This document summarizes Nuno Brás's PhD dissertation on developing new approaches for magnetic induction tomography (MIT). The dissertation aimed to implement a new MIT prototype and numerical framework to handle large numbers of measurements. It developed an experimental moving coil system that achieved state-of-the-art sensitivity. A new 3D eddy current solver was also created with competitive speed and low error. Additionally, an alternating direction method of multipliers was developed and shown to provide advantages for large datasets, being applied to 2D and 3D inverse problems.
An Empirical Study on Faith-based Microfinance as an Alternative Tool of Poverty Alleviation. The doctoral study discussed the role of FBOs in microfinance.
An Empirical Study on Faith-based Microfinance as an Alternative Tool of Poverty Alleviation. The doctoral study discussed the role of FBOs in microfinance.
PhD thesis defense presentation for my topic "Improving Content Delivery and Service Discovery in Networks" for wireless and other networks. Columbia University, 2016.
Presentation slides for my PhD thesis dissertation on machine learning algorithm development to analyze multi dimensional genomic data such as microarrays
Luz rello - Ph.D. Thesis presentation - DysWebxia: A Text Accessibility Model...Luz Rello
Ph.D. Presentation
Title: DysWebxia: A Text Accessibility Model for People with Dyslexia
Author: Luz Rello
Advisors: Ricardo Baeza-Yates and Horacio Saggion
Abstract: Worldwide, 10% of the population has dyslexia, a cognitive disability that reduces readability and comprehension of written information. The goal of this thesis is to make text more accessible for people with dyslexia by combining human computer interaction validation methods and natural language processing techniques. In the initial phase of this study we examined how people with dyslexia identify errors in written text. Their written errors were analyzed and used to estimate the presence of text written by individuals with dyslexia in the Web. After concluding that dyslexic errors relate to presentation and content features of text, we carried out a set of experiments using eye tracking to determine the conditions that led to improved readability and comprehension. After finding the relevant parameters for text presentation and content modification, we implemented a lexical simplification system. Finally, the results of the investigation and the resources created, lead to a model, DysWebxia, that proposes a set of recommendations that have been successfully integrated in four applications.
This E-nano Newsletter special double issue
contains the updated version of the nanoICT
position paper on Carbon Nanotubes (CNTs)
summarising state-of-the-art research in this field
as well as a description of the possible electrical,
electronic and photonic applications of carbon
nanotubes, the types of CNTs employed and the
organisations or groups that are most proficient
at fabricating them.
In the second paper, the Nanoelectronics
European Research Roadmap is addressed
focusing on the main European Programmes
supporting the short, medium and long-term
research activities.
This issue also contains a catalogue (insert),
compiled by the Phantoms Foundation
providing a general overview of the
Nanoscience and Nanotechnology
companies in Spain and in particular the
importance of this market research,
product development, etc.
We would like to thank all the authors
who contributed to this issue as well as
the European Commission for the
financial support (project nanoICT No.
216165).
Dr. Antonio Correia
Editor - Phantoms Foundation
www.phantomsnet.net
CupCarbon simulator: Simulating the D-LPCN algorithm to find the boundary nodes of a WSN by Ahcene Bounceur, University of Bretagne Occidentale, Brest, France
07b. Nanotechnologies for diagnostics and nanomedicine
Lab on a chip: Miniaturization, Soft lithographies, microfluidics (Navier-Stokes equations, laminar flow in microchannels, main microfluidic components), Selected applications to chemical microreactors, separation systems and Lab On a Chip.
PhD thesis defense presentation for my topic "Improving Content Delivery and Service Discovery in Networks" for wireless and other networks. Columbia University, 2016.
Presentation slides for my PhD thesis dissertation on machine learning algorithm development to analyze multi dimensional genomic data such as microarrays
Luz rello - Ph.D. Thesis presentation - DysWebxia: A Text Accessibility Model...Luz Rello
Ph.D. Presentation
Title: DysWebxia: A Text Accessibility Model for People with Dyslexia
Author: Luz Rello
Advisors: Ricardo Baeza-Yates and Horacio Saggion
Abstract: Worldwide, 10% of the population has dyslexia, a cognitive disability that reduces readability and comprehension of written information. The goal of this thesis is to make text more accessible for people with dyslexia by combining human computer interaction validation methods and natural language processing techniques. In the initial phase of this study we examined how people with dyslexia identify errors in written text. Their written errors were analyzed and used to estimate the presence of text written by individuals with dyslexia in the Web. After concluding that dyslexic errors relate to presentation and content features of text, we carried out a set of experiments using eye tracking to determine the conditions that led to improved readability and comprehension. After finding the relevant parameters for text presentation and content modification, we implemented a lexical simplification system. Finally, the results of the investigation and the resources created, lead to a model, DysWebxia, that proposes a set of recommendations that have been successfully integrated in four applications.
