SlideShare a Scribd company logo
1 of 33
Download to read offline
Multimodal Analysis of Speech, Handwriting
and Gait for the Assessment of Patients with
Parkinson’s Disease
Student: Juan Camilo V´asquez Correa
Advisors:
Prof. Juan Rafael Orozco Arroyave1, Prof. Elmar N¨oth2
1GITA research group, University of Antioquia UdeA.
2Pattern recognition Lab. Friedrich Alexander Universit¨at. Erlangen-N¨urnberg.
jcamilo.vasquez@udea.edu.co
1 / 33
Outline
Introduction
Methods
Speech Analysis
Gait Analysis
Handwriting Analysis
Multimodal Analysis
Results
Multi-view learning using GCCA
Convolutional neural networks for speech analysis
Ongoing work
Conclusion
2 / 33
Introduction: Parkinson’s Disease
Second most prevalent
neurological disorder
worldwide.
Patients develop sev-
eral motor and non-
motor impairments. (O.
Hornykiewicz 1998).
3 / 33
Introduction: Parkinson’s Disease II
Motor impairments
Bradykinesia
Rigidity
Resting tremor
Micrographia
Dysartrhia
Non–Motor impairments
Depression
Sleep disorders
Cognitive impairments
Sensory system deficits
Evaluated by neurologist experts according to the
MDS-UPDRS-III scale (C. G. Goetz et al. 2008).
4 / 33
Introduction: Parkinson’s Disease II
Motor impairments
Bradykinesia
Rigidity
Resting tremor
Micrographia
Dysartrhia
Non–Motor impairments
Depression
Sleep disorders
Cognitive impairments
Sensory system
deficits
Evaluated by psychologist experts.
5 / 33
Introduction: Research Problem
Motor evaluation is expensive and time–consuming.
Neurologists evaluate perceptually the motor deficits of the
patients.
The assessment of the motor capabilities provides suitable
information to update the treatment and the medication.
6 / 33
Introduction: Research Problem
Motor evaluation is expensive and time–consuming.
Neurologists evaluate perceptually the motor deficits of the
patients.
The assessment of the motor capabilities provides suitable
information to update the treatment and the medication.
7 / 33
Introduction: Research Problem
Motor evaluation is expensive and time–consuming.
Neurologists evaluate perceptually the motor deficits of the
patients.
The assessment of the motor capabilities provides suitable
information to update the treatment and the medication.
8 / 33
Introduction: Justification
if the motor evaluation is performed with bio–signals such
as speech, handwriting and gait, the treatment could be fol-
lowed in a more objective way
9 / 33
Introduction: Proposal
These three bio-signals constitute a reliable source of
information to describe several symptoms of PD patients.
The combination of such sources of information allows to
perform an accurate quantification of the neurological state
of the patients.
The multimodal analysis that includes information
from different kind of sensors for the analysis of PD
has not been enough studied (Q. W. Oung, et al. 2015)
10 / 33
Introduction: Proposal
These three bio-signals constitute a reliable source of
information to describe several symptoms of PD patients.
The combination of such sources of information allows to
perform an accurate quantification of the neurological state
of the patients.
The multimodal analysis that includes information
from different kind of sensors for the analysis of PD
has not been enough studied (Q. W. Oung, et al. 2015)
11 / 33
Introduction: Proposal
These three bio-signals constitute a reliable source of
information to describe several symptoms of PD patients.
The combination of such sources of information allows to
perform an accurate quantification of the neurological state
of the patients.
The multimodal analysis that includes information
from different kind of sensors for the analysis of PD
has not been enough studied (Q. W. Oung, et al. 2015)
12 / 33
Methods: Speech analysis
Speech impairments can be assessed using four dimensions
(J. R. Orozco-Arroyave 2016)
Phonation
Articulation
Prosody Intelligibility
pataka pataka
13 / 33
Methods: Speech analysis
