MemoCard lernen mit System

Voxel & Pixel

- Smallest information unit in a 3D image

- It contains scalar values sorted by components

- Pixel in 2D

How many levels can we have for 1 byte/ 2 bytes?

- 256 for 1 byte (i.e., 28-1)

- 65536 levels for 2bytes (216)

- 0 à black ; 255/ 65535 à white

What is DICOM?

Digital Imaging and Communications in Medicine (DICOM). Standard for storing, handling, printing, transmitting information in medical data

What is the orientation TAG describing? DICOM?

The orientation TAG is described via two cosines directions. Setting/reading the image orientation is crucial to have a good readout of pixel data in world coordinates. -

Pros of landmark-based methods:

- Compact representation

- Computationally efficient

Automated landmark detection:

- Manual landmarks can also be used to evaluate automated methods developed to perform landmark detection, or to evaluate the accuracy of registration techniques.

- Slicer3D software and its annotation module.

Take home message of voxel-based representation:

- Home-take message here is to think whether your clinical application requires the level of detail provided by voxel-based representations. Think of the computational and memory requirements of the clinical scenario you are dealing with.

Curves landmarks

Curves are not very popular but can be very descriptive in certain cases

E.g., brain sulci and gyri

Landmark based methods:

- On the other end of the spectrum regarding the amount of represented information (comparison with voxel-based rep.)

- Store very little data, corresponding tp location of a set of (anatomically meaningful or otherwise chosen) landmarks

- Landmarks: Points, curves, surfaces

When are meshes a good choice?

Good choice when developing a lightweight but rich shape representation.

DICOM Network Protocol

- Search studies

- Restore

- Control treatment

- Scheduling, reporting, sharing workload

What is the difference between study and series?

Study: patient A has a CT scan

Series: patient A has a CT scan + MRI + Xray --> set of studies

Why is the description of a voxel/pixel in real units so important?

Because image data is used to locate structures, pathologies, plan treatment (e.g. brain tumor surgery or radiotherapy), measure volumes, growth (of a disease), etc.

Which color model is mostly used?

Medical Images mostly use gray scale --> one channel

Voxel based representation

- These representations are derived directly from images

- Assign a label to each structure in the image

Where is hypo intensity and hyper intensity on the grayscale?

- Hypo intensity --> black

- Hyper intensity --> white

What DICOM combines?

- Combines image data with meta-data (patient data, acquisition parameters)

What does the pixel depth mean?

The pixel depth, or bit depth, determines the range of values that a particular raster file can store, which is based on the formula 2^n where n is the pixel depth.

Cons of meshes:

Not dense (may lack detail)

Pros of voxel-based representation

- Directly related to the source image

- Dense resolution

Cons of landmark-based methods:

- Incomplete information (how to place landmarks)

- Need accurate localization of landmarks (not evident in 3D)

Pros of meshes:

- Topological flexibility (just by changing the connectivity function)

- Simple and easy to manipulate

- There exist simplification and refinement methods to reduce the number of nodes

Why is DICOM used?

Enables integration of scanners, servers, workstations, printers, and other hardware from multiple vendors into a picture archiving and communication system (PACS)

What is the most common mesh representation?

Triangulations

What would be the TAG value in this image? (Here a image would be needed)

- There is a 45 degrees rotation (and cos(45)=0.707) - The cosines are then 6 values, and the TAG values would llike this: (0.707,-0.707,0, 0.707,0.707,0) the negative -0.707 corresponds to the component of the (new) LR axis moving in the negative direction of the (old) RA axis.

What is the PACS system?

At the core of DICOM resides the PACS system, which is the computational unit at hospitals and clinics where imaging information is stored, and used to handle patient information related to imaging scans.

Three elements that uniquely describe a voxel position in real units (E.g., mm, cm)

- Image Origin; must not always be (0,0,0); (0,0)

- Image dimension

- Voxel/Pixel spacing (voxel/pixel) size

Point landmarks

- Can be used to derive measures (e.g. distances between two points), drive an elastic thin-plate representation (middle bottom) characterizing shape changes in the human skull, etc.

- It is crucial to select landmarks that describe the shape, and it can be time consuming to annotate data

How to compute distance maps?

To compute distance maps is very simple:

- for each voxel inside an structure, compute the closest boundary point and assign its distance as voxel value.

Famous mesh algorithm and its properties?

- Marching cubes algorithm

- Based on defining sets of intersection cases of a surface and a voxel

- Set of 15 configuration suffice covers all possible combination of background/ forground voxels.

For what are distance map used?

They are commonly be used as a probabilistic map, guide registration of images (registering distance maps, statistics of shapes, etc.

Why is pixel depth important for medical images?

Pixel depth is then important for correct reading and storing of pixel data, but also for computational and memory consumption efficiency

Name different methods for image representation:

- Voxel-based representations

- Landmarks (points, curves, Surfaces)

- Meshes

- Distance maps

Cons of voxel-based representation

Very large amount of data

- Time consuming

- Complex to handle (e.g., interaction with the structure)

Structure of DICOM file format

A DICOM file, with extension .dcm, describes a single slice. It includes a header and data sets. The header has a so-called anarchic zone, left to the user, and it’s followed by the four chars D, I, C and M. The data sets follow the displayed structure. TAGs are predefined and part of the standard. In the next slides we will focus on some specific tags.

Pros and cons of distance maps???

- mümmer no usesueche

What can DICOM handle?

- Hardware includes vendor type, specific parameter values (e.g. energy values for a CT scan), etc. Treatment and patient data are crucial as well to follow-up patients and to inform treating doctors.

- The second component (bottom) deals with the network protocol. Searching, querying, scheduling, etc. are part of this protocol

What are the parameters saved in the DICOM File format?

- Hardware, imaging à Model, Software

- Treatment-related à contrast agent

- Patient data --> Name, age, date of birth

Are distance maps implicit or explicit representation?

Distance maps are an implicit representation of the anatomy.

Distance maps:

- Shape is defined in a relative way

- Each point inside of the shape is non-zero

- Each point on the boundary of the shape is zero

- Distance maps are then an implicit way of representing shape.