Assignment 2

Due Today !

Dr. Rudolph Pienaar

Fetal Neonatal Neuroimaging and Developmental Science Center

3/1 at 12pm

We are ready!

Everyone will present at least 1 paper this semester

You will work in a group of 3!

2-3 papers per Journal Club

Sign-up now!

Images are 2D Arrays!

it's all numb3rs but you need to figure out the conventions...

Coding time!

Image Filtering

Input

Output

Kernel

Thresholding

Input

Input > 200

Count

Color

everything < 200 set to 0

everything < 247 set to 0

Histogram

How to segment the tumor?

Binary segmentation mask

Segmentation

means dividing up an image into different structures

foreground vs. background

Label Map

each identified structure has a unique ID

Datatype?

Comparison of Segmentation Techniques

Automatic Segmentation is a hard problem!

so much variability across subjects!

Semi-Automatic Segmentation

user places seeds

Manual Segmentation or Expert Segmentation is the "Groundtruth"

extremely time-consuming

2 Experts worked on this for 6 months

Registration

Registration

a

 

 

b

 

 

c

Registration

Registration

mapping one subject to a standard space

mapping one image modality to a standard space

once in standard space, we can apply knowledge or compare

Atlas-based Segmentation

Registration / Alignment

Images are 2D Arrays!

2D Image Modalities

Ultrasound

X-Ray

What about CT, MRI, EM?

Image Volumes

Fetal MRI

2D

3D

Marching Cubes creates 3D Meshes

Bill Lorensen 1987

Label Maps

R.I.P.

Slice-based Volume Rendering

Volume Rendering with Ray Casting

Volume Rendering with Ray Casting

Cinematic Volume Rendering

DICOM

Study

Patient

Series

2D Images

Picture archiving and communication system (PACS)

From DICOM...

...to .NRRD, .NIFTI, and others!

ton of converters available

CS666 Lecture 15

By Daniel Haehn

CS666 Lecture 15

Slides for CS666 Biomedical Signal and Image Processing at UMass Boston. See https://cs666.org!

  • 66