Critical exponents of the phase transition have been associated with the intermittency and the corresponding factorial moments [X] [X].
Scientific
Work
Achievement: The death of endeavor and the birth of disgust.
Ambrose Bierce,
The Enlarged Devil’s Dictionary
You’ll come to learn a great deal if you study the
Insignificant in depth
Odysseus
Elytis, Greek
poet
The
low energy nuclear era
Electromagnetic and nuclear transition matrix elements for the Zr Isotopes
My first
scientific activity is connected with the accomplishment of my
diploma thesis at the Max-Planck Institute of nuclear physics
Heidelberg in
Germany. I was a member of the Q3D group of Dr. C. A. Wiedner. Using
a Q3D magnetic spectrograph our group studied the reaction Zr(a
,
a*)Zr.
Differential cross sections of low-lying excited states of the even
Zr Isotopes 90,92,94 and 96 were measured [X]
[X]. From DWBA (Distorted
Wave Born Approximation) calculations it was possible to assign spin
and parity for a large number of excited states. It was therefore
possible also to clarify previously contradictory results of similar
experiments. Additionally I observed a Coulomb Nuclear Interference
(CNI) effect which is clear visible for the low lying 2+
and 3- states. Using this effect our Q3D group studied the
isospin dependence of the cross section for these states. The
contribution of neutrons was larger than expected [X].
The results of my diploma thesis have been published in two articles
in Nuclear Physics A and one in Zeitschrift für Physik A.
Various experiments have been performed later and the results have
been finally been confirmed showing a strong dependence with
increasing neutron number that current direct nuclear models cannot
explain.
The
high energy nuclear era
In March 1987
I did a study at the Institute of
High Energy physics
in Heidelberg, about the possibility of using Cerenkov counters in a
Kaon magnetic spectrograph for a GSI experiment. This study was
completed with an internal report. In September 1987 I joined the
NA35 Experiment in CERN, Geneva. Aim of this experiment was the study
of the properties of nuclear matter under extreme conditions of
pressure and temperature. A special interest was the production and
detection of a transition of nuclear matter to a quark-gluon plasma
state. In comparison to energies of 10 MeV per nucleon in my diploma
thesis now 200 GeV per nucleon were at least theoretical necessary to
produce the energy density for the deconfinement.
Transverse
momentum spectra
My
first work was to analyze the transverse momentum spectra of the
produced negatively charged pions. I compared these with spectra of
pions produced in p-p collisions at a comparable energy. The
difference of the transverse momentum spectra was found to be large.
The pion spectra could not be described by a single stationary
source. An excess of pions with a small transverse momentum was
found. The events are quite complex since more than 300 particles are
produced in central collisions of O-Nuclei, with a total kinetic
energy of 3.2 TeV, on stationary Gold nuclei.
Search
for fluctuations
In
1988 the CERN SPS accelerator complex provided beams of 32S
nuclei with a 6.4 TeV total kinetic energy. The mean total
multiplicity of produced particles was then 600-700. Different models
like Lund-Fritjof or Venus (which are phenomenological without a
phase transition incorporated), did not succeed to describe the
transverse momentum spectra. Since it is probably that only in a
small fraction of events a Quark-Gluon plasma state is formed, there
is a possibility that taking indiscriminately all events, possibly
signals of the formation of a Quark-Gluon plasma will be attenuated
or even worst smeared out of existence. Therefore motivated by Leo
van Hoves Model describing the hadronization of Quark-Gluon plasma
with effects visible in the rapidity spectrum, I used the factorial
moment method to filter events as possible candidates. This was one
of the first uses of factorial moments in a heavy ion experiment. The
result was that the factorial moments are well suited for the study
of fluctuations. No clear intermittency was observed in the first
one-dimensional analysis. Only few events have been observed to have
the required rapidity distribution showing a slightly larger than the
average transverse momentum, a prediction also of van Hoves model.
The very poor statistical significance, due to the small number of
available events showed that intermittency was very small. As the
statistic later has been increased it has been found that it can be
explained by Bose-Einstein correlations.
