AI Lesion detection Flashcards
What is Artificial Intelligence?
Mimicking the intelligence or behavioural pattern of humans or any other living entity
What is Machine Learning?
A technique by which a computer can LEARN from DATA without using a complex set of different rules. This approach is mainly based on training a model from datasets
What is Deep Learning?
A technique to perform machine learning inspired by our brain’s own network of neurons
Relationship of AI, Machine Learning and Deep Learning
AI>Machine Learning>Deep Learning
Types of Machine Learning
In the past SVM (support vector machines), now deep learning (multiple hidden layers), U-Net
Which network will you use to image segmentation?
U-Net: Convulutional Network for Biomedical Image Segmentation
Example of use of segmentation
White matter spots in MS
Types of learning ?
Supervised : labels Classification - Regression
Unsupervised : No labels, Clustering
How does unsupervised learning work
The machine will try to find similarities and put those input data in 2 or more clusters
Supervised ML PROS and CONS (!!!)
\+ smaller number of data \+ can classify a disease - need EXPERT annotated data - what is gold standard? - will not detect unexpected diseases/conditions
Unsupervised ML PROS and CONS (!!!)
+ can detect unexpected or unknown categories
+ no annotation needed
- larger dataset
- detected clusters/patterns not necessarily equal to diseases
How do you narrow the input data ?
You select features
What is the problem with features selection?
Double Dipping
Cross Validation technique (!!!)
Discriminate the entire data in 10 parts use 9 for training and 1 for testing
What is the best way for a training vs testing dataset? (!!!)
Completely different training and testing data sets
