 THz), 3. X-ray (300 PHz), 4. Gamma (30EHz, PET SPECT)
What other energy sources are used for medical imaging? Give an example.
Acoustic: aeroacoustics, microphones measure sound pressure at different locations
Ultrasonic: US waves
Electronic: sensor placed behind a camera lens to translate an image into an electronic signal
What are usual imaging resolution ranges in medical imaging?
Imaging resolution: MRI > CT > US > OCT
Imaging penetration depth: MRI = CT > US > OCT
For lungs of Covid-19 patients: CT, due to depth and resolution
Which modalities are useful for full-body imaging?
Radiography/X-Ray
Flat panel detectors are used to convert X-rays to light photons which are then measured by CCD/CMOS sensors (indirect measurement)
+ RT: Relatively low cost, Deep penetration with good resolution, Wide availability and standards
- CT: higher radiation exposure than X-Ray – from different angles: carcinogenic
- MR: expensive, takes very long + noisy -> discomfort
- US: cumbersome, since body contact would be necessary everywhere, time-consuming
What is functional imaging? Give an example.
= Imaging method to visualize function of organs/tissue
* Information on changes e.g. Metabolism, blood flow (e.g. detection of cancer)
* Opposed to structural imaging, specific areas within a certain tissue or organ imaged
* Often use tracers or probes to reflect spatial distribution within the body based on bounding
Examples: Functional MRI, PET/SPECT, EEG, Optical imaging, Angiography
. Which imaging modality would you use for imaging the lungs, e.g. of Covid-19 patients? Give a rationale.
CT
+ Good penetration depth to resolution ratio
+ Series of X-ray images from different angles
+ Imaging time in CT is less than 5 minutes
+ CT to visualize moving objects, e.g. heart, lung
+ Increase in detector elements, results in higher imaging speeds / higher resolution
+ In modern CT scanners images consist of 512 × 512 pixels representing the CT number in Hounsfield units (HU): lung -600…-900 HU: here differenciation air or water in the lungs
What is multi-modal imaging and how is it used for nuclear imaging?
= Combination of different modalities = multi-modal
* Each modality is fitted for certain tasks / body parts
* Get complimentary information
Example ▪ CT/PET ▪ MR/PET ▪ OCT/PAM
Detailed CT image of area and in addition PET image (tracer to bind to tumor cells for representation) to be lied above
What are the main characteristics of optical imaging methods? When would you use them?
Light and special properties of photons used to probe tissue, similar to ultrasound but using electromagnetic waves
Measuring of reflection, using interferometry
+ Non-ionizing radiation (visible, ultraviolet, and infrared light) => no exposure to harmful radiation => used for repeated procedures to monitor progression of disease
+ Structural and functional imaging early (detect metabolism changes => markers of abnormal functioning)
+ Usually, non-invasive diagnostic tool
+ Combination with other imaging techniques
Examples: OCT (Optical Coherence Tomography), Endoscopy, Photoacoustic imaging…
What are the main properties of a digitized image or volume?
- Image elements: pixel (2D), voxel (3D)
- Image depth = range of values for image function: 8, 12 or 16 bit (2^8 …)
- Dimension
Representation as: Surface plot, numerical array, visual intensity array
What makes medical image formats special compared to conventional consumer image formats?
Compressed formats (jpg) not very popular as possible loss of diagnostic information
* Big images (>8bit): need to be handled by special image processing programs
* 3D imaging format with details: Slice thickness, Distance between slices
* Information on relevant patient data/image aquisition is not provided in conventional image header formats
* Structured link from patient to image
Example: .tif
Explain the DICOM format: Which information can be stored? How is the information organized? What are application entities and how do they communicate?
“Digital Imaging and Communications in Medicine” = container for medical imaging data (image exchange, visualization and acquisition)
* 3D images of digital radiography, CT, MRI, US, …
* Link image data to medical procedure, image-acquisition parameters (Patient 3D position, sizes, orientations, slice thickness, radiation doses and exposures, image processing filters)
* Encoding of medical data (e.g. patient name, current patient diagnosis)
* Real-world data are viewed by DICOM as objects with attributes
▪ Image height – „Rows“ attribute
▪ Image width – „Columns“ attribute
▪ Image pixel data – „Pixel data“ attribute
▪ Unique identifier
Communication: Captured data can be transmitted and processed between DICOM devices and software – Application Entities (AE)
▪ AEs provide services to each other
▪ Entity: Clearly identifiable object through which information is stored or processed
▪ Relationship between AEs: Service Class Provider (SCP) → Service Class User (SCU)
Which information systems are usually involved in a hospital radiological workflow?
- PACS: Pictures archiving and communication system: consists of medical image and data acquisition, storage, and display subsystems
- HIS: Hospital information system
- RIS: Radiology information system
- MDR: Medical data record software
Tasks: Patient scheduling, billing, financials, reports, worklists, labs, … - HL7 standard: Messaging and document implementation in XML
How can telemedicine be used in the area of medical imaging? Give a scenario.
People living on rural areas need CT scan of lung
Establishment available but no expert in Image evaluation/Diagnosis/Therapy
* Tele-diagnosis: „Point-of-care“ diagnosis and treatment: Ability to test and treat patients rapidly at sites close to where they live
* Tele-consultations: Physician in rural area sends medical information such as CT, MRI scans to specialist in a distant location and receives advice
* Tele-management: Immediate management care to the patient: receive information, monitoring of advised therapy (over two instances) …
What types of digital images are usually used in medical image computing? Give an example for each type.
