cerebro - engineering.tamu.edu · App Goals-2017 app ts er device via r . Title:...
Transcript of cerebro - engineering.tamu.edu · App Goals-2017 app ts er device via r . Title:...
Summ
ary
The primary objectives to upgrading Cerebro this sem
ester were:
Expanding the data analysis to run in a more generalized w
ay, i.e., the ability to w
ork with data not sorted in a specific w
ay.
Initializing a headset prototype (prototyping both the design and the electronics) in order to accurately obtain biosignals.
Updating the app in order to make it user friendly and provide versatility in
emergency response options.
Initial Approach
Create a design that will practical, functional, m
oderately aesthetic, and non-invasive
Compare and contrast different m
aterials
Considerations included durability, flexibility, strength
Examples: ABS, Polyjet, N
injaFlex and Tango
3D printing Research
Types of printers available on campus, m
aterials available
EEG biosensors w
earable electronics in the market
Design Specifications
Initial design was developed using SolidW
orks software
Advantages:
Allows for placem
ent of electrodes to be moved to necessary spots
Thinner piece around ears for comfort
Comfortable, thin headband
First Prototype
●Used 3-D printer:
○EDEN
260 V with dim
ensional accuracy of 7 microns and a building
time of g^3 per 8.2 m
in
●M
aterial: Tango, a rubber-like material
Spring 2017 -Goals
●H
ouse electronics inside the prototype
●Create adjustable m
echanism for different head sizes
●Im
prove design:
○Com
bine materials to provide a headset structure
○Design spring m
echanism for housing of dry electrodes
○Incorporate a battery w
ith a useful life of at least 5 hrs
Overview
of the Semester
Figuring out conventionally placed electronics
Discerning the Necessary Electronics to be H
oused
Exploration of Types of Electrodes
Researching Capacitive Electrodes
Initiating the Prototype
Electronics in Conventional Headsets
Electrode with N
oise Reduction Module
Application Specific Integrated Circuit
Analog-Digital Converter
Bluetooth Module
What Parts are N
eeded?
Rephrasing the Question: W
hat parts does the algorithm take care of?
Algorithm takes care of Application Specific IC, therefore no Signal Processing
required in Electronics.
What m
ust be covered in Electronics:
Electrode
Noise Reduction
Analog Digital Converter
Serial Transfer to Device (Bluetooth)
The Quest for the Ideal Electrode
The Capacitive Electrode is Ideal because:
Relatively new technology, scope for an additional claim
in patent
Doesn’t require direct contact
More com
fortable
Eliminates the effects of artefacts
Downsides
Relatively new technology, no com
mercial electrodes available to play w
ith
Very nuanced and precise construction (Layered Electrode Surface)
Initializing the Prototype
Capacitive Electrodes require more tim
e, so we decided to m
odel the prototype based on dry or w
et electrodes
Start with a basic m
odel and upgrade in Phases
Using Wet Electrode
Using data acquisition electronics based on Dr. Vu’s EKG system
Other Prototype Initiatives
Serial Protocol Output (via USB)
Using Teensy Microcontrolling Unit
Using standard 9600 baud
Transferring 16-bit data at 256 Hz sam
pling rate
Figure: An apparatus to utilize
Teensy AD
C converter using an
Oscilloscope.
Plans for Spring 2017
Upgrade prototype from conventional electrodes to Capacitive Electrode
Better Integration with H
ardware (electronic housing, ergonom
ic factor)
Integration with Sm
artphone Application
Executive Summ
aryG
ained understanding of random forest m
odels & m
achine learning
Gained understanding of relevant features (R
elative Phase/P
hase Synchronization,
Snow
ball PS
D, S
nowball P
SD
Phase A
ngle)
Trained & Tested A
lgorithm on single patient (Tested patient 1 -Trained on files 10-
24 and tested on files 1-5 for patient)
Autom
ated patient input in order to train and test on more than one patient at a tim
e
Autom
ated annotation creation in order to allow for patients to be selected in a
random order
BackgroundData pulled from
PhysioBank ATM via
physionet.org
Close to 1000 hours of EEG data from
24 patients at the Children’s H
ospital of Boston (data collected by M
IT)
Analyze T7-FT9 & P8-O2 voltage potential
differences
Critical features for algorithm are relative phase,
snowball of pow
er spectral density (PSD), and snow
ball of PSD phase angle.
Training data structure: 1 row per second-25 colum
ns (1 for annotation class label, 8 for each of the 3 feature types)
Random
Forest
Ensemble m
achine learning for event classification
The algorithm is a forest because it utilizes a large num
ber of decision trees
The algorithm is random
because each tree is trained on a random subset of
the training data. Also, each binary node split selects a random set of
predictor features.
