Scribd
Upload
43 views29 pages

Lecture 11

This document discusses image features for medical image analysis. It defines an image feature as a measurable piece of data in an image that is unique to a specific object, such as color, shape, or edges. Good image features should discriminate one image or object from others. Examples of image features that could be used to classify images of a motorbike and horse are discussed. The role of features in machine learning is to transform raw image data into feature vectors that capture the characteristics of objects to allow algorithms to learn. Common features that could be used to classify horses and zebras are discussed as color and stripes.

Uploaded by

ahsan puri
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
43 views29 pages

Lecture 11

This document discusses image features for medical image analysis. It defines an image feature as a measurable piece of data in an image that is unique to a specific object, such as color, shape, or edges. Good image features should discriminate one image or object from others. Examples of image features that could be used to classify images of a motorbike and horse are discussed. The role of features in machine learning is to transform raw image data into feature vectors that capture the characteristics of objects to allow algorithms to learn. Common features that could be used to classify horses and zebras are discussed as color and stripes.

Uploaded by

ahsan puri
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 29

IN THE NAME OF

ALLAH THE MOST


MERCIFUL AND
BENIFICIENT
MEDICAL IMAGE
ANALYSIS
DR. ZOBIA SUHAIL
PHD IN MEDICAL IMAGE PROCESSING
A B E R Y S T W Y T H U N I V E R S I T Y, WA L E S , U K ( 2 0 1 9 )
MEDICAL IMAGE ANALYSIS

Overview of Previous Lecture:

• Canny Edge Detection


MEDICAL IMAGE ANALYSIS
Image Features:
Image feature is a measureable piece of data in an image which is unique to the specific
Object.

It can be a:

Color
Shape
Edges

The main role of features in computer vision(and not only) is to transform visual information into
the vector space. 
This give us possibility to perform mathematical operations on them, for example finding similar
vector(which lead us to similar image or object on the image)
MEDICAL IMAGE ANALYSIS
Image Features:

Good Image feature should discriminate an image from other images / an object in an image
From the object in other images.

Image features are also called as image descriptors and normally consists of more then
One values that will be used as image identifier.
MEDICAL IMAGE ANALYSIS
Image Features:

Examples:
MEDICAL IMAGE ANALYSIS
Image Features:

Examples:

What would be a good feature that can be used in these two images for discrimination?
MEDICAL IMAGE ANALYSIS
Image Features:

Examples:

What would be a good feature that can be used in these two images for discrimination?
What if we use WHEEL as a feature
MEDICAL IMAGE ANALYSIS
Image Features:
Examples:

What would be a good feature that can be used in these two images for discrimination?
What if we use WHEEL as a feature  Yes WHEEL is a good feature to discriminate motorbike
And horse image.
MEDICAL IMAGE ANALYSIS
Image Features:
Examples:

Can WHEEL still be used to discriminate between the images of horse and
Zebra?
If NO WHY ???? And what could be a good feature then?
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

In machine learning, feature is any measurable characteristics of any thing that we want to observe.
Features are used as input to our machine learning algorithm that in turns provide us
With some prediction / classification of the under observation phenomena.

Example:
If we want to measure the prize of car: The input features would be:
Model
Manufacturer
Color
Etc…

A good feature vector results in good prediction / classification and hence improve the overall
Efficiency and accuracy of machine learning algorithms.
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
In machine learning projects, we want to transform the raw data (image) into
a features vector to show our learning algorithm how to learn the characteristics of the object.
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
In machine learning projects, we want to transform the raw data (image) into
a features vector to show our learning algorithm how to learn the characteristics of the object.

IMAGE  FEATURE VECTOR


MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
In machine learning projects, we want to transform the raw data (image) into
a features vector to show our learning algorithm how to learn the characteristics of the object.

IMAGE  FEATURE VECTOR


1D array if image features
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
In general feature extraction means to recognize general features from multiple
Images of same type, instead of just taking features from one image.

Like in previous example, the extracted features are for only one particualr type of bike.
Whereas if we talk about the general features for all type of motorbikes, it will be something like
All the motorbikes have a common feature that they have a wheels, where a wheel is something
That have circular shape with some patterns in it.
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

So if we look at 1000 of motorbikes and extract common wheel feature it will be like:

Image Source: https://freecontent.manning.com/the-computer-vision-pipeline-part-4-feature-extraction/


MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

Machine Learning model relies a lot on features provided as input.

It means good feature extraction is very important in order to build a powerful


Machine learning model.

Example:
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
Example:

What would be good features to classify a horse and zebra….


MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
Example:

What would be good features to classify a horse and zebra….


SPEED! 
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
Example:

What would be good features to classify a horse and zebra….


SPEED!  On Average both Horse and Zebra run very fast, so speed can’t be a good
Feature to discriminate a horse and a zebra
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
Example:

What would be good features to classify a horse and zebra….


Height! 
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
Example:

What would be good features to classify a horse and zebra….


Height!  Zebras are around 1.2-1.5 meters in height where a horse can grow to 1.8 meters. 
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:
Example:

What would be good features to classify a horse and zebra….


Colour!  Zebras have black and white stripes, whereas horses are normally
Have symmetric colour. 
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

Characteristics of good feature:

1. Identifiable

2. Easy to track and compared

3. Consistent

4. Still visible in noisy images


MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

How to extract image features:

There are two ways of getting features from image, first is an image descriptors
(white box algorithms) also called as hand-crafted features,
second is a neural nets(black box algorithms) also called as automatic feature extraction.
We will be working with the first one approach of feature extraction i.e white box algorithms
/ Traditional Machine Learning using hand-crafted features.
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

Hand-Crafted Features:
MEDICAL IMAGE ANALYSIS
Image Features in Machine Learning:

Automatic Feature Extraction:


MEDICAL IMAGE ANALYSIS
Image Feature:

End of Lecture: Image Feature _1:

You might also like