This E-nano Newsletter special double issue
contains the updated version of the nanoICT
position paper on Carbon Nanotubes (CNTs)
summarising state-of-the-art research in this field
as well as a description of the possible electrical,
electronic and photonic applications of carbon
nanotubes, the types of CNTs employed and the
organisations or groups that are most proficient
at fabricating them.
In the second paper, the Nanoelectronics
European Research Roadmap is addressed
focusing on the main European Programmes
supporting the short, medium and long-term
research activities.
This issue also contains a catalogue (insert),
compiled by the Phantoms Foundation
providing a general overview of the
Nanoscience and Nanotechnology
companies in Spain and in particular the
importance of this market research,
product development, etc.
We would like to thank all the authors
who contributed to this issue as well as
the European Commission for the
financial support (project nanoICT No.
216165).
Dr. Antonio Correia
Editor - Phantoms Foundation
www.phantomsnet.net
CupCarbon simulator: Simulating the D-LPCN algorithm to find the boundary nodes of a WSN by Ahcene Bounceur, University of Bretagne Occidentale, Brest, France
07b. Nanotechnologies for diagnostics and nanomedicine
Lab on a chip: Miniaturization, Soft lithographies, microfluidics (Navier-Stokes equations, laminar flow in microchannels, main microfluidic components), Selected applications to chemical microreactors, separation systems and Lab On a Chip.
Applications of Computer Science in Environmental ModelsIJLT EMAS
Computation is now regarded as an equal and
indispensable partner, along with theory and experiment, in the
advance of scientific knowledge and engineering practice.
Numerical simulation enables the study of complex systems and
natural phenomena that would be too expensive or dangerous, or
even impossible, to study by direct experimentation. The quest
for ever higher levels of detail and realism in such simulations
requires enormous computational capacity, and has provided the
impetus for dramatic breakthroughs in computer algorithms and
architectures. Due to these advances, computational scientists
and engineers can now solve large-scale problems that were once
thought intractable. Computational science and engineering
(CSE) is a rapidly growing multidisciplinary area with
connections to the sciences, engineering, and mathematics and
computer science. CSE focuses on the development of problemsolving
methodologies and robust tools for the solution of
scientific and engineering problems. We believe that CSE will
play an important if not dominating role for the future of the
scientific discovery process and engineering design. The
computation science is now being used widely for environmental
engineering calculations. The behavior of environmental
engineering systems and processes can be studied with the help
of computation science and understanding as well as better
solutions to environmental engineering problems can be
obtained.
1. New Approaches Towards a Higher
Resolution Biomedical Imaging
Magnetic Induction Tomography
04/06/2010
Nuno Brás
PhD Thesis Presentation
Universidade Técnica de Lisboa
Instituto Superior Técnico
PhD Dissertation
Doctoral Program In Electrical and Computer
Engineering
António C. Serra and Raúl C. Martins (advisors)
2. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of the Art Issues
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
2
3. Outline
04/06/2010Nuno B. Brás - PhD Dissertation 3
! Motivation
! The Problem in Hands
! State of the Art Issues
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
4. Motivation
Imaging Systems
In Natural Sciences
Astrophysics; Biophysics; Geophysics, Biology…
In Engineering
biomedical; industrial processes; oil and water
prospecting; search and rescue instrumentation;…
04/06/2010Nuno B. Brás - PhD Dissertation 4
5. Motivation
Magnetic Induction Tomography
(as a Biomedical Imaging System)
Tomography - recreates maps or images from peripheral
measurements
! MIT is:
! Active Tomographic method (instead of Passive);
! Harmless (instead of Harmful)
! In-vivo or in-vitro;
! Functional or Steady Imaging
04/06/2010Nuno B. Brás - PhD Dissertation 5
6. Motivation
What does MIT try to solve?