Each tree gives a binary “yes/no” classification, and the average of all tree classifications can be used to predict an event
Our R
andom Forest
Utilizes Matlab TreeBagger()
Function
400 decision trees used
If greater than 80 trees (20%) predict
a seizure, the model predicts a
seizure
Power Spectral D
ensity (PSD)
A method of representing pow
er as a function of frequency
Used in both Snowball of PSD, and
Snowball of PSD Phase Angle
features
Snowball
Making sm
all changes currently can lead to large changes in the future (causing an event)
Changes that have occurred in the features are added up
Snowball of PSD
& Snowball of PSD
Phase Angle
Benefit of Snowball Effects: Analyzes
1 channel (T7-FT9) instead of two
as seen in the RP features
Phase Synchronization for Snowball
of PSD Phase Angle: Utilize Hilbert
transform to break PSD signal
down into real and im
aginary parts. Extract phase angle.
Relative Phase Feature (R
P)
Also uses Hilbert transform
Unlike the snowball effects, this
feature captures the differences betw
een two
signals (channels T7-FT9 & P8-O
2)
According to NBT, m
& n equal 1 for the relative phase equation, but m
oving forward w
e should test this hypothesis to confirm
optimal
values
Training & Testing Code on Patient 1Understand how
the algorithms w
ere implem
ented by previous team
Analyzed Patient 1
First the algorithm w
as run to create a set of Train and Testing data
Next, annotations w
ere input based on provided information. Finally, Random
Forest algorithm
was run to for sezuire predciton
Accuracy: 0.9645
Precision: 0.1051
Sensitivity: 0.0342
Specificity: 0.9915
Patient Input Autom
ation
Allow for grow
th of number of Patient Files to be selected w
hen running algorithm
Code allows for developer to select w
hich random Patients w
anted to be analyzed
The algorithm w
ill then stitch together all of the files inside each of the Patient Files
This data can then be run through feature extraction
Annotation A
utomation
In addition to stitching together patient files, annotations times needed to be autom
ated
.CSV files containing Patient, Files, Start and Stop Times
Using loop indexed the files to the appropriate start and stop time of each seizure
Accounted for the previous time elapsed from
each file
Plotted Results (Patients 3 & 23)
The plot to the right demonstrates
our fully automated code
correctly plotting the times of the
9 seizures between patients 3
and 23 (plotted to demonstrate
new capabilities of both m
ultiple patient input autom
ation & annotation creation autom
ation)
Total of 9 in both patients 3 and 23 data
Spring 2017
Have received access to the ADA Supercom
puter
Can’t train algorithm on all patients w
ithout supercom
puter; data too large for comm
ercial com
puters as Matlab crashes
Create Test and Train data with greater num
ber of patients
Run RandomForest fully w
ith 400 trees
Integrate with App Team
for access on Server, and configure code to w
ork with live data
Overall A
pp Goal
To create a user interface that provides a connection between the live ECG
data collected and the seizure detection algorithm
that will notify the patient
of an oncoming seizure
Semester A
pp Progress
Improved user interface from
last semester
Visual appeal, better organization
Created template to receive bluetooth data from
EEG device in progress
Gained understanding of java, app developm
ent, and android studio
Explored comm
unication between android and online servers
Features-“Alarm
” tab (Hom
e Screen)
5 bottom tabs used for navigation betw
een features
This will show
any active seizure predictions (after user receives initial alert)
This will show
all previous alarms
Provides advice on what to do during a seizure (next slide)
Allows user to set alarm
preferences (next slide)
Features-“Alarm
” tab
“Prepare”●
Provides user with
information on how
to prepare for a seizure
○W
ill autom
atically appear w
ith an alert
●“B
egin” button provides inform
ation for som
ebody (besides user) to provide assistance during a seizure
“Manage”●
Allows user to
specify how far
in advance they w
ant to receive an alarm
Features “Tracker” & “Contacts”
“Tracker”●
Where user can go
back and log in activities they recall before their seizure, im
mediately after
their seizure has ceased.
“Contacts”●
Where em
ergency contact info can be inputted to be notified in case of seizure
Future Features-
“Brainwaves” tab
●W
hen device is complete, this w
ill show the patient’s live EEG
signal being received via bluetooth
“Personal Info” tab●
This will have patient’s personal info (for exam
ple, medication) that can
be used in case of emergency
App D
eletion Notification
●E
mergency contacts are im
mediately notified if app is deleted off
patient’s phone.