- Not in medical imaging: Binary image (1 bit)
- Greyscale image (N bit): usually 8 bits: Breast X-ray
- Colour image (n x N bit): types of digital images are usually used in medical image: 24 bits
(8 bits for each colour): PET scan brain - Multichannel image (n x N bit): information outside normal human perceptron range (infrared, ultraviolet, x-ray) => Processed, used in medical research: Divide intensity range of brain CT images
What are point and neighbourhood image transformations and what are they used for?
In order to transform an image for extension of intensity levels or compression of intensity levels
Point image transformations: only intensity and location of single pixel is processed, surrounding pixels are irrelevant
* i.e. histogram, thresholding, grayscale transformation, contrast stretching
Neighbourhood image transformations: area is processed, usually centre pixel and surrounding pixels within specific distance
* Correlation, convolution, morphological operators, spartial filtering
What are image histograms?
Provide frequency of grey values
Horizontal axis → grey values
Vertical axis → number of pixels with grey value
Give information about contrast and colour (dark or bright): Aim: Create an image with equally distributed brightness levels over the whole brightness scale. ➔ Enhance the contrast!
What is spatial (räuml.) filtering and how is a spatial filter defined? Given an example of a spatial filter and what is it used for?
= Neighbourhood image transformation
* Compute a function of the local neighbourhood of each pixel in the image
* Function = filter kernel (called mask, template, window) saying how to combine values from neighbours
Used for:
▪ Image enhancement (denoise, resize, etc.)
▪ Extract information (texture, edges, etc.)
▪ Pattern recognition (template matching)
Example: Moving average 2D = smoothing (compressing data)
How can the Fourier transform be applied to medical images and what is used for?
Fourier transform = weighting function
To be multiplied with integral of sines/cosines of function is not periodic
Convolution and correlation can be done efficiently in the Fourier domain
* process takes a complex-valued function (images are complex-valued functions with zero imaginary component) and returns a complex-valued function
* representing the image as a sum of sinusoids
* Fourier transform considering the phase and magnitude of the complex function
What are image artifacts?
= false images, or parts of images, that do not represent true anatomic structures
* Noise
* Intensity inhomogeneity
* Motion
But: “Higher noise, higher contrast – lower noise, lower contrast”
How can you find edges in an image?
Edges = change in pixel intensity
* Rate of change found by derivative (Ableitung)
* Usually: neighbourhood processing
* Example algorithms: Sobel, canny edge
Find edges in 2D images:
* Smooth first to remove noise
* Calculate image gradient
* Gradient points in direction of most rapid change in intensity
* First derivatives show where the edges are, Zero crossings of second derivatives can be used to extract them
Explain the main steps in the Canny Edge detection algorithm.
(1) Smooth image with Gaussian filter to remove noise
(2) Compute gradient magnitude and angle images
(3) Apply non-maxima suppression to the gradient images
(4) Use double thresholding (hysteresis) and connectivity analysis to detect and link edges
What is the Hough transform and what is used for?
= feature extraction technique (edge detection)
* Locate shapes by linking edge segments in images
* Extract lines, circles and ellipses (or conic sections)
You have to binarize an image based on intensity only. How do you proceed to perform the task by setting a manual threshold? Which approach would you choose for an automated segmentation for this problem?
Manual approach for threshold: Histogram -> Minima = Threshold
Eventually smoothing first to remove noise
Automated approach: Otsu method: Uses grey-value histogram of the given image as input and aims at providing the best threshold in a sense that the overlap between two classes (set of object and background pixels) is minimized
How do active contours work? What are they used for?
= Allow model-based segmentation based on regional parameters
To segment an image where the data are distorted by noise or artefacts
- Represent boundary as a parametric curve
- The curve is associated to an energy function E
- Snake smoothly follows high intensity gradients if they reliably reflect the object boundary
- Internal forces: Designed to keep the model smooth during deformation
- External forces: Move the model towards an object boundary or other desired feature within an image
- Energy minimization based on internal (contour properties) and external energy (image)
What is k-means clustering?
= Method for seeking clusters in data based on observations, e.g. intensity
How can images be represented as graphs? Which graph-based segmentation methods do you know?
Representation can help to find features (i.e. edges)
Image graph:
* Segmentation of the image as weighted, (un)directed graph
* Pixels are nodes
* Edge is connection between pairs of nearby pixels
* Every node is connected to its 4 (or 8) x-y neighbours
Goal of segmentation using graphs: Order the nodes in diverse subsets, where similarity within subset is big and out of subset is low
Examples: Graph cut image segmentation, Watershed segmentation, Cell segmentation, Dijkstra‘s shortest path algorithm (Compute minimum cost path from one seed to all other pixels)
How are segmentation algorithms evaluated? What are useful metrics?
Based on execution time and correctness of the result:
* Accuracy: ability of the method to mirror standard of reference measurements or diagnosis.
* Precision: ability of the method to provide repeatable measurements i.e. low variability due to noise, different acquisition procedure
* Specificity: Ratio of negative cases correctly classified as negative
* Sensitivity: Ratio of true cases correctly classified as true
What is the principal component analysis? What is it used for?
Given example vectors (design matrix)
2. Calculate covariance matrix, and its eigenvector decomposition
3. Represent data in the new space with reduced dimensionality (also Autoencoder)
Reduce the dimensions of data in space => large feature vectors (Get rid of useless/redundant information)
▪ Feature selection (Visualization)
▪ Shape normalization
▪ Model shape
▪ Classification
* To visualize high dimensional datasets
* Each example is represented by a vector containing the values for all features (E.g. 3 for colour vectors)
