Selective sampling for assessing structural spatial frequencies with specific contrast mechanisms
The disclosed embodiments provide a method for acquiring MR data at resolutions down to tens of microns for application in in vivo diagnosis and monitoring of pathology for which changes in fine tissue textures can be used as markers of disease onset and progression. Bone diseases, tumors, neurologic diseases, and diseases involving fibrotic growth and/or destruction are all target pathologies. Further the technique can be used in any biologic or physical system for which very high-resolution characterization of fine scale morphology is needed. The method provides rapid acquisition of signal at selected values in k-space, with multiple successive acquisitions at individual k-values taken on a time scale on the order of microseconds, within a defined tissue volume, and subsequent combination of the multiple measurements in such a way as to maximize SNR. The reduced acquisition volume, and acquisition of only signal values at select places in k-space, along selected directions, enables much higher in vivo resolution than is obtainable with current MRI techniques.
1. A method for calibration of Magnetic Resonance (MR) tissue texture measurement comprising:
using microCT, MRI microscopy, or pathology to obtain high resolution 2D or 3D tissue data sets from selected tissue samples;
simulating data acquisition using the data sets as input for applying a selected contrast mechanism, selectively exciting a simulated volume of interest (VOI) employing a plurality of simulated time varying radio frequency signals and applied gradients, applying a simulated encoding gradient pulse to induce phase wrap to create a spatial encode for a specific k-value and orientation, the specific k-value determined based on the texture within the VOI, initiating a series of simulated gradients to produce k-value encodes, a resulting k-value set being a subset of that required to produce an image of the VOI, recording multiple sequential samples of simulated NMR RF signal encoded with the k-value set and post processing the recorded NMR signal samples to produce a data set of signal vs k-values for k-values in the k-value set, to characterize a simulation of the textural features of tissue in the VOI;
comparing features in the 2D/3D data sets with the simulation of textural features to provide comparative datasets;
repeating the step of simulating data acquisition across a high number of VOIs positioned within the tissue datasets;
applying supervised machine learning to the simulation of textual features and to the comparative datasets to optimize acquisition parameters including VOI dimensions and acquisition direction, using best resolution of the targeted feature measure as an endpoint.
2. The method as defined in claim 1 further comprising:
using unsupervised machine learning across the defined VOIs in tissue with specific disease markers to identify salient features additional to that called out for supervised learning;
using machine learning algorithms to correlate those features with information known regarding disease onset and progression in the tissue samples towards biomarker identification;
determining a sparsely sampled data set needed for measuring the tissue biomarkers towards a disease diagnostic assessment;
using machine learning algorithms to determine the strength of the diagnostic assessment;
acquiring data in the actual SNR environment of a MR scanner by applying the selected contrast mechanism, selectively exciting a volume of interest (VOI) employing a plurality of time varying radio frequency signals and applied gradients, applying an encoding gradient pulse to induce phase wrap to create a spatial encode for a specific k-value and orientation, the specific k-value determined based on texture within the VOI, initiating a series of gradients to produce k-value encodes, a resulting k-value set being a subset of that required to produce an image of the VOI, recording multiple sequential samples of the NMR RF signal encoded with the k-value set and post processing the recorded NMR signal samples to produce a data set of signal vs k-values for k-values in the k-value set, to characterize textural features of tissue in the VOI, on the same tissue samples, for comparison to the ground truth datasets;
repeating the recited steps to obtain optimization of acquisition parameters and calibration of the embodiments disclosed towards high resolution, robust textural measure.
3. A method for interpretation of Magnetic Resonance (MR) tissue texture measurement for determining pathology of a tissue type comprising:
selecting a contrast mechanism enhancing the contrast between component tissue types in a multiphase biologic sample for measurement with a MR imaging process;
applying the selected contrast mechanism;
selectively exciting a volume of interest (VOI) employing a plurality of time varying radio frequency signals and applied gradients;
applying an encoding gradient pulse to induce phase wrap to create a spatial encode for a specific k-value and orientation, the specific k-value determined based on texture within the VOI;
initiating a series of gradients to produce k-value encodes, a resulting k-value set being a subset of that required to produce an image of the VOI;
recording multiple sequential samples of the NMR RF signal encoded with the k-value set;
post processing the recorded NMR signal samples to produce a data set of signal vs k-values for k-values in the k-value set, to characterize textural features of tissue in the VOI; and
applying machine learning to a power density distribution of a textural wavelength of the k-value set to identify bio-markers for diagnosis of pathology of the tissue.
4. The method of claim 3 further comprising applying machine learning to identify a correlation between textural features and features in a power density spectrum of the textural wavelengths.
5. The method of claim 3 further comprising applying machine learning to the textural features and diagnostic information sources using additional sources of diagnostic information such as patient histories, exam records, imaging, serum markers, physical performance, and cognitive tests for extraction of diagnostic data to determine a disease assessment.
6. The method of claim 5 further comprising applying machine learning to determine weighting of the various diagnostic information sources in the ultimate diagnosis.
7. The method of claim 3 further comprising:
selecting a plurality of biologic phantoms having tissue pathology from healthy through diseased;
selecting a contrast mechanism enhancing the contrast between component tissue types in each biologic phantom for measurement with a MR imaging process;
applying the selected contrast mechanism;
selectively exciting a volume of interest (VOI) in each biologic phantom employing a plurality of time varying radio frequency signals and applied gradients;
applying an encoding gradient pulse to induce phase wrap to create a spatial encode for a specific k-value and orientation, the specific k-value determined based on texture within the VOI;
initiating a series of gradients to produce k-value encodes, a resulting k-value set being a subset of that required to produce an image of the VOI; and,
recording multiple sequential samples of the NMR RF signal encoded with the k-value set to provide texture measurement of each of the biologic phantoms.
post processing the recorded NMR signal samples to produce a data set of signal vs k-values for k-values in the k-value set, to characterize textural features of tissue in the VOI.
8. A method for pathology assessment employing tissue texture using magnetic resonance (MR) comprising:
selectively exciting a volume of interest (VOI) in tissue subject to motion employing a plurality of time varying radio frequency signals and applied gradients;
applying an encoding gradient pulse to induce phase wrap to create a spatial encode for a specific k-value and orientation, the specific k-value determined based on texture within the VOI;
initiating a series of gradients to produce k-value encodes giving a resulting k-value set;
recording multiple sequential samples of the NMR RF signal encoded with the k-value set within a single excitation of the VOI, the selectively excited VOI moving with the tissue;
post processing the recorded NMR signal samples to produce a data set of signal vs k-values for k-values in the k-value set, to characterize textural features of tissue in the VOI.