Development of advanced signal processing and source imaging methods for superparamagnetic relaxometry.

Superparamagnetic relaxometry (SPMR) is a highly sensitive technique for the in vivo detection of tumor cells and may improve early stage detection of cancers. SPMR employs superparamagnetic iron oxide nanoparticles (SPION). After a brief magnetizing pulse is used to align the SPION, SPMR measures the time decay of SPION using super-conducting quantum interference device (SQUID) sensors. Substantial research has been carried out in developing the SQUID hardware and in improving the properties of the SPION. However, little research has been done in the pre-processing of sensor signals and post-processing source modeling in SPMR. In the present study, we illustrate new pre-processing tools that were developed to: (1) remove trials contaminated with artifacts, (2) evaluate and ensure that a single decay process associated with bounded SPION exists in the data, (3) automatically detect and correct flux jumps, and (4) accurately fit the sensor signals with different decay models. Furthermore, we developed an automated approach based on multi-start dipole imaging technique to obtain the locations and magnitudes of multiple magnetic sources, without initial guesses from the users. A regularization process was implemented to solve the ambiguity issue related to the SPMR source variables. A procedure based on reduced chi-square cost-function was introduced to objectively obtain the adequate number of dipoles that describe the data. The new pre-processing tools and multi-start source imaging approach have been successfully evaluated using phantom data. In conclusion, these tools and multi-start source modeling approach substantially enhance the accuracy and sensitivity in detecting and localizing sources from the SPMR signals. Furthermore, multi-start approach with regularization provided robust and accurate solutions for a poor SNR condition similar to the SPMR detection sensitivity in the order of 1000 cells. We believe such algorithms will help establishing the industrial standards for SPMR when applying the technique in pre-clinical and clinical settings.

A reconstructed metastasis model to recapitulate the metastatic spread in vitro.

Metastasis remains a leading cause of morbidity and mortality from solid tumors. Lack of comprehensive systems to study the progression of metastasis contributes to the low success of treatment. We developed a novel three-dimensional in vitro reconstructed metastasis (rMet) model that incorporates extracellular matrix (ECM) elements characteristic of the primary (breast, prostate, or lung) and metastatic (bone marrow, BM) sites. A cytokine-rich liquid interphase separates the primary and distant sites, further recapitulating circulation. Similar to main events underlying the metastatic cascade, the rMet model fractionated human tumor cell lines into sub-populations with distinct invasive and migratory abilities: (i) a primary tumor-like fraction mainly consisting of non-migratory spheroids; (ii) an invasive fraction that invaded through the primary tumor ECM, but failed to acquire anchorage-independence and reach the BM; and (iii) a highly migratory BM-colonizing population that invaded the primary ECM, survived in the "circulation-like" media, and successfully invaded and proliferated within BM ECM. BM-colonizing fractions successfully established metastatic bone lesions in vivo, whereas the tumor-like spheroids failed to engraft the bones, showing the ability of the rMet model to faithfully select for highly aggressive sub-populations with a propensity to colonize a metastatic site. By applying the rMet model to study real-time ECM remodeling, we show that tumor cells secrete collagenolytic enzymes for invading the primary site ECM but not for entering the BM ECM, indicating possible differences in ECM remodeling mechanisms at primary tumor versus metastatic sites.