https://github.com/jcvasquezc/NeuroSpeech1
1J. R. Orozco-Arroyave, J. C. V´asquez-Correa, et. al. “NeuroSpeech: an
open-source software for Parkinson’s speech analysis”. In: Digital Signal Precessing
and SoftwareX, (Under review) (2017).
14 / 33
Methods: Gait analysis
Gait acquisition
15 / 33
Methods: Gait analysis
Heel
strike
Toe
off
Heel
strike
Stance time Swing time
Stride time & Stride length
Toe off / Heel strike
angles
( J. Klucken et al. 2013; P. A. P´erez-Toro 2016 )
16 / 33
Methods: Gait analysis
Time–frequency analysis
0 5 10 15 20
Time (s)
0
10
20
30
40
50
Frequency(Hz)
Healthy Control
0 5 10 15 20 25
Time (s)
Patient
Short time Fourier Transform
Wavelet transform
Modulation spectra
Wigner Ville distribution
17 / 33
Methods: Handwriting analysis
Handwriting acquisition
18 / 33
Methods: Handwriting analysis
Speed of the stroke
Acceleration
In–air movement
Pressure of the pen
Azimuth
(P. Drot´ar et al. 2016)
Static handwriting analysis
(Z. Naiquian et. al 2017 )
19 / 33
Methods: Multimodal analysis
GCCA
(J. C. V´asquez-Correa, et
al. 2017)
Early fusion
Weak learners
Multimodal convolutional
neural networks
Deep autoencoders.
20 / 33
Methods: Pattern analysis
Classical machine learning
Support vector
machines.
Support vector
regressors.
minimize
w,b,ξ
1
2 ||w||2
+ C N
i=1 ξi
subject to yi · (xT
i w + b) ≥ 1 − ξi,
ξi ≥ 0
21 / 33
Methods: Pattern analysis
Deep learning
Convolutional neural networks.
Recurrent neural networks and LSTM
Variational deep autoencoders
Convolutional layer 1 Convolutional layer 2Pooling layer Pooling layer 2 RNN-LSTM
Input
y
J hidden filters K hidden filters
22 / 33
Results: Multi-view learning using GCCA2
To obtain a new feature representation when multimodal
data is not available.
To predict missing information.
Machine learning methods are trained with the new feature
representation.
arg min
Uj
J
j=1
G − XjUj
2
F
s.t. GT
G = I
2J. C. V´asquez-Correa, et al. “Multi-view representation learning via GCCA for
multimodal analysis of Parkinson’s disease”. In: 42nd International Conference on
Acoustic, Speech, and Signal Processing (ICASSP). 2017.
23 / 33
Results: Multi-view learning using GCCA3
Baseline GCCA
Classification PD vs. HC 77% 78%
Neurological state prediction 0.36 0.40
Speech quality prediction 0.67 0.71
Table: Results GCCA
The proposed approach is suitable to map the features
from other modalities that are not always available.
3J. C. V´asquez-Correa, et al. “Multi-view representation learning via GCCA for
multimodal analysis of Parkinson’s disease”. In: 42nd International Conference on
Acoustic, Speech, and Signal Processing (ICASSP). 2017.
24 / 33
Results: CNNs for speech analysis4
Convolution layer I Convolution layer IIMax-pool. layer 1 Max-pool layer 2 Fully conected MLP
Input layer
PD
vs.
HC
Feature maps 1
Feature maps 2
convolutional neural networks (CNN) learns high–level
representations from the low–level raw data.
CNN is formed with an array of convolutional filters and
subsampling layers.
HL(i, j, d) = (I ∗ Kd )(i, j) = m n I(i + m, j + n)Kd (m, n)
4J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural
Network to Model Articulation Impairments in Patients with Parkinson’s Disease”. In:
18th International Conference of the Speech and Communication Association
(INTERSPEECH). 2017.
25 / 33
Results: CNNs for speech analysis5
Voiced-Unvoiced transitions are modeled with CNNs and
TFRs.
The STFT and the continuous wavelet transform (CWT)
are considered.
Speech of PD patients in three languages: Spanish,
German and Czech.
5J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural
Network to Model Articulation Impairments in Patients with Parkinson’s Disease”. In:
18th International Conference of the Speech and Communication Association
(INTERSPEECH). 2017.
26 / 33
Results: CNNs for speech analysis6
Language Accuracy
Spanish 85.9%
German 75.0%
Czech 89.4%
Table: Classification of PD vs. HC
using CNNs
