NEUROINFORMATICS- DO NOT BID IF YOU DO NO KNOW HOW TO USE NeuCom SOFTWARE!! Must have Computer Science Skills to be able to do it.
sheila18
Assignment_11.pdf
SECMS Paper COMP800, Neuroinformatics, 2018
Assignment 1 Specification School of Computer and Mathematical Sciences
Zohreh Doborjeh KEDRI
Auckland University of Technology
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
1
CONTENT OF THE ASSIGNMENT
5 Summarise the results that you obtained from the data analysis
1 Chose a topic in the area of NI and describe its rationale
2 Present a brief literature review on how this topic has been approached in the past
3 Select a collected and available brain data relevant to the problem
4 Analyse and visualise the data using at least 3 different methods from the NeuCom computational tool or other machine learning tools
2
Assessment: 25% for the final report, 5% for an oral report about the collected data, and 10% for a short 5 min presentation in the class.
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
EEG
fMRI
Massive amount of Spatio-Temporal Brain data
STBD, such as EEG, fMRI, etc. are available in
the area of Neuroinformatics.
Sufficient methods are required for modelling,
learning, visualisation and classification of data in
order to:
Better understating and knowledge discovery;
Analysing of spatio-temporal interaction
between data variables...
3
1. Chose a topic in the area of NI and describe its rationale
The study and understanding of the human brain has become an important focus among researchers, due to the increase of the neurological disorders;
There is a huge amount of collected STBD data which needs to be analysed using proper computational methods;
Spatio-Temporal data is a complex data which is difficult to analyse; Traditional Machine Learning techniques often provide insufficient ability
to represent outcomes of the study when dealing with spatio-temporal data;
Most of the models used, fail to reflect the real nature of the STBD process;
STBD computational models require to be developed to offer prediction possibilities of the data patterns;
Needs of an appropriate biologically model to understand brain data and the process that generates it;
4 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
2. Present a brief literature review on how this topic has been approached in the past
Analysis of Neurodegenerative
Diseases
E.g. Alzheimer’s
Disease
Spatio-temporal brain data (STBD)
Cognitive impairment and memory loss is one of the first symptoms
Study the ability of a Machine Learning tools for classifying STBD
“As the case study, you can choose any type of brain data recorded from either patients or healthy subjects”
5
Example:
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
3. Select brain data relevant to the topic already collected and available
I. Select available spatio-temporal data II. Describe the data III. Describe where the data was collected IV. Describe how the data was collected V. Describe the names and the meaning of the variables: what are the spatial features and the temporal features of the data?
VI. Describe data’s previous use
6 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
7
I. Select available data
http://www.cs.colostate.edu/eeg/eegSo ftware.html#keirndata
Example: cognitive EEG data
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
8
Z.A. Keirn master thesis Purdue University 1988
II. Describe the data
III. Describe where the data was collected
IV. Describe how the data was collected
V. Describe the names and the meaning of the variables
Example: cognitive EEG data
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
9
Cited 139
Cited 318
VI. Describe data’s previous use
Example: cognitive EEG data
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
10
VI. Describe data’s previous use
Example: cognitive EEG data
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
BCI Competition IV: http://www.bbci.de/competition/iv/
BNCI-Horizon: http://bnci-horizon-2020.eu/database/data-sets
Physionet: https://physionet.org/
Available EEG data
4. Analyse and visualise the data using NeuCom
I. Download the NeuCom Student V.0.919 at: https://kedri.aut.ac.nz/areas-of- expertise/data-mining-and-decision-support/neucom
I. Prepare the raw data set for NeuCom, as ordered vectors of real value data;
II. Use at least 3 different methods from the NeuCom computational tool to analyse your data;
12 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
I. Download the NeuCom Student
13 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
A Case Study on Moral Dilemma Judgement
A moral dilemma is a conflict in which you have to choose between two or more
actions and have moral reasons for choosing each action.
For example, the trolley dilemma and the foot bridge dilemma are examples of moral
dilemmas
Impersonal Dilemma (ID) Personal Dilemma (PD)
14
10 subjects
20 EEG channels
30 trial dilemma
15
II. Prepare the raw data set for NeuCom
Feature 1 Feature 2 ..... Feature n Class Label
Sample 1
Sample 2
.....
Sample n
Save data into a .txt file
16
Class label should be positive integer
Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
Data preparing for NeuCom
Ch 1 Ch 2 Class label
Sample 1 Mean of 512
data points
1 or 2 or 3 or 4
Sample 2 1 or 2 or 3 or 4
…
Sample 40 1 or 2 or 3 or 4
Ch 1 Ch 2 … Ch 20
…
.
.
.
… Average
Yes answer to trolley dilemma class1 No answer to trolley dilemma class 2 Yes answer to bridge dilemma class 3 No answer to bridge dilemma class 4
Time point 1
Time point 512
EEG data variables
17
Data visualisation in NeuCom
18
Feature selection with NeuCom
• We want to select the most appropriate variables when creating a computational model;
• There are different methods for feature selection; we will use two filtering methods available in NeuCom: correlation and signal-to-noise ratio (SNR).
SNR correlation 19 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
Feature selection: Signal-to-noise ratio SNR
SNR is computed for 20 variables Variables are ranked based on SNR values
20
Class Distribution F7, P5 and T5
21
III. Use different classification methods from the computational tool NeuCom
Cross validation: Classification methods obtained with LOOCV (leave one out cross validation) using different methods, Super Vector Machine (SVM) Multi-layer Perceptron (MLP) Multi Linear Regression (MLR) Cross validation mode: Leave one out cross validation (LOOCV)
One sample for validation, 39 samples for training
Methods SVM MLR MLP
Overall
classification
Accuracy
Normalized 90% 80% 84%
Non-Normalized 90% 80% 78%
22 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
Interpretation: Activated brain regions
• Frontal lobe
• Parietal lobe
• Temporal lobe
23
5. Summarise the results that you obtained from the data analysis
What do you want to do further with this data (if at all)?
What have you learnt from the data that you selected and analysed via NeuCom ?
24 Assignment 1 - Neuroinformatics (COMP800) [email protected] www.kedri.aut.ac.nz
