Earth Sciences & Env.
Engineering & Tech.
Information & Comm. Tech.
Life Sciences & Biotech
Use this search facility to find out more about the profile of our HPC-Europa2 visitors, the type of work they have been doing, and their project achievements.
The main objectives of the participation in the HPC-Europa2 program were the efficient execution of Monte Carlo Simulations for medical purposes using anthropomorphic phantoms.
In the work plan described in the project, first main goal was the successful installation of GATE Monte Carlo toolkit on the CPUs provided by the HPC. The benchmarks of GATE would be the first simulations that would run in order to justify the correct installation of the GATE and all appropriate programs – libraries. This would verify that GATE is properly working on the platform of the project.
Realistic Monte Carlo simulations for medical purposes using Geant4/GATE are highly computationally demanding. In order to minimize the simulation time Variance Reduction Techniques will be assessed, to optimize performance. Initially the accuracy of simple VRTs will be assessed. These methods will include:
i) variations in phantoms voxel size, ii) reduced number of phantom slices, iii) emission of the photons in a limited angle iv) emission of photons only towards the direction of the gamma camera, in case of SPECT v) emission of back to back photons using in a limited solid angle towards the PET ring in case of PET. In addition more advanced VRTs will be used such as:
i) Variations in random number generator ii) Photon splitting iii) Forced detection
After these initial objectives, the main goal of the project is to create a database of simulated data using NCAT and MOBY phantom. Clinical data will be used in order to attach the computational phantoms on the characteristics of the clinical one. The database will include a set of simulations comparable to each clinical situation. For this:
Initially typical normal exams will be simulated. Radioactivity will be assigned to basic organs or structures (e.g bones, heart) and proper radioisotopes will be used. The aim is to obtain indicative data regarding the normal exams. Different sizes of patients and organs will be simulated. ii) Then typical abnormal exams will be simulated. Radioactivity will be assigned to basic organs or structures (e.g bones, heart) and proper radioisotopes will be used; In addition organs parameters will be changed to model some typical abnormalities. As an example part of the heart will be blocked or alterations in respiratory or heart rate will be introduced. The aim is to obtain indicative data regarding abnormal exams. Again, different sizes of patients and organs will be simulated, as well as different levels of pathologies. iii) Moreover, small tumors of various sizes and shapes will be introduced in the NCAT or MOBY phantom and different levels of radioactivity will be assigned to them. The aim is to form a data set of tumor data, based on literature input. This set will be further used to improve imaging protocols.
Finally, an important objective of the visit is to explore further cooperation opportunities between proposers and host institutions. The interest in simulation of real preclinical and clinical exams is really significant. Besides protocols optimization there is major concern regarding dosimetry both in small animals and humans. In the case of animals –and especially in repeated studies- the amount of radioactivity may affect results. In addition small doses are required to ensure minimum interaction; thus it is critical to define total dose in several organs in different imaging scenarios. Although such ideas have already been implemented and explored by the proposer they will not be included in this project. However, their future realization will be investigated, as well as potential interested partners of the host institute.
Part of the objectives that were included in the proposal of the project have been achieved. More specifically, the first goal of installing GATE in the platform provided by HPC-Europa2 program was achieved successfully. In the beginning there were some problems, which were identified when running the benchmarks of GATE. The technical support team of the BSC made an excellent work, by re-installing the program of GATE and all the other appropriate programs and libraries (CLHEP, Geant4, ROOT), taking into account the error messages that were recorded.
The simulation hours that were provided to our project (equal to 30000 hours) were not enough in order to achieve the major goal of creating a full database of whole body NCAT simulated data. For this reason, we decided to crop the NCAT phantom and simulate only specific bed positions. This reduces dramatically the computational time needed. For a whole body PET acquisition using the NCAT phantom it was precalculated that about ~24000 hours would be necessary. The simulated acquisition would last for 36 min at, in 18 bed positions. The voxel size of the NCAT phantom was 4x4x4 mm3.