50 100 150
Time (ms)
0
1000
2000
3000
4000
Frequency(Hz)
50 100 150
Time (ms)
Low Energy High Energy
Figure: Output of the CNN after
the last max–pool layer: PD
patient (left) and a HC speaker
(right)
6J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural
Network to Model Articulation Impairments in Patients with Parkinson’s Disease”. In:
18th International Conference of the Speech and Communication Association
(INTERSPEECH). 2017.
27 / 33
Ongoing work
Combination of convolutional and recurrent neural
networks for multimodal analysis of PD.
Convolutional layer 1 Convolutional layer 2Pooling layer Pooling layer 2 RNN-LSTM
Speech
y
J hidden filters K hidden filters
Gait
Handwriting
28 / 33
Ongoing work
Combination of convolutional and recurrent neural
networks for multimodal analysis of PD.
𝒙(𝑡)
𝒉(𝑡−1)
+
𝑼 𝑓 𝑾 𝑓
𝑓: forget gate
+
𝑼 𝑾
𝒙(𝑡)
𝒉(𝑡−1)
𝒙(𝑡)
𝒉(𝑡−1)
+
𝑼 𝑔 𝑾 𝑔
𝑔: input gate
∗+∗
𝒔(𝑡)
𝒔(𝑡−1)
𝑠: hidden state
𝑡𝑎𝑛ℎ
𝒉(𝑡)
𝒙(𝑡)
𝒉(𝑡−1)
+
𝑼 𝑞 𝑾 𝑞
𝑞: output gate
LSTM cell
29 / 33
Conclusion
Speech, handwriting and gait signals constitute a reliable
source of information to describe several symptoms of PD
patients.
The combination of such sources of information allows to
perform an accurate quantification of the neurological state
of the patients.
Several features and pattern analysis approaches could be
considered to improve the classification of the disease, and
the monitoring of the neurological state of the patients.
30 / 33
References I
J. Klucken et al. “Unbiased and mobile gait analysis detects motor impairment in Parkin-
son’s disease”. In: PloS one 8.2 (2013), e56956.
C. G. Goetz et al. “Movement Disorder Society-sponsored revision of the Unified Parkin-
son’s Disease Rating Scale (MDS-UPDRS): Scale presentation and clinimetric test-
ing results”. In: Movement disorders 23.15 (2008), pp. 2129–2170.
J. C. V´asquez-Correa, et al. “Multi-view representation learning via GCCA for multimodal
analysis of Parkinson’s disease”. In: 42nd International Conference on Acoustic,
Speech, and Signal Processing (ICASSP). 2017.
J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural Net-
work to Model Articulation Impairments in Patients with Parkinson’s Disease”. In:
18th International Conference of the Speech and Communication Association (IN-
TERSPEECH). 2017.
J. R. Orozco-Arroyave. Analysis of Speech of People with Parkinson’s Disease. Ger-
many: Logos Verlag Berlin, 2016.
J. R. Orozco-Arroyave, J. C. V´asquez-Correa, et. al. “NeuroSpeech: an open-source
software for Parkinson’s speech analysis”. In: Digital Signal Precessing and Soft-
wareX, (Under review) (2017).
O. Hornykiewicz. “Biochemical aspects of Parkinson’s disease”. In: Neurology 51.2
Suppl 2 (1998), S2–S9.
P. A. P´erez-Toro J. C. V´asquez-Correa, et. al. “An´alisis motriz en las extremidades in-
feriores para el monitoreo del estado neurol´ogico de pacientes con enfermedad de
Parkinson”. In: XXI Symposium on Image, Signal Processing and Artificial Vision
(STSIVA). 2016.
References II
P. Drot´ar et al. “Evaluation of handwriting kinematics and pressure for differential diag-
nosis of Parkinson’s disease”. In: Artificial intelligence in Medicine 67 (2016), pp. 39–
46.
Q. W. Oung, et al. “Technologies for assessment of motor disorders in Parkinson’s dis-
ease: a review”. In: Sensors 15.9 (2015), pp. 21710–21745.
Z. Naiquian et. al. “Toward Monitoring Parkinson’s through Analysis of Static Handwriting
Samples: A Quantitative Analytical Framework”. In: IEEE journal of biomedical and
health informatics 21.2 (2017), pp. 488–495.
Multimodal Analysis of Speech, Handwriting
and Gait for the Assessment of Patients with
Parkinson’s Disease
Student: Juan Camilo V´asquez Correa
Advisors:
Prof. Juan Rafael Orozco Arroyave1, Prof. Elmar N¨oth2
1GITA research group, University of Antioquia UdeA.
2Pattern recognition Lab. Friedrich Alexander Universit¨at. Erlangen-N¨urnberg.
jcamilo.vasquez@udea.edu.co
33 / 33

