WP2 - Imaging and Quantification

Leaders: R. Faccini, S. Bortolussi


  • WP2 has the goal to assess the distribution and kinetics of tracers in cell cultures, mice models and humans and to this aim has three tasks:

Task 2.1)

Assess the receptivity of the tracers used by WP4 on DU145 and PANC-1 models by means of three techniques. The first two, Liquid Chromatography with High Resolution Mass Spectrometry and Neutron Autoradiography with CR39 detectors 11B at the TRIGAMark II reactor, are complementary, and test the presence of the specific molecule and of the Boron nuclei, respectively. The outcome of these tests is input of the radiobiological experiments and of the animal tests.

The third technique aims at understanding not only how much tracer is internalized by the cells when exposed to a given concentration, but also which concentration of tracers reaches the cells. To this aim we will perform tests on mice with PANC-1 tumors.

Task 2.2)

In order to assess the actual concentration on fluorine or boron reaches the tumor, a technique sensitive to relatively high concentrations and that can be performed in-vivo is required. To this aim, PET/SPECT imaging would allow to test concentrations up to

10-16 Moles/ml, while those required for sensitization are eleven orders of magnitude larger. We therefore suggest to exploit the fact that the gyromagnetic factor of 19F is close to 1H to perform Magnetic Resonance Imaging (MRI) of the distribution of 19F.

The absence of intrinsic signals in living tissues allows indeed in vivo visualization of fluorinated tracers, with a signal-to-noise ratio (SNR) close to that of 1H-MRI

even with only 1mMole/ml of 19F. Nonetheless, the limit of MRI with is the relatively high concentrations required to have a good SNR. Since the limiting factor is the scarce signal due to low concentration of 19F-compounds, the electronic noise of the readout chain, the low efficiency of the radiofrequency coils, this task explores the HW modifications that can increase the sensitivity to 19F-MRI.

In particolar:

● we are selecting 19F-compounds with more 19F identical nuclei and sufficient long T2-relaxation times to increase 19F signal;

● we are studying the performances of the RF antennas to design a more performant one;

● we are studying the used amplifiers to exploit the possibility to improve it either by changing components or by introducing cooling;

● we are implementing an SDR system in order to be able to test different digital signal reconstructions and optimize the system performances;

To test such HW improvements a stand is being setup in the NMR laboratory of Silvia Capuani (CNR), placed inside the Department of Physics. Such test stand is made of a 0,35T scanner and a mobile NMR scanner, both produced by BRUKER and accessible both in input (antenna) and in output (amplifier/SDR). Moreover, to test 19F-compounds a high field MRI scanner is used.

Task 2.3)

In addition to hardware improvements a better resolution of 19F-MRI can be achieved with software improvements: the problem to be solved is equivalent to the search of a small signal in a large background. To this aim we plan to apply machine learning in three different tasks:

● Noise reduction: one of the main hindrances to clinical application of 19F-MRI is the low signal-to-noise ratio. Recent developments in deep learning neural networks based denoisers (DNN) have shown promising results in noise reduction tasks. DNNs often outperform state-of-the-art conventional denoisers [4, 10, 6] in image restoration and they may be helpful in maximizing the SNR in 19F-MRI.

● Registration of 1H and 19F MRI: Image registration is the process of matching and superimposing two or more images taken at different times, with different equipments or different protocols. It is of critical importance in the integration of multiple sources of information in medical image analysis.

● Automated segmentation of anatomical structure in 1H MRI: we plan to identify the set of voxels which make up the volume of the object(s) of interest.

Papers:

  • R. Ferrari et al., “MR-based artificial intelligence model to assess response to therapy in locally advanced rectal cancer”EUROPEAN JOURNAL OF RADIOLOGY, 118-, (2019)