The next step was to reduce the number of slices of the NCAT phantom, so as to make an acquisition of 10 bed positions for the whole body, which was also quite demanding. A first approach was to reduce the activity used in the simulation. A first trial was done by using the 1/10 of the total activity in the phantom. Processing the results we decided that reducing the activity, resulted to simulations with quite low statistics. This means that the reconstruction of the images were not realistic, and not comparable to the real clinical data. We decided to put all effort in order to standardize the method of obtaining realistic results, and to utilize for this purpose the computational time that was provided by the HPC-Europa2 program.
We decided to do all these tests by using only a small part of the NCAT phantom. More specifically 20.8 cm of the whole body was used, while the field of view (FOV) of the PET scanner is about ~15 cm. So we could have a realistic simulation of one bed position of the NCAT phantom, including coincidences that come outside the FOV.
The simulated results were reconstructed with the FBP and OSEM algorithm of STIR software, adopted to the Biograph-6 PET system, of our simulation. For this reason we changed the voxel size and the number of the slices of the NCAT phantom, as STIR software required. Having done all these configurations, we implemented a simulation of the NCAT phantom, without having any kind of abnormalities or tumors, a so called “healthy” phantom.
We repeated the same simulation by using the same anatomy of the phantom and inserting a simulated tumor using 2 different approaches. Firstly, the tumor was considered as a homogeneous structure having the same activity distribution allover its volume. Secondly, the tumor was inserted taking into account the variability of the tumor, at voxel-by-voxel level.
Concluding, the main achievement of the HPC program is that GATE toolkit was correctly installed and the method for using computational phantoms, in clinical acquisitions, such as NCAT was standardized.
Moreover, the visit at the host institute, (Prof G. Kontaxakis), provided a great experience on how to use and handle phantoms like NCAT and MOBY. The cooperation among our laboratory and their institute still exists, and we are very encouraged to do more future work related to projects of common interest.
To summarize, GATE Monte Carlo toolkit was installed for the first time in the BSC (Barcelona Supercomputing Center – MareNostrum). So, installing GATE in a HPC center provides a very good opportunity for using this simulation tool for further studies related to medical physics (e.g. Nuclear Medicine, Radiotherapy-Dosimetry, Brachytherapy etc). Having it available allows more users to setup their own simulations.
Nontoxic drug delivery systems for efficient transmembrane transport are central issue in the successful therapy of a number of diseases. Appropriate building blocks of reversible drug carrying micelles are soluble surfactants possessing oligo(ethylene glycol) hydrophilic heads and alkane hydrophobic tails (Cy(EO)x). The use of a coarse-grained (CG) model for these systems allows extension of our all-atom work to much larger systems and longer time-scales. The project aims include also finding the structural aspects of the amphiphilic organization in aqueous medium by generating sufficiently long MD CG trajectories and then processing the simulation data for models of spherical micelles built of different numbers of surfactant molecules in explicit solvent (water). Examination of the results for the selected micelles permits the establishment of size-stability relationship. Also the probability of surfactant exchange between micelles or the possibility of micelles fusing could be investigated. On the other hand, the capacity of nonionic surfactants C12(EO)5 to pack antimicrobial peptides (AMPs) could be considered at the coarse-grained level.
The objectives of the project have been fully achieved. The structural aspects of the amphiphilic organization in aqueous medium are obtained from the simulation data for models of spherical micelles consisting of a various number of C12(EO)5 molecules in water solvent. Information about the micelle structure formed by C12(EO)5 and the surfactant organization therein, i.e., the shape, stability and water penetration into the hydrophobic core is acquired on the basis of radius of gyration and radial distribution functions. Analyses of the CG trajectories for the selected micelles determine the size-stability relationship and a typical aggregation number, which corresponds very well to the experimental one. For a more comprehensive description of the process of micelle formation, a self-assembly process, which starts from a configuration where the surfactants are randomly distributed in the water solution, is simulated. The results thereof are in very good agreement with those obtained with the other models. Finally, a small AMP is incorporated in the most stable micelle and subject to CG MD simulation. Alongside with analysis of structural features and position of the AMP in the micelle, the stability of the formed (C12(EO)5)n/peptide complex is evaluated by means of binding energy assessment.