Bose-Einstein
correlations
1989
I was a Research Assistant at the Institute of nuclear physics in
Frankfurt. There I started my work on Bose-Einstein Correlations in
Heavy Ion Collisions. With this method the space-time geometry of a
Boson source can be measured experimentally. Many aspects of Boson
correlations for static and dynamic expanding sources and first
experimental results have been the result of my doctor thesis "A
study of Bose-Einstein Correlations in central O+ Au and S+S
collisions at a projectile energy of 200 GeV per nucleon".
The results suffering from
statistics showed a quite large and therefore long lived pion sources
at mid-rapidity formed in the collisions of ultrarelativistic nuclei.
The source was found to expand parallel to the beam axis very fast
showing relativistic effects. The longitudinal expansion is similar
to that of the Big Bang. Some of these observations were influenced
by wrong track measurements and statistical fluctuations. My thesis
considers also three pion correlations and tried to explain it in
terms of two-particle correlations. It included a theoretical
analysis from an experimentalists view. It was clear that
relativistic effects were present due to the longitudinal expansion.
It took years until dedicated experiments using electronic methods
have acquired enough statistics to produce reliable results. My
conclusions that the Bose-Einstein effect would rapidly evolve
theoretically and experimentally became reality. I mentioned the
effects of protons may have in the correlation spectra. Later this
was observed although not expected. Some ideas about the Coulomb
correction and correlations between pion pairs of opposite charge
have been continuously been developed and new correction methods
later have been proposed.
With the
observation of strangeness enhancement, fluctuations in multiplicity,
excess of particles with small transverse momentum and large and long
lived pion sources, NA35 has provided a large number of phenomena
that still wait for a theoretical explanation.
Additional
from measurements of the transverse energy, an energy density of 2-3
GeV/fm3 in the first moments of the collisions was found,
this is within the limits where theory predicts a phase transition.
The results of the NA35 Experiment have been published in various
articles in Zeitschrift für Physik C, Nuclear Physics A, and
Physical Review Letters [X].
I was
interested in the analysis of Bose-Einstein Correlations such as in
Pb-Pb collisions, where the multiplicity is so large, that multiboson
effects could be visible in various spectra. For multiplicities of
2000 particles on an event by event basis the source size can be
measured from Bose-Einstein Correlations. New optical methods as the
Speckle Interferometry can be used for the first time in High Energy
Physics for these events (and in particular at LHC Energies). From
the study of Bose-Einstein Correlations of bosons with different
freeze out times, e.g. photons produced in the high density state,
and kaons, pions and the corresponding particle spectra, it will be
possible to find the temporal and spatial evolution of excited and
dense nuclear matter formed in high energy collisions.
It is also
not excluded that multibosons effects play a significant and
important role in fluctuations, and therefore being even responsible
for the formation of galaxies, or for the small fluctuations observed
in the cosmic background radiation of approximately 3 K.
During my
thesis and later there was a continuous discussion whether the
results are due to a quark-gluon plasma formation. It took more years
after I stopped working in this field that people agreed that the
phase-transition actually takes place. The European center of high
energy physics, CERN, announced that the evidence of the creation of
the plasma state is compelling. For a short period I occasionally
worked with a theoretical group from Athens which
considered the effects of the strange quark mass on the order of the
phase transitions and the associated fluctuations in rapidity
density.
Critical exponents of the phase transition have been
associated with the intermittency and the corresponding factorial
moments [X] [X].
The study of a phenomenon such as the QGP continues until now and probably will last for another 10-20 years. It needs still some more experimental and theoretical work and probably study of the nuclear matter properties at higher and lower energies to obtain a clearer picture of the transition. I was quite at the begin of this high energy nuclear physics area, but my attitude is not to spent more than approximately 5 years in one field as there is always the danger for "overspecialization" or as the Germans say to become a "Fachidiot". Of course today as things are more complex it is sometimes necessary to become an expert in "nothing".
Some experience I have gained from discussions of theoretical and experimental issues of holographic interferometry with an old friend, Evangelos Madadakis from the Forschungszentrum in Karlsruhe in Germany, which used it for the study of transport coefficients in fluids.
From
3D ultrasound to radiotherapy treatment planning problems
3D
ultrasound
Down
back to the earth I improved my computer skills and made a transition
from the stone-age FORTRAN to C and C++. My
first non-scientific work was the development of a Windows NT version
from an experimental UNIX version (and therefore of course not
commercial J) for the volume rendering
software InViVo (Interactive visualization of volumetric data) from
the Fraunhofer computer graphics institute and especially the 3D
ultrasound acquisition version known now as InViVoScanNT [X]
.