More Related Content

Similar to Multimodal Analysis of Speech, Handwriting and Gait for the Assessment of Patients with Parkinson’s Disease

Machine Learning Based Approaches for Prediction of Parkinson's Disease
Machine Learning Based Approaches for Prediction of Parkinson's Disease  Machine Learning Based Approaches for Prediction of Parkinson's Disease
Machine Learning Based Approaches for Prediction of Parkinson's Disease mlaij
 
Brain structural connectivity and functional default mode network in deafness
Brain structural connectivity and functional default mode network in deafnessBrain structural connectivity and functional default mode network in deafness
Brain structural connectivity and functional default mode network in deafnessAntonio Carlos da Silva Senra Filho
 
Recent Advances in Deep Learning Techniques for Electronic Health Record
Recent Advances in Deep Learning Techniques for Electronic Health RecordRecent Advances in Deep Learning Techniques for Electronic Health Record
Recent Advances in Deep Learning Techniques for Electronic Health Recordkingstdio
 
Micro-Neuro-Sensor Recording of STN Neurons of the Human Brain
Micro-Neuro-Sensor Recording of STN Neurons of the Human BrainMicro-Neuro-Sensor Recording of STN Neurons of the Human Brain
Micro-Neuro-Sensor Recording of STN Neurons of the Human BrainMangaiK4
 
POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...
POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...
POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...ijbesjournal
 
ARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGY
ARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGYARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGY
ARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGYconvegnonazionaleaiic
 
Leibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific NotationLeibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific Notationkhinsen
 
Wavelet-based EEG processing for computer-aided seizure detection and epileps...
Wavelet-based EEG processing for computer-aided seizure detection and epileps...Wavelet-based EEG processing for computer-aided seizure detection and epileps...
Wavelet-based EEG processing for computer-aided seizure detection and epileps...IJERA Editor
 
Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...
Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...
Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...ijtsrd
 
Elements of Theory for Multi-Neuronal Systems
Elements of Theory for Multi-Neuronal SystemsElements of Theory for Multi-Neuronal Systems
Elements of Theory for Multi-Neuronal SystemsSSA KPI
 
DevFest19 - Early Diagnosis of Chronic Diseases by Smartphone AI
DevFest19 -  Early Diagnosis of Chronic Diseases by Smartphone AIDevFest19 -  Early Diagnosis of Chronic Diseases by Smartphone AI
DevFest19 - Early Diagnosis of Chronic Diseases by Smartphone AIGaurav Kheterpal
 
A novel convolutional neural network based dysphonic voice detection algorit...
A novel convolutional neural network based dysphonic voice  detection algorit...A novel convolutional neural network based dysphonic voice  detection algorit...
A novel convolutional neural network based dysphonic voice detection algorit...IJECEIAES
 
Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...
Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...
Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...Codiax
 
Cross correlation analysis of
Cross correlation analysis ofCross correlation analysis of
Cross correlation analysis ofcsandit
 
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPYCROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPYcscpconf
 
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPYCROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPYcscpconf
 
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING hiij
 
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING hiij
 
How to quantify hierarchy?
How to quantify hierarchy?How to quantify hierarchy?
How to quantify hierarchy?Dániel Czégel
 

Similar to Multimodal Analysis of Speech, Handwriting and Gait for the Assessment of Patients with Parkinson’s Disease (20)

Machine Learning Based Approaches for Prediction of Parkinson's Disease
Machine Learning Based Approaches for Prediction of Parkinson's Disease  Machine Learning Based Approaches for Prediction of Parkinson's Disease
Machine Learning Based Approaches for Prediction of Parkinson's Disease
 
Brain structural connectivity and functional default mode network in deafness
Brain structural connectivity and functional default mode network in deafnessBrain structural connectivity and functional default mode network in deafness
Brain structural connectivity and functional default mode network in deafness
 
Recent Advances in Deep Learning Techniques for Electronic Health Record
Recent Advances in Deep Learning Techniques for Electronic Health RecordRecent Advances in Deep Learning Techniques for Electronic Health Record
Recent Advances in Deep Learning Techniques for Electronic Health Record
 
Micro-Neuro-Sensor Recording of STN Neurons of the Human Brain
Micro-Neuro-Sensor Recording of STN Neurons of the Human BrainMicro-Neuro-Sensor Recording of STN Neurons of the Human Brain
Micro-Neuro-Sensor Recording of STN Neurons of the Human Brain
 
ANN presentataion
ANN presentataionANN presentataion
ANN presentataion
 
POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...
POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...
POWER SPECTRAL ANALYSIS OF EEG AS A POTENTIAL MARKER IN THE DIAGNOSIS OF SPAS...
 
ARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGY
ARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGYARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGY
ARTIFICIAL INTELLIGENCE BASED WEB PLATFORM FOR DIGITAL NEUROLOGY
 
Leibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific NotationLeibniz: A Digital Scientific Notation
Leibniz: A Digital Scientific Notation
 
Wavelet-based EEG processing for computer-aided seizure detection and epileps...
Wavelet-based EEG processing for computer-aided seizure detection and epileps...Wavelet-based EEG processing for computer-aided seizure detection and epileps...
Wavelet-based EEG processing for computer-aided seizure detection and epileps...
 
Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...
Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...
Comparative Analysis of Original Wave & Filtered Wave of EEG signal Used in t...
 
Elements of Theory for Multi-Neuronal Systems
Elements of Theory for Multi-Neuronal SystemsElements of Theory for Multi-Neuronal Systems
Elements of Theory for Multi-Neuronal Systems
 
DevFest19 - Early Diagnosis of Chronic Diseases by Smartphone AI
DevFest19 -  Early Diagnosis of Chronic Diseases by Smartphone AIDevFest19 -  Early Diagnosis of Chronic Diseases by Smartphone AI
DevFest19 - Early Diagnosis of Chronic Diseases by Smartphone AI
 
A novel convolutional neural network based dysphonic voice detection algorit...
A novel convolutional neural network based dysphonic voice  detection algorit...A novel convolutional neural network based dysphonic voice  detection algorit...
A novel convolutional neural network based dysphonic voice detection algorit...
 
Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...
Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...
Fireside chat: Newton Howard, Director of the MIT Synthetic Intelligence Lab ...
 
Cross correlation analysis of
Cross correlation analysis ofCross correlation analysis of
Cross correlation analysis of
 
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPYCROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
 
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPYCROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
CROSS CORRELATION ANALYSIS OF MULTI-CHANNEL NEAR INFRARED SPECTROSCOPY
 
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
 
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
DETECTION OF CRACKLES AND WHEEZES IN LUNG SOUND USING TRANSFER LEARNING
 
How to quantify hierarchy?
How to quantify hierarchy?How to quantify hierarchy?
How to quantify hierarchy?
 

Recently uploaded

GEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLE
GEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLEGEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLE
GEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLEselvakumar948
 
Verification of thevenin's theorem for BEEE Lab (1).pptx
Verification of thevenin's theorem for BEEE Lab (1).pptxVerification of thevenin's theorem for BEEE Lab (1).pptx
Verification of thevenin's theorem for BEEE Lab (1).pptxchumtiyababu
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueBhangaleSonal
 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfJiananWang21
 
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptxS1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptxSCMS School of Architecture
 
PE 459 LECTURE 2- natural gas basic concepts and properties
PE 459 LECTURE 2- natural gas basic concepts and propertiesPE 459 LECTURE 2- natural gas basic concepts and properties
PE 459 LECTURE 2- natural gas basic concepts and propertiessarkmank1
 
kiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal loadkiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal loadhamedmustafa094
 
Wadi Rum luxhotel lodge Analysis case study.pptx
Wadi Rum luxhotel lodge Analysis case study.pptxWadi Rum luxhotel lodge Analysis case study.pptx
Wadi Rum luxhotel lodge Analysis case study.pptxNadaHaitham1
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaOmar Fathy
 
Thermal Engineering Unit - I & II . ppt
Thermal Engineering  Unit - I & II . pptThermal Engineering  Unit - I & II . ppt
Thermal Engineering Unit - I & II . pptDineshKumar4165
 
Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...
Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...
Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...Call Girls Mumbai
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdfKamal Acharya
 
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments""Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"mphochane1998
 
Tamil Call Girls Bhayandar WhatsApp +91-9930687706, Best Service
Tamil Call Girls Bhayandar WhatsApp +91-9930687706, Best ServiceTamil Call Girls Bhayandar WhatsApp +91-9930687706, Best Service
Tamil Call Girls Bhayandar WhatsApp +91-9930687706, Best Servicemeghakumariji156
 
Kuwait City MTP kit ((+919101817206)) Buy Abortion Pills Kuwait
Kuwait City MTP kit ((+919101817206)) Buy Abortion Pills KuwaitKuwait City MTP kit ((+919101817206)) Buy Abortion Pills Kuwait
Kuwait City MTP kit ((+919101817206)) Buy Abortion Pills Kuwaitjaanualu31
 
Hospital management system project report.pdf
Hospital management system project report.pdfHospital management system project report.pdf
Hospital management system project report.pdfKamal Acharya
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.Kamal Acharya
 
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...Amil baba
 
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptxOrlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptxMuhammadAsimMuhammad6
 

Recently uploaded (20)

GEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLE
GEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLEGEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLE
GEAR TRAIN- BASIC CONCEPTS AND WORKING PRINCIPLE
 
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak HamilCara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
 
Verification of thevenin's theorem for BEEE Lab (1).pptx
Verification of thevenin's theorem for BEEE Lab (1).pptxVerification of thevenin's theorem for BEEE Lab (1).pptx
Verification of thevenin's theorem for BEEE Lab (1).pptx
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torque
 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdf
 
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptxS1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
 
PE 459 LECTURE 2- natural gas basic concepts and properties
PE 459 LECTURE 2- natural gas basic concepts and propertiesPE 459 LECTURE 2- natural gas basic concepts and properties
PE 459 LECTURE 2- natural gas basic concepts and properties
 
kiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal loadkiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal load
 
Wadi Rum luxhotel lodge Analysis case study.pptx
Wadi Rum luxhotel lodge Analysis case study.pptxWadi Rum luxhotel lodge Analysis case study.pptx
Wadi Rum luxhotel lodge Analysis case study.pptx
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS Lambda
 
