• Research of magnetic field deformations caused by weak magnetic materials using NMR imaging methods
• Pair-wise temporal pooling method for rapid training of the HTM networks used in computer vision applications
  
• Study of the influence of ECG electrodes position inaccuracy on the localization of ischemic lesions
  
• Concentration measurement of magnetic nanoparticles in biological tissues
  

Result 1: Research of magnetic field deformations caused by weak magnetic materials using NMR imaging methods

Projects: VEGA  2/0090/11 and ŠPVV 2003SP200280203, project ASFEU CEKOMAT, ITMS No. 26240120006 (Ústav merania SAV).

Authors: I. Frollo, P. Andris, J. Přibil, D. Gogola, L. Valkovič, P. Szomolányi

 Objects inserted into the homogeneous magnetic field of an NMR tomograph, its magnetic properties (susceptibility), cause deformations of this field that are possible to measure and image using special sensor. Methods of imaging of magnetic field deformations are able to image very small magnetic changes in biological or physical samples. These changes are influenced by stationary homogeneous magnetic field of the imager, radiofrequency field, impulse gradient magnetic field and acoustical vibrations, that arise under the influence of magnetic gradient impulses.
For weak magnetic material samples a dispersive magnetic fields were theoretically calculated and tested on NMR imager 0.178 T. A homogeneous phantom (reference medium) was realized - fill up with water solution. Resultant images represent magnetic field changes in a homogeneous phantom.
For detection of these changes new sequence Gradient-echo that is sensitive to small changes of magnetic field was suggested. An analysis of frequency properties of acoustical noise for diversely types of 3-D MR sequences was executed.
The method was tested on small samples. It will find its application in further research of influence of magnetic field on living organisms and in research of new magnetic material.

Pic. 1. Left: NMR sensor - horizontal radio-frequency coil and plastic holder (homogeneous phantom), working area diameter 90 mm.
Right: NMR image of a material sample - polymer fiber, thickness 0.14 mm, sample were impregnated with layer of Fe3O4 nanoparticles.

Publications:

• FROLLO, Ivan - ANDRIS, Peter - GOGOLA, Daniel - PŘIBIL, Jiří - VALKOVIČ, Ladislav - SZOMOLÁNYI, Pavol. Magnetic field variations near weak magnetic materials studied by magnetic resonance imaging techniques. In IEEE Transactions on Magnetics, 2012, vol. 48, no. 8, p. 2334-2339. ISSN 0018-9464. (1.363-IF2011)
• ANDRIS, Peter - FROLLO, Ivan. Measurement of magnetic field with background using a low field NMR scanner. In Measurement Science and Technology, 2012, vol. 23, no. 6, art. no. 065006. ISSN 0957-0233. (1.494-IF2011) 
• PŘIBIL, Jiří - PŘIBILOVÁ, A. - FROLLO, Ivan. Analysis of spectral properties of acoustic noise produced during magnetic resonance imaging. In Applied Acoustics, 2012, vol. 73, p. 687-697. ISSN 0003-682X. (1.050-IF2011) 
  

Result 2: Pair-wise temporal pooling method for rapid training of the HTM networks used in computer vision applications

Projects: VEGA 2/0019/10 (Institute of Measurement Science, SAS), PhD thesis project supported by „e-talent“ grant of Tatra bank foundation N. 2010et019.

Authors: S. Štolc, I. Bajla, K. Valentín, R. Škoviera

In 2012, we have proposed several modifications to the conventional learning algorithms of the Hierarchical Temporal Memory (HTM), a bio-inspired large-scale model of the neocortex. Firstly, an alternative spatial pooling method has been introduced, which makes use of a random pattern generator exploiting the Metropolis-Hastings algorithm. The main contribution of the research consists in the proposal of a novel temporal pooling method –the pair-wise explorer – which allows faster and more reliable training of the HTM networks in static images. While the conventional temporal pooler trains the HTM network on a finite segment of the smooth Brownian-like random walk across the training images, the proposed method performs training by means of the pairs of patterns randomly sampled from a virtually infinite smooth random walk. We have conducted a set of experiments with the single-layer HTM network applied to the position, scale, and rotation-invariant recognition of geometric objects. The obtained results provide a clear evidence that the pair-wise method yields significantly faster convergence to the theoretical maximum of the classification accuracy with respect to both the length of the training sequence and the number of training patterns. The pair-wise explorer is particularly recommended, when the exhaustive training is impossible due to the complexity of the given task.

