Introduction to Medical Image Analysis - compute.dtu.dk - week8.pdf · DTU Compute 3 DTU Compute, Technical University of Denmark Introduction to Medical Image Analysis 21/3/2018

DTU Compute

Introduction to Medical Image Analysis

Slides adapted from Jens E. Wilhjelms lectures

Rasmus R. PaulsenDTU Compute

[email protected]

http://www.compute.dtu.dk/courses/02511http://www.compute.dtu.dk/courses/02512

mailto:[email protected]://www.compute.dtu.dk/courses/02511

DTU Compute

21/3/2018Introduction to Medical Image Analysis2 DTU Compute, Technical University of Denmark

Lecture 8 X-ray imaging and CT scanning

9.00 Lecture

Exercises

12.00 13.00 Lunch break

13.00 - Exercises

DTU Compute


What can you do after today? Describe the basic use of X-rays Describe the X-ray as an electromagnetic wave Explain the basic technique in an X-ray tube Estimate how simple materials will look on an X-ray Use Lambert-Beers law to compute material attenuation Compute the attenuation of simple non-homogenous

materials Describe the concept of tomographic reconstruction Describe the relation between Hounsfield units and tissue

types Compute a CT slice using a simple and idealised setup

DTU Compute


X-ray imaging The most used form of

medical imaging Simple Cheap Fast Radiation

DTU Compute


The history of X-ray Discovered by chance

DTU Compute


Wilhelm Conrad Rntgen German physics professor Experimented with a Crookes

tube Discovered that an unknown ray

could be captured on photographic plates

Named them X-rays Other call them Rntgen-rays

Had no idea they were dangerous

Made an X-ray of his wife's hand First medical X-ray

Wilhelm Rntgen's first medical X-ray, of his wife's hand, taken on 22 December 1895

Crookes tube

DTU Compute


Quick popularity X-ray became popular

extremely fast Shoe fitting Examine your bones in coin

machines Wedding pictures

X-ray clinics in small normal apartments

DTU Compute


Dangers People started to realise that

exposure to X-rays could be dangerous

Hands of X-ray pioneer Mihran Kassabian

DTU Compute


Brief history of X-ray and CT 1895 Wilhelm Conrad Rntgen discovered X-ray 1896 Within a week X-ray was not known worlwide 1896 GE and Siemens begin selling X-ray equipment 1904 Dangers of radiation are described 1956 First reconstruction algorithm is described 1958 First prototype CT scanner without computer 1968 Hounsfields method for CT patented 1972 Hounsfields method for CT demonstrated in the US 1989 First spiral CT scanner enters the market

DTU Compute


X-rays as electromagnetic radiation It has a frequency f

(v is used for frequency in notes) Measured in Hertz [Hz]

It has a wavelength (lambda) Measured in meters [m]

It has a speed The speed of light c

http://upload.wikimedia.org/wikipedia/en/1/1a/Wavelength.svg

DTU Compute


Wavelength 10 pm < < 10 nm

pm = picometer = 11012 m nm = nanometer = 1109 m

DTU Compute


Production of X-rays Electrons are accelerated using a cathode Some hit the anode (heavy metal target) Slowed down in the anode material

Generating heat A small part of the energy is transformed

to X-rays The electron comes very close to the

nucleus Electromagnetic interaction causes a

deviation of the trajectory The electron looses energy and an X-ray

photon is emitted.

DTU Compute


Electron volts 1 eV is the energy increase that an electron

experiences, when accelerated over a potential difference of 1 V.

In medical imaging 20 keV < E < 150 keV

keV = kilo-electron-volts

Cathode Anode

- +100 kV

DTU Compute


X-ray tube

The anode rotates to avoid over heating

http://upload.wikimedia.org/wikipedia/en/9/9b/Xraytubeinhousing.png

DTU Compute


X-ray tube

Jackson X-ray tube, 1896. Modern rotating anode tube

http://upload.wikimedia.org/wikipedia/commons/5/51/Rontgenbuis-draaianode.jpg

DTU Compute


Full X-ray system

DTU Compute


X-ray film Until recently real film was

used Being replaced by digital

radiography The first step was to use

phosphor plates that was later scanned

DTU Compute


Contrast in X-ray images Some materials absorb more

X-rays than others We see the X-rays that got

through Dark area high radiation

Air Soft-tissue Fat

Bright area low radiation Metals Bone

Scanned X-ray film

DTU Compute


X-ray attenuation (dmpning) X-rays hits an object and

travels through it is the intensity at the

entrance is what is left on the

other side after a length of x

The rest disappears in several different ways

Computed using Lambert-Beers law

I (x)I 0

I 0

I (x)

x

DTU Compute


Lambert-Beers law Materials can be described

using a linear attenuation coefficient How much do the material

dampen X-rays High coefficient

Metal and bone Low coefficient

Soft tissue and fat

: Linear attenuation coefficient

0 1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Distance in object (cm)

exp(

x)

High Low

I (x)I 0

x

DTU Compute


Non-homogeneous material Non-homogeneous not the same everywhere

x1

x2 x5x3

x4

I 0

I 0

I 0

I 0I 1

film

I 2

I 3?