While other volume rendering systems such as VTK are quite
sophisticated InViVo is one of the fastest volume rendering programs.
Together with Michael Richtscheid, Markus Grimm and Stefan Walter
this version, first developed by G. Sakas in UNIX, now is
commercially available from MedCom.
For a short period I worked together with George Ioannidis in
Offenbach from
NTUA in Athens in
image distortion corrections and mosaicking problems.
HDR
brachytherapy
After
one approximately year in the Fraunhofer institute in Darmstadt,
Germany, I started working for medical physics problems. This
opportunity I had by Prof. D. Baltas from the Medical Physics and
Engineering department of Klinikum Offenbach. I looked at the problem
of the optimization of dose distributions in brachytherapy
and in external beam radiotherapy. Only in the USA more
than 600000 persons from a total of more than 1 million with cancer
must be treated by radiation therapy. Worldwide more than 100 million
persons have cancer. While there is a tremendous progress in
molecular biology a solution of the cancer problem still requires
probably a few decades (who knows?) and still radiotherapy plays an
important role in cancer treatment. The goals of conformal radiation
therapy are to limit the high dose to the cancer volume while
simultaneously protecting healthy tissue. Additional sensitive
structures in the body must be protected from excessive radiation.
These multiple goals have been until currently been added into a
single objective scoring function. Goal of the optimization is the
determination of intensity profiles or temporal and spatial
distributions of radioactive sources which produce dose distributions
which satisfies all the objectives and constraints. The objectives
have been added into a single term using so called importance factors
which to some extend are arbitrary and their influence on the result
is not known. Treatment planning systems still provide only this
method for optimization. The treatment planner either accepts the
solution or by trial and error tries to find an acceptable solution.
After the
work of O. C. Haas who used multiobjective algorithms for solving
radiotherapy optimization problems with geometric methods I presented
the first true a posteriori multiobjective dose optimization
methods for radiation therapy [X]
.
The interest in multiobjective optimization is increasing both from a
mathematical viewpoint but more for solving practical scientific and
engineering problems. Most of the true problems are multiobjective
problems with not only a single but many competing solutions. One
special class of such solutions is the so called Pareto optimal or
efficient solutions. In cooperation with Prof. D. Baltas who has
experience in the field of high-dose rate brachytherapy we are in the
stage of propagating the true multiobjective optimization problem in
radiotherapy and especially in brachytherapy.
Other
problems which I considered were methods for increasing the quality
of solutions in brachytherapy which includes studies for dose
calculations using Fast Fourier transforms and the convolution
theorem
[X],
stratified sampling methods
[X]
and dose calculation methods based on Delaunay and Voronoi graphs.
Additional some computational geometry based methods have been
developed [X] [X]
to
improve the distribution of sampling points for the calculation of
dose volume histograms. Part of this work has been obtained from the
work of Thorsten Kemmerer from the University of applied sciences in
Frankfurt/Main Germany, Maria Papagianopoulou from Patras University
in Greece and Kostas Karouzakis from the National Technical
University Athens in Greece which I supervised to a great extent
their diploma thesis. These studies have been published in the two
most important medical physics journals: Medical Physics and Physics
in Medicine and Biology.
The European
project MITTUG (Minimally Invasive Therapy for Tumors 3D Ultrasound
Guided) in cooperation with the Nucletron company combines together
the experience in brachytherapy of the medical physics department of
the Strahlenklinik Offenbach (Germany), the 3D ultrasound imaging
experience of the Fraunhofer/MedCom company and my experience for
multiobjective anatomy based dose optimization in high-dose rate
brachytherapy.
I have
developed WinOpt-HDR a Microsoft-Windows based experimental toolkit
for multiobjective dose optimization in HDR-brachytherapy. Kostas
Karouzakis has written a significant part of WinOpt-HDR. I developed
methods to perform fast multiobjective dose optimization with
deterministic or evolutionary algorithms
and
fast calculation methods.