Fig. 1. The left picture represents an illustration of the principle of the image pattern codebook generation using Metropolis-Hastings algorithm. On the right: -graphical presentation of an example of the HTM network training by the method of temporal pooling of image patterns randomly sampled from a virtually infinite Brownian-like random walk across the training images.

Publications:

Result 3: Study of the influence of ECG electrodes position inaccuracy on the localization of ischemic lesions

Projects: VEGA 2/0210/10 „Methods and systems for multichannel measurement and evaluation of bioelectric signals of the heart and brain", APVV-0513-10 „Measuring, communication and information systems for monitoring of the cardiovascular risk in hypertension patients"

Authors: J. Lenková, J. Švehlíková, M. Tyšler

The influence of inaccuracy of positions of ECG electrodes on the chest on the result of the noninvasive localization of ischemic lesions using solution of the inverse problem of electrocardiography was analyzed. Twelve ischemic lesions and corresponding ECGs on the surface of a realistic torso were simulated. The locations of ischemic lesions were then determined by inverse computations using 3 inhomogeneous torso models: 1. Realistic model with positions of 62 electrodes obtained from MRI, 2. General model with shape adapted according to the real torso dimensions and with electrodes placed according the lead set definition, 3. Model with the same torso shape as in model 2 but with real electrode positions as in model 1.
The mean lesion localization errors were 0.5 cm, 1.8 cm and 1.0 cm when torso models 1, 2 and 3 were used, respectively. From the results it can be concluded that not only the shape of the torso but also the correct position of electrodes and the relations between the heart and electrode positions have to be modeled very truly to minimize the lesion localization error.

Publications:

Result 4: Meranie koncentrácie magnetických nanočastíc v biologických tkanivách

Projects: VEGA 2/0160/10

Authors: M. Škrátek, I. Šimáček, A. Dvurečenskij, A. Cigáň

Bola navrhnutá a odskúšaná magnetometrická metóda kvantifikácie obsahu magnetických nanočastíc Fe₃O₄ injektovaných prostredníctvom magnetických kvapalín do tkaniva  pokusných zvierat (potkanov). Koncentrácia nanočastíc v meranom objekte a vo vzorkách sa určovala na základe magnetickej odozvy na striedavé magnetické pole jednokanálovým SQUID gradiometrom. Výsledky sa porovnávali s meraním magnetického momentu mikrovzoriek roztokov magnetických kvapalín a  vody získaných pomocou SQUID magnetometra Quantum Design MPMS XL 7 AC. Kombináciou týchto dvoch meraní  bolo možné spresniť výsledky a analyzovať chyby kvantifikačných meraní spôsobené rôznou veľkosťou meraného objektu a jeho vzdialenosťou od snímača. Predpokladaná citlivosť merania koncentrácii nanočastíc nachádzajúcich sa v biologických tkanivách a koloidných roztokoch bude pri tejto metóde 3 až 5 µg_Fe3O4 /cm³_H2O.

Publication:

ŠKRÁTEK, Martin - ŠIMÁČEK, Ivan - DVUREČENSKIJ, Andrej - CIGÁŇ, Alexander. Identification of magnetic nanoparticles by SQUID biosusceptometric system. In Lékař a technika : biomedicínské inženýrství a informatika, 2012, vol. 42, no. 2, p. 35-37. ISSN 0301-5491.