Good things about MIT:
! (very) low price comparing with other tomographic systems
(e.g. MRI machine up to 2.5 million € + installation (25%))
! biological passivity (low power radiofrequency)
! excellent penetration abilities in biological phantoms, even in bone-
like tissues (unlike ultrasound tomography )
! full reconstruction is theoretically achievable
04/06/2010Nuno B. Brás - PhD Dissertation 6
7. Motivation
Not so good things about MIT:
1. The underlying image reconstruction problem is a
large, complex and non-linear problem (non-convex)
2. Its behavior is strongly dependent on the number of
measurements and their intrinsic accuracy
Actual Context
There are no commercial equipments, just prototypes.
First applications are being explored
04/06/2010Nuno B. Brás - PhD Dissertation 7
8. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
8
9. What is MIT, conceptually
It is a distributed parameter estimation (DPE) problem
under Electromagnetic Partial Differential equations
(PDE)
04/06/2010Nuno B. Brás - PhD Dissertation
The Problem in Hands
9
10. 04/06/2010Nuno B. Brás - PhD Dissertation
where is the kernel of the underlying physical process.
Consider an electromagnetic physical process ruled by the following relation:
The Problem in Hands
10
Distributed Parameter Estimation Problems in
electromagnetics
11. where is the kernel of the underlying physical process.
Electric and
magnetic
fields or potentials
(state variables)
Boundary
Conditions
+
Sources
Set Differential
Operators
+
parameters
04/06/2010Nuno B. Brás - PhD Dissertation
The Problem in Hands
11
Distributed Parameter Estimation Problems in
electromagnetics
Consider an electromagnetic physical process ruled by the following relation:
12. Distributed Parameter Estimation Problems in
electromagnetics
04/06/2010Nuno B. Brás - PhD Dissertation
where is the kernel of the underlying physical process.
The Problem in Hands
12
The inverse problem is the distributed parameter identification problem given by:
The forward problem (well-posed, and typically linear) is the PDE equation
Consider an electromagnetic physical process ruled by the following relation:
13. The General Parameter Estimation Model
(Outline of this Thesis)
Experimental Setup (PART 1)
The Forward Problem (PART 2)
The Inverse Problem (PART 3)
04/06/2010Nuno B. Brás - PhD Dissertation
Problem in Hands
13
14. The MIT Setup
04/06/2010Nuno B. Brás - PhD Dissertation
Source Currents: Harmonic , tens of
kHz to some MHz;
Current Amplitude: Up to 1 A;
Setup Radius: Typically 15 cm;
Conductivity Values: complex with
absolute value between 0.1 S/m to 2
S/m;
Source and sensing coils size: around
5 cm and 3 cm correspondingly
The Problem in Hands
14
19. 04/06/2010Nuno B. Brás - PhD Dissertation
The Problem in Hands
19
The MIT Setup
Each d should be as large, independent
and accurate as possible
20. Outline
04/06/2010Nuno B. Brás - PhD Dissertation 20
! Motivation
! The Problem in Hands
! State of the Art Issues
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
21. ! Experimental Issues
! Typical layout generates ambiguities (gradiometers);
! Fixed system of acquisition (fixed and low number is allowed)
! Forward Problem
! Developed solvers in MIT context are accurate but typically slow;
(it is not possible to use commercial solvers)
! Inverse Problem
! Non-linear approaches are the best known reconstruction methods.
! The state of the art method is not efficient when used with a high
number of acquisitions.
04/06/2010Nuno B. Brás - PhD Dissertation
State of the Art Issues
21
22. Outline
04/06/2010Nuno B. Brás - PhD Dissertation 22
! Motivation
! The Problem in Hands
! State of the Art Issues
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
23. 04/06/2010Nuno B. Brás - PhD Dissertation
Objectives - Developed Work
23
General Objective
Implementation of a new Magnetic Induction Tomography prototype
and numerical framework to deal with large number of acquisitions
Developed Work
! Experimental
! A new moving prototype was implemented for large number of
acquisitions while attaining the state of the art sensitivity (SCR)
! Forward Problem
! A new 3D eddy current PDE solver was developed with a
competitive processing time and low relative error.
! Inverse Problem
! A new ADMM method was developed and used in 2D and 3D IP.
! Its feasibility was proved and its advantage was clearly shown for
large datasets scenarios.
24. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
24
27. Experimental Work
04/06/2010Nuno B. Brás - PhD Dissertation 27
! Twin Coil Setup
Source
Current
Amplifier
Capacitive
Shields
Source
Coil
Motor
systems
28. 04/06/2010Nuno B. Brás - PhD Dissertation 28
Experimental Work
Involved Areas While Designing This Prototype
! Electromagnetic Compatibility (EMC)
! Mechanical Setup Design and Characterization
! Current Source Design
! Acquisition System
! Signal Processing
29. 04/06/2010Nuno B. Brás - PhD Dissertation 29
Experimental Work
Involved Areas While Designing This Prototype
! Electromagnetic Compatibility (EMC)
! Mechanical Setup Design and Characterization
! Current Source Design
! Acquisition System
! Signal Processing
Measuring System
30. 04/06/2010Nuno B. Brás - PhD Dissertation 30
Symmetry Axis
Without object:
V1-V2 = Residual V ≈ 0, for any angle
With object:
V1-V2 = ∆V
V1
V2
Experimental Work
! Twin Coil Setup
Useful V =∆V– Residual V
31. ! Shielded Differential coils and cables
! Gain (low noise) Amplifier =100; with a low-pass filter
! ADC with 12 bits, up to 60 MSamples/s
! The Goertzel Transform (second order IIR filter ) to calculate the frequency
bin amplitude and phase (others were tested)
! Average sliding window applied to avoid 50Hz modulation and reduce noise
! Principal Component Analysis (PCA) and a reference coil were used to
eliminate long term trends from the source current.
04/06/2010Nuno B. Brás - PhD Dissertation 31
Experimental Work
Acquisition System
Signal Processing System
32. 04/06/2010Nuno B. Brás - PhD Dissertation 32
Experimental Work
! Results a differential coil pair, during 300 sec time
33. 04/06/2010Nuno B. Brás - PhD Dissertation 33
Experimental Work
! Results - Stability
Applying an averaging sliding window with 50 values
A minimum measurable SCR of
was obtained which is the stated state of the art value in this case with
a moving system of sensing coils
34. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problem
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
34
35. The Forward Solver abilities:
! Isotropic parameter model
! Constant
! Diffusion model approach (no skin currents)
! Harmonic and stationary equations
! Multi- level grid allowing mutligrid (MG) techniques and/or
adaptive mesh refinements (AMR) in future iterations
! Analytical sources, using the Biot-Savart law
The Forward Problem
04/06/2010Nuno B. Brás - PhD Dissertation 35
36. Normal Component of = 0
The Forward Problem
04/06/2010Nuno B. Brás - PhD Dissertation 36
Formulation:
Imposed directly by (1)
(1)
in
Boundary ConditionsInterface Conditions
where and
37. The Forward Problem
The Discretization : Finite Integration Technique
04/06/2010Nuno B. Brás - PhD Dissertation 37
38. The Forward Problem
04/06/2010Nuno B. Brás - PhD Dissertation 38
The proposed FIT formulation:
! The resultant system is non-singular - the used gauging
ensures a robust regularization of the system;
! A Laplacian-type instead a curl-curl-type problem, where a
larger set of numerical methods are available;
! Subgriding implementation;
Moreover…
! Several Numerical Optimization aspects were implemented;
! Solved with iterative preconditioned methods (here, iLU
factorization was used);
42. The Forward Problem
How this interacts with the inverse problem?
04/06/2010Nuno B. Brás - PhD Dissertation 42
43. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problems
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
43
44. Inverse Problems
04/06/2010Nuno B. Brás - PhD Dissertation 44
! Implemented Methods :
! Two versions of the Alternating Direction Method of
Multipliers (Augmented Lagrangian Method) in an
elliptic 2D inverse problem with total variation
regularization and Wavelets regularization
! Gauss Newton in ellipitc 2D Inverse Problem
! ADMM version adapted to solve the MIT 3D
46. Inverse Problems
04/06/2010Nuno B. Brás - PhD Dissertation 46
The ADMM – a sequence of simple problems
Closed form
Either implementing a fixed point iteration for a
continuous approximation of
(ADMM with fixed point iteration)
or
Using a second Alternate Direction split
Closed Solution (DIPESAL Method)
48. Inverse Problems
04/06/2010Nuno B. Brás - PhD Dissertation 48
Gauss Newton using an
unconstrained version of
the problem
ADMM with a fixed
point iteration and
TV Regularization
ADMM with a fixed point
iteration and Wavelet
Regularization
51. Inverse Problems
04/06/2010Nuno B. Brás - PhD Dissertation 51
DIPESAL
ADMM with fixed point iteration
2.2% better reconstruction and 7% faster,
with better image reconstruction quality
5% noise and LARGER problem
52. Inverse Problems
04/06/2010Nuno B. Brás - PhD Dissertation 52
! Some remarks for 3D MIT inverse problem
! Two things changes for MIT problem:
! Size: Each Iterations of the ADMM problem have to be
solved iteratively: The Second Order Stationary method was
applied.