Thermal Engineering Unit - I & II . ppt
Thermal Engineering  Unit - I & II . pptThermal Engineering  Unit - I & II . ppt
Thermal Engineering Unit - I & II . ppt
 
Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...
Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...
Bhubaneswar🌹Call Girls Bhubaneswar ❤Komal 9777949614 💟 Full Trusted CALL GIRL...
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdf
 
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments""Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
 
Tamil Call Girls Bhayandar WhatsApp +91-9930687706, Best Service
Tamil Call Girls Bhayandar WhatsApp +91-9930687706, Best ServiceTamil Call Girls Bhayandar WhatsApp +91-9930687706, Best Service
Tamil Call Girls Bhayandar WhatsApp +91-9930687706, Best Service
 
Kuwait City MTP kit ((+919101817206)) Buy Abortion Pills Kuwait
Kuwait City MTP kit ((+919101817206)) Buy Abortion Pills KuwaitKuwait City MTP kit ((+919101817206)) Buy Abortion Pills Kuwait
Kuwait City MTP kit ((+919101817206)) Buy Abortion Pills Kuwait
 
Hospital management system project report.pdf
Hospital management system project report.pdfHospital management system project report.pdf
Hospital management system project report.pdf
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.
 
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
 
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptxOrlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
 

Multimodal Analysis of Speech, Handwriting and Gait for the Assessment of Patients with Parkinson’s Disease