DTU Compute


X-rays in practise Several parameters to tune

KV Exposure time Etc

Settings dependent on body part Radiologists are trained for this Still a lot of human knowledge in

taking good X-rays

DTU Compute


Uhomogentmateriale A) 1

B) 2C) 3D) 4E) 5F) 6

Chart1

1

1

1

1

1

1

-

-

-

-

-

-

Sheet1

A1

B1

C1

D1

E1

F1

DTU Compute


CT scanning

DTU Compute


Invention of Computed Tomography 1972: G.N. Hounsfield, scientist in Middlesex, England

Announced computed axial transverse scanning Presented cross-sectional images of the head showing tissues

inside the brain as separate structures of gray matter, white matter, cerebrospinal fluid, and bone

Pathologic processes such as blood clots, tumors, and infarcts could be easily seen

Dr. Hounsfield's discovery completely revolutionized the practice of medicine: Structures inside the human body that had never been imaged before, could now be visualized.

1979 Nobel Prize in Medicine

DTU Compute


Development of CT Early 70s : Several minutes to acquire single slice Today: Less than a minute for a full body scan

DTU Compute


Basic principle of CT Super Sudoku It is assumed that the object

consists of small cubes of homogeneous material

We want to find the linear attenuation coefficient in each cube

I 0

I 0

I 0 I 0

DTU Compute


Quiz? What is A, B, C, D? Can be formulated as a

system of equations 4 unknowns 4 equations

A

C

B

D

A+B = 13

C+D = 10

B+D = 11A+C = 12 DISCLAIMER: just an illustration not solvable

DTU Compute


Basic principle of CT Super Sudoku It is assumed that the

object consists of small cubes of homogeneous material

We want to find the linear attenuation coefficient in each cube

We measure the radiation that comes through

I 0

I 0

I 0 I 0

DTU Compute


Basic principle of CT Super Sudoku What is ?

I 0

I 0

I 0 I 0

x

DTU Compute


Basic principle of CT Super Sudoku Four equations Four unknowns Can be solved directly

I 0

I 0

I 0 I 0

x

DTU Compute


Real CT More than 4 pixels per slice 512 x 512 pixels Many projections Enormous system of

equations Not solvable by direct

methods

DTU Compute


Hounsfield Units Instead of using a linear

attenuation coefficient Dimensionless Calibrated

Air: -1000 Water: 0

Linear attenuation coefficient in voxel

Different equation used in book (eq (1))

DTU Compute


Hounsfield Units

DTU Compute


Real CT scanning

DTU Compute


Real CT scanning Many projections Advanced reconstruction

algorithms Filtered back projection Radon transform

DTU Compute


Different versions of CT machines

Rotating detectors

DTU Compute


Rotating source stationary detectors

DTU Compute


Modern CT

High Speed Advantage GE Spiral CT

DTU Compute


Multi-slice CT

128 slice scanner

DTU Compute


What can be seen on the CT image

Photograph of cryosectioned head CT scan

DTU Compute


Uhomogent MaterialeA) 1B) 2C) 3D) 4E) 7

0

11

3

6

0

A B C D E

Chart1

0

11

3

6

0

Sheet1

A0

B11

C3

D6

E0

DTU Compute


Todays exercise Automatic algorithm to locate the spleen Mandatory hand in for both 02511 and 02512 Several methods

Pixel classification Morphology BLOB analysis

Training data Validation data why?

DTU Compute


Bonus ChallengeThe Spleen Challenge I have three secret images

Locate the spleen on these Upload your Matlab code on CampusNet

The Spleen Challenge Input: file name of DICOM file Output: Binary image with the Spleen

I will run your function on the images and compare the results to the gold truth

The winner is the algorithm that is closest to the ground truth Give your algorithm a funky name

Deadline 4. April

DTU Compute


Previous highlights

DTU Compute


DTU Compute


Epic_SpleenMasterOrz9000Extreme_supreme_premium_Deluxe_Edition

DTU Compute


DTU Compute


What can you do after today? Describe the basic use of X-rays Describe the X-ray as an electromagnetic wave Explain the basic technique in an X-ray tube Estimate how simple materials will look on an X-ray Use Lambert-Beers law to compute material attenuation Compute the attenuation of simple non-homogenous

materials Describe the concept of tomographic reconstruction Describe the relation between Hounsfield units and tissue

types Compute a CT slice using a simple and idealised setup

DTU Compute


Next week Geometric transformations

DTU Compute


Exercises

?

Introduction to Medical Image AnalysisLecture 8 X-ray imaging and CT scanningWhat can you do after today?X-ray imagingThe history of X-rayWilhelm Conrad RntgenQuick popularityDangersBrief history of X-ray and CTX-rays as electromagnetic radiationSlide Number 11Production of X-raysElectron voltsX-ray tubeX-ray tubeFull X-ray systemX-ray filmContrast in X-ray imagesX-ray attenuation (dmpning)Lambert-Beers lawNon-homogeneous materialX-rays in practiseUhomogentmaterialeCT scanningInvention of Computed TomographyDevelopment of CTBasic principle of CT Super SudokuQuiz?Basic principle of CT Super SudokuBasic principle of CT Super SudokuBasic principle of CT Super SudokuReal CTHounsfield UnitsHounsfield UnitsHounsfield UnitsReal CT scanningReal CT scanningDifferent versions of CT machinesRotating source stationary detectorsModern CTMulti-slice CTWhat can be seen on the CT imageUhomogent MaterialeTodays exerciseBonus ChallengeThe Spleen ChallengePrevious highlightsSlide Number 50Slide Number 51Epic_SpleenMasterOrz9000Extreme_supreme_premium_Deluxe_EditionSlide Number 53Slide Number 54What can you do after today?Next weekExercises

Introduction to Medical Image Analysis - compute.dtu.dk - week8.pdf · DTU Compute 3 DTU Compute, Technical University of Denmark Introduction to Medical Image Analysis 21/3/2018

Documents