The algorithms developed have been implemented in the treatment planning system for prostate brachytherapy SWIFTTM provided by the Nucletron Company. Everyday patients are now treated by this system that uses deterministic and multiobjective evolutionary algorithms. The algorithms have been tested and further improved by the Pi-Medical company. Not only a higher quality such as conformity of the dose distributions has been achieved but also the treatment time has significantly been reduced. Since Nucletron is a world leader in this field the system will soon be used worldwide.
My hope is
that other treatment planning systems for HDR- and LDR- brachytherapy
will adopt this methodology which produces solutions of superior
quality than other algorithms and does not require arbitrary
importance factors for the various objectives.
External
beam radiotherapy
In
cooperation with Cristian Cotrutz from the University
of Patras
in Greece
and now at
Stanford University
we did a
first study of the use of a gradient based algorithm for dose
optimization in intensity modulated beam radiotherapy (IMRT)
[X].
Using a simply mapping technique we have eliminated the need of
corrections or inclusions of constraints for non-physical solutions
with negative values which naturally emerge from gradient and
deterministic in general optimization procedures. We have applied
this algorithm for a true multiobjective optimization method by
studying for the first time Pareto fronts of the objective space.
Multiobjective
optimization problems in general can not be solved by deterministic
algorithms. Medical and industrial applications require robust and
reliable algorithms. Evolutionary algorithms (EA) are becoming more
stable and their performance has reached a quite impressive level.
Multidimensional
problems in radiotherapy involves as many as 300 parameters to be
optimized has allowed us to study the efficiency of various
evolutionary algorithms. In terms of probability distributions of
solutions in the objective space in some cases we have estimated the
performance of evolutionary algorithms in comparison to random search
methods or deterministic gradient methods if they can be applied to
the problem. These results have been published in the Lecture Notes
in Computer Science series
[X]
[X]. For IMRT the number of
parameters may exceed 5000.
We have
emphasized the importance of the initialization and the use of hybrid
algorithms. Additional insights we hope to have from comparisons of
Pareto and local search based methods. Our first study showed that
for the intensity modulated beam radiotherapy problem standard
evolutionary algorithms are not suited. Our hope is to combine the
best properties of both into a more powerful hybrid method.
End of 2002 I
started again my studies of multiobjective optimization algorithms
for external radiotherapy with Eduard Schreibmann. We presented first
systematic results of the use of gradient based optimization
algorithms and showed that the results for variance based objectives
are global optimal. We compared results with zero and clinical
acceptable critical values for the OARs. Gradient based algorithms
like L-BFGS are very fast and very robust.
I was
involved in the optimization with geometric objectives developed by
Eduard Schreibmann and Prof. Dimos Baltas. Studies have shown that
the geometric and dosimetric objectives are correlated and thus
geometric methods could be used in the beam orientation optimization
or optimization for arc therapy
.
A hybrid evolutionary algorithm for multiobjective inverse planning
for IMRT was developed that provides a representative set for
non-dominated solutions that can be used for the selection of a
optimal number of beams, their orientation and which is independent
of importance factors
.
Cancer
Statistics modeling
With Dr. R. F Mould I started 2004 a cooperation extending his work on Cancer Statistics and modelling. Dr Mould was able to show that cancer can be cured, i.e. that the patient who had cancer and were cured could not statistically be distinguished from healthy patients (The idea was that these patients will die if not from the cancer they had from other causes more likely than other healthy persons who did not have cancer), and that it is possible to reduce the follow up time. Usually the old methodology required to consider a large number of patients for many years to determine the effects of drugs and treatment methods. With the new statistical methods it is possible to reduce this time and the costs of cancer treatment. Our first work considered the statistical modeling of mesothelioma cancer patients. Using statistical data collected (the biggest ever) we considered the dependence on age, sex and region and other dependencies. This study provided a reference base line, not only using the median, that could be used to control new treatment methods for this absolute deadly disease. The work will continue considering other cancer types.
The
future?
My hope is that the multiobjective approach will be used also in IMRT and radiation therapy in general and will replace the guesswork of the past. I also hope to be able to have the opportunity to contribute in some other fields even if it is more likely to qualify as a "jack of all trades," than a master of any; But as Shakespeare? says: I want “..to read a little more from the Nature's infinite book”s” of secrecy” (and not only one)..