  • 1. Multimodal Analysis of Speech, Handwriting and Gait for the Assessment of Patients with Parkinson’s Disease Student: Juan Camilo V´asquez Correa Advisors: Prof. Juan Rafael Orozco Arroyave1, Prof. Elmar N¨oth2 1GITA research group, University of Antioquia UdeA. 2Pattern recognition Lab. Friedrich Alexander Universit¨at. Erlangen-N¨urnberg. jcamilo.vasquez@udea.edu.co 1 / 33
  • 2. Outline Introduction Methods Speech Analysis Gait Analysis Handwriting Analysis Multimodal Analysis Results Multi-view learning using GCCA Convolutional neural networks for speech analysis Ongoing work Conclusion 2 / 33
  • 3. Introduction: Parkinson’s Disease Second most prevalent neurological disorder worldwide. Patients develop sev- eral motor and non- motor impairments. (O. Hornykiewicz 1998). 3 / 33
  • 4. Introduction: Parkinson’s Disease II Motor impairments Bradykinesia Rigidity Resting tremor Micrographia Dysartrhia Non–Motor impairments Depression Sleep disorders Cognitive impairments Sensory system deficits Evaluated by neurologist experts according to the MDS-UPDRS-III scale (C. G. Goetz et al. 2008). 4 / 33
  • 5. Introduction: Parkinson’s Disease II Motor impairments Bradykinesia Rigidity Resting tremor Micrographia Dysartrhia Non–Motor impairments Depression Sleep disorders Cognitive impairments Sensory system deficits Evaluated by psychologist experts. 5 / 33
  • 6. Introduction: Research Problem Motor evaluation is expensive and time–consuming. Neurologists evaluate perceptually the motor deficits of the patients. The assessment of the motor capabilities provides suitable information to update the treatment and the medication. 6 / 33
  • 7. Introduction: Research Problem Motor evaluation is expensive and time–consuming. Neurologists evaluate perceptually the motor deficits of the patients. The assessment of the motor capabilities provides suitable information to update the treatment and the medication. 7 / 33
  • 8. Introduction: Research Problem Motor evaluation is expensive and time–consuming. Neurologists evaluate perceptually the motor deficits of the patients. The assessment of the motor capabilities provides suitable information to update the treatment and the medication. 8 / 33
  • 9. Introduction: Justification if the motor evaluation is performed with bio–signals such as speech, handwriting and gait, the treatment could be fol- lowed in a more objective way 9 / 33
  • 10. Introduction: Proposal These three bio-signals constitute a reliable source of information to describe several symptoms of PD patients. The combination of such sources of information allows to perform an accurate quantification of the neurological state of the patients. The multimodal analysis that includes information from different kind of sensors for the analysis of PD has not been enough studied (Q. W. Oung, et al. 2015) 10 / 33
  • 11. Introduction: Proposal These three bio-signals constitute a reliable source of information to describe several symptoms of PD patients. The combination of such sources of information allows to perform an accurate quantification of the neurological state of the patients. The multimodal analysis that includes information from different kind of sensors for the analysis of PD has not been enough studied (Q. W. Oung, et al. 2015) 11 / 33
  • 12. Introduction: Proposal These three bio-signals constitute a reliable source of information to describe several symptoms of PD patients. The combination of such sources of information allows to perform an accurate quantification of the neurological state of the patients. The multimodal analysis that includes information from different kind of sensors for the analysis of PD has not been enough studied (Q. W. Oung, et al. 2015) 12 / 33
  • 13. Methods: Speech analysis Speech impairments can be assessed using four dimensions (J. R. Orozco-Arroyave 2016) Phonation Articulation Prosody Intelligibility pataka pataka 13 / 33
  • 14. Methods: Speech analysis https://github.com/jcvasquezc/NeuroSpeech1 1J. R. Orozco-Arroyave, J. C. V´asquez-Correa, et. al. “NeuroSpeech: an open-source software for Parkinson’s speech analysis”. In: Digital Signal Precessing and SoftwareX, (Under review) (2017). 14 / 33
  • 15. Methods: Gait analysis Gait acquisition 15 / 33
  • 16. Methods: Gait analysis Heel strike Toe off Heel strike Stance time Swing time Stride time & Stride length Toe off / Heel strike angles ( J. Klucken et al. 2013; P. A. P´erez-Toro 2016 ) 16 / 33
  • 17. Methods: Gait analysis Time–frequency analysis 0 5 10 15 20 Time (s) 0 10 20 30 40 50 Frequency(Hz) Healthy Control 0 5 10 15 20 25 Time (s) Patient Short time Fourier Transform Wavelet transform Modulation spectra Wigner Ville distribution 17 / 33
  • 19. Methods: Handwriting analysis Speed of the stroke Acceleration In–air movement Pressure of the pen Azimuth (P. Drot´ar et al. 2016) Static handwriting analysis (Z. Naiquian et. al 2017 ) 19 / 33
  • 20. Methods: Multimodal analysis GCCA (J. C. V´asquez-Correa, et al. 2017) Early fusion Weak learners Multimodal convolutional neural networks Deep autoencoders. 20 / 33
  • 21. Methods: Pattern analysis Classical machine learning Support vector machines. Support vector regressors. minimize w,b,ξ 1 2 ||w||2 + C N i=1 ξi subject to yi · (xT i w + b) ≥ 1 − ξi, ξi ≥ 0 21 / 33
  • 22. Methods: Pattern analysis Deep learning Convolutional neural networks. Recurrent neural networks and LSTM Variational deep autoencoders Convolutional layer 1 Convolutional layer 2Pooling layer Pooling layer 2 RNN-LSTM Input y J hidden filters K hidden filters 22 / 33
  • 23. Results: Multi-view learning using GCCA2 To obtain a new feature representation when multimodal data is not available. To predict missing information. Machine learning methods are trained with the new feature representation. arg min Uj J j=1 G − XjUj 2 F s.t. GT G = I 2J. C. V´asquez-Correa, et al. “Multi-view representation learning via GCCA for multimodal analysis of Parkinson’s disease”. In: 42nd International Conference on Acoustic, Speech, and Signal Processing (ICASSP). 2017. 23 / 33