! Different discretization and equations: The new equations
were referred before;
60. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problems
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
60
61. Conclusions and Original Contributions
! Experimental Work
! A new prototype was implemented for moving sensing and source
coils, allowing acquire large sets of accurate data
! System sensitivity, a state of the art value, is stable during a large
amount of time (>300 sec), allowing to implement a moving setup
with high sensitivity.
04/06/2010Nuno B. Brás - PhD Dissertation 61
This work resulted in 7 congress papers during the PhD period
62. Conclusions and Original Contributions
! Forward Problem
! A new hybrid formulation of Finite Integration Technique
! The processing time was optimized to be included in an
inverse problem attaining a low relative error
! Total relative error ~ 1.5 %
! Processing time ~19 sec. per eddy current problem
04/06/2010Nuno B. Brás - PhD Dissertation 62
This work was published on IEEE Transactions on Magnetics
(May 2010)
63. Conclusions and Original Contributions
04/06/2010Nuno B. Brás - PhD Dissertation 63
! 2D Elliptic Inverse Problem
! A new method ADMM algorithm (DIPESAL) was implemented
with closed solution for Total Variation regularization
! Lower relative error (2,2%);
! 7% faster.
! Better qualitative images in general
This work was submitted to IEEE Transactions on Image Processing
(May 2010)
64. Conclusions and Original Contributions
04/06/2010Nuno B. Brás - PhD Dissertation 64
! 3D MIT Inverse Problem Solution
! The ADMM algorithm feasibility in 3D MIT
problems was achieved
! A complete new approach with clear advantages in large
number of measurements conditions.
! Its application to the MIT originates state of the art
simulated reconstructed maps.
This work was submitted to IEEE Transactions on Medical Imaging
(May 2010)
65. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problems
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
65
66. Further Work
04/06/2010Nuno B. Brás - PhD Dissertation 66
! Experimental
! Use large arrays of Giant Magnetic Resistors as sensors – This can
dramatically change MIT since it can increase dramatically the number
of acquisitions and their accuracy.
! Use strong permeability material cores to increase magnetic field.
! Use of transient signals to increase SNR
! Numerical
! Multigrid scheme for preconditioning and code parallelization to
accelerate even more the forward and inverse problem.
! Field Regularization – there is very experiments clearly shows
advantages in using this
! Improve ADMM to MIT – There is space to improve, namely in the
regularization process
67. Outline
04/06/2010Nuno B. Brás - PhD Dissertation
! Motivation
! The Problem in Hands
! State of The Art
! Objectives and Developed Work
! Experimental Work
! Forward Problem
! Inverse Problems
! Conclusions and Original Contributions
! Further Work
! Acknowledgements
67
68. Acknowledgements
! Supervisors: A. C. Serra And Raúl C. Martins
! Other professor and researchers
! Professor José Bioucas Dias (Elect. and Comp. Dep.)
! Professor Artur Lopes Ribeiro (Elect. and Comp. Dep.)
! Professor Paulo Martins (Mechanical Dep.)
! Professor Helena Ramos (Elect. and Comp. Dep.)
! Dr. Tomas Radil (Elect. and Comp. Dep.)
! Professor Pedro Santos (Mathematics Dep.)
! Professor Carlos Alves (Mathematics Dep.)
! Eng. Luis Soares
! Eng. Alexandre Pestana
! Eng. José Gouveia
04/06/2010Nuno B. Brás - PhD Dissertation 68
69. New Approaches Towards a Higher
Resolution Biomedical Imaging
Magnetic Induction Tomography
04/06/2010
Nuno Brás
PhD Thesis Presentation
Universidade Técnica de Lisboa
Instituto Superior Técnico
PhD Dissertation
Doctoral Program In Electrical and Computer
Engineering
António C. Serra and Raúl C. Martins (advisors)