In the upcoming studies we plan to simulate the transport of the latter (AMP in a micelle) across a lipid bilayer at the coarse-grained level employing the MARTINI CG force field. This will provide insight into the interaction of the micelle with the bilayer and the various structural transformations of the separate components of the system taking place due to the presence of the micellar aggregate.
Drug delivery systems are extensively studied due to their possible use in anticancer therapy. We proposed very interesting solution, the so called “nanocapsule” - nanodevice composed of open-ended carbon nanotube and two ferromagnetic nanoparticles attached to its tips. The aim of our research was to study influence of its structure on properties and behaviour under external magnetic field. We studied behaviour of such nanocapsule composed of single-walled carbon nanotube (25,0) and ferromagnetic nanoparticles by means of Monte Carlo simulations. The results have shown very promising properties, that is capping the nanotube in the absence of EMF and uncapping when large enough EMF is applied. Moreover, applying lower EMFs give no change of the structure, making it possible to use low EMFs for targeted delivery.
The next step of our research was to study properties of similar structures. Finding optimal structure of nanocarrier is very important for future synthesis. Therefore, we focused on studies of the dependence of the nanocapsule properties on its parameters, i.e. type of nanotube (its diameter, chirality), size and magnetic state of nanoparticles, how nanoparticles are bound to the nanotube tips, etc. We were also interested in dynamic properties of the nanocapsule under the external magnetic field. Time-scale of the structural transformation is the crucial parameter making the nanocapsule applicable as drug delivery system. Behaviour of the nanocapsule under the alternating magnetic fields was also of our interest. Therefore, the Molecular Dynamics code with the implemented Brenner potential was being developed for such purposes. During my visit I planned to finish its development and optimize it. Next step would be code parallelization using OpenMP.
Further, with help of Dr Philip Camp, I planned to choose appropriate approach to study the behaviour of such a nanocapsule when composed of superparamagnetic nanoparticles and modify the MD code accordingly.
I started research as planned and worked on developing and optimizing the Molecular Dynamics code. Using the software available on Hector I managed to optimize it. Improvement of the most time-consuming part of the code made the whole code worked more than 10x faster.
Discussions with Dr Philip Camp resulted in choice of the appropriate potential for description of magnetic interactions of superparamagnetic nanoparticles with the external magnetic field and modification of the MD code accordingly.
Preliminary calculations based on the MD code revealed an enormous sensitivity to the choice of the initial structure of the nanocapsule which resulted in the program crash. After performing careful debugging and testing we stated that the best approach will be to find a relaxed structure by means of Monte Carlo simulations of to perform the total energy minimization before the MD run. Finally we decided to implement MC code which additionaly gave us a notion about the influence of size parameter on the behaviour of the nanocapsule.
Additionally, I started to use different, more realistic potential to describe nanoparticles-nanotube interactions. Previously, they were described by Lennard-Jones potential which is regarded as crude approximation in this case. Currently the Hamaker potential is implemented in the code. Results obtained from the MC simulations for both potentials seemed to be very promising, therefore we kept using this approach to obtain more information about the influence of various factors on the properties and behaviour of the nanocapsule with and without the external magnetic field. I determined how the size and type of the nanoparticles and type of the nanotube affects the nanodevice behaviour. I also developed an auxiliary code that is capable of creating various structures of nanodevice, making possible to prepare a lot of simulation runs quickly. These studies will be continued and their results will be published.
Moreover, as soon as that part of research will be completed, we will continue to study the dynamic properties of the nanocapsule.