  • 24. Results: Multi-view learning using GCCA3 Baseline GCCA Classification PD vs. HC 77% 78% Neurological state prediction 0.36 0.40 Speech quality prediction 0.67 0.71 Table: Results GCCA The proposed approach is suitable to map the features from other modalities that are not always available. 3J. C. V´asquez-Correa, et al. “Multi-view representation learning via GCCA for multimodal analysis of Parkinson’s disease”. In: 42nd International Conference on Acoustic, Speech, and Signal Processing (ICASSP). 2017. 24 / 33
  • 25. Results: CNNs for speech analysis4 Convolution layer I Convolution layer IIMax-pool. layer 1 Max-pool layer 2 Fully conected MLP Input layer PD vs. HC Feature maps 1 Feature maps 2 convolutional neural networks (CNN) learns high–level representations from the low–level raw data. CNN is formed with an array of convolutional filters and subsampling layers. HL(i, j, d) = (I ∗ Kd )(i, j) = m n I(i + m, j + n)Kd (m, n) 4J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural Network to Model Articulation Impairments in Patients with Parkinson’s Disease”. In: 18th International Conference of the Speech and Communication Association (INTERSPEECH). 2017. 25 / 33
  • 26. Results: CNNs for speech analysis5 Voiced-Unvoiced transitions are modeled with CNNs and TFRs. The STFT and the continuous wavelet transform (CWT) are considered. Speech of PD patients in three languages: Spanish, German and Czech. 5J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural Network to Model Articulation Impairments in Patients with Parkinson’s Disease”. In: 18th International Conference of the Speech and Communication Association (INTERSPEECH). 2017. 26 / 33
  • 27. Results: CNNs for speech analysis6 Language Accuracy Spanish 85.9% German 75.0% Czech 89.4% Table: Classification of PD vs. HC using CNNs 50 100 150 Time (ms) 0 1000 2000 3000 4000 Frequency(Hz) 50 100 150 Time (ms) Low Energy High Energy Figure: Output of the CNN after the last max–pool layer: PD patient (left) and a HC speaker (right) 6J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural Network to Model Articulation Impairments in Patients with Parkinson’s Disease”. In: 18th International Conference of the Speech and Communication Association (INTERSPEECH). 2017. 27 / 33
  • 28. Ongoing work Combination of convolutional and recurrent neural networks for multimodal analysis of PD. Convolutional layer 1 Convolutional layer 2Pooling layer Pooling layer 2 RNN-LSTM Speech y J hidden filters K hidden filters Gait Handwriting 28 / 33
  • 29. Ongoing work Combination of convolutional and recurrent neural networks for multimodal analysis of PD. 𝒙(𝑡) 𝒉(𝑡−1) + 𝑼 𝑓 𝑾 𝑓 𝑓: forget gate + 𝑼 𝑾 𝒙(𝑡) 𝒉(𝑡−1) 𝒙(𝑡) 𝒉(𝑡−1) + 𝑼 𝑔 𝑾 𝑔 𝑔: input gate ∗+∗ 𝒔(𝑡) 𝒔(𝑡−1) 𝑠: hidden state 𝑡𝑎𝑛ℎ 𝒉(𝑡) 𝒙(𝑡) 𝒉(𝑡−1) + 𝑼 𝑞 𝑾 𝑞 𝑞: output gate LSTM cell 29 / 33
  • 30. Conclusion Speech, handwriting and gait signals constitute a reliable source of information to describe several symptoms of PD patients. The combination of such sources of information allows to perform an accurate quantification of the neurological state of the patients. Several features and pattern analysis approaches could be considered to improve the classification of the disease, and the monitoring of the neurological state of the patients. 30 / 33
  • 31. References I J. Klucken et al. “Unbiased and mobile gait analysis detects motor impairment in Parkin- son’s disease”. In: PloS one 8.2 (2013), e56956. C. G. Goetz et al. “Movement Disorder Society-sponsored revision of the Unified Parkin- son’s Disease Rating Scale (MDS-UPDRS): Scale presentation and clinimetric test- ing results”. In: Movement disorders 23.15 (2008), pp. 2129–2170. J. C. V´asquez-Correa, et al. “Multi-view representation learning via GCCA for multimodal analysis of Parkinson’s disease”. In: 42nd International Conference on Acoustic, Speech, and Signal Processing (ICASSP). 2017. J. C. V´asquez-Correa, J. R. Orozco-Arroyave, and E. N¨oth. “Convolutional Neural Net- work to Model Articulation Impairments in Patients with Parkinson’s Disease”. In: 18th International Conference of the Speech and Communication Association (IN- TERSPEECH). 2017. J. R. Orozco-Arroyave. Analysis of Speech of People with Parkinson’s Disease. Ger- many: Logos Verlag Berlin, 2016. J. R. Orozco-Arroyave, J. C. V´asquez-Correa, et. al. “NeuroSpeech: an open-source software for Parkinson’s speech analysis”. In: Digital Signal Precessing and Soft- wareX, (Under review) (2017). O. Hornykiewicz. “Biochemical aspects of Parkinson’s disease”. In: Neurology 51.2 Suppl 2 (1998), S2–S9. P. A. P´erez-Toro J. C. V´asquez-Correa, et. al. “An´alisis motriz en las extremidades in- feriores para el monitoreo del estado neurol´ogico de pacientes con enfermedad de Parkinson”. In: XXI Symposium on Image, Signal Processing and Artificial Vision (STSIVA). 2016.
  • 32. References II P. Drot´ar et al. “Evaluation of handwriting kinematics and pressure for differential diag- nosis of Parkinson’s disease”. In: Artificial intelligence in Medicine 67 (2016), pp. 39– 46. Q. W. Oung, et al. “Technologies for assessment of motor disorders in Parkinson’s dis- ease: a review”. In: Sensors 15.9 (2015), pp. 21710–21745. Z. Naiquian et. al. “Toward Monitoring Parkinson’s through Analysis of Static Handwriting Samples: A Quantitative Analytical Framework”. In: IEEE journal of biomedical and health informatics 21.2 (2017), pp. 488–495.
  • 33. Multimodal Analysis of Speech, Handwriting and Gait for the Assessment of Patients with Parkinson’s Disease Student: Juan Camilo V´asquez Correa Advisors: Prof. Juan Rafael Orozco Arroyave1, Prof. Elmar N¨oth2 1GITA research group, University of Antioquia UdeA. 2Pattern recognition Lab. Friedrich Alexander Universit¨at. Erlangen-N¨urnberg. jcamilo.vasquez@udea.edu.co 33 / 33