A machine learning approach that measures pH using acidoCEST MRI of iopamidol.

Tumor acidosis is an important biomarker for aggressive tumors, and extracellular pH (pHe) of the tumor microenvironment can be used to predict and evaluate tumor responses to chemotherapy and immunotherapy. AcidoCEST MRI measures tumor pHe by exploiting the pH-dependent chemical exchange saturation transfer (CEST) effect of iopamidol, an exogenous CT agent repurposed for CEST MRI. However, all pH fitting methodologies for acidoCEST MRI data have limitations. Herein we present results of the application of machine learning for extracting pH values from CEST Z-spectra of iopamidol. We acquired 36,000 experimental CEST spectra from 200 phantoms of iopamidol prepared at five concentrations, five T1 values, and eight pH values at five temperatures, acquired at six saturation powers and six saturation times. We also acquired T1 , T2 , B1 RF power, and B0 magnetic field strength supplementary MR information. These MR images were used to train and validate machine learning models for the tasks of pH classification and pH regression. Specifically, we tested the L1-penalized logistic regression classification (LRC) model and the random forest classification (RFC) model for classifying the CEST Z-spectra for thresholds at pH 6.5 and 7.0. Our results showed that both RFC and LRC were effective for pH classification, although the RFC model achieved higher predictive value, and improved the accuracy of classification accuracy with CEST Z-spectra with a more limited set of saturation frequencies. Furthermore, we used LASSO and random forest regression (RFR) models to explore pH regression, which showed that the RFR model achieved higher accuracy and precision for estimating pH across the entire pH range of 6.2-7.3, especially when using a more limited set of features. Based on these results, machine learning for analysis of acidoCEST MRI is promising for eventual in vivo determination of tumor pHe.

Differentiation of Myositis-Induced Models of Bacterial Infection and Inflammation with T2-Weighted, CEST, and DCE-MRI.

We used T2 relaxation, chemical exchange saturation transfer (CEST), and dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) to assess whether bacterial infection can be differentiated from inflammation in a myositis-induced mouse model. We measured the T2 relaxation time constants, %CEST at 5 saturation frequencies, and area under the curve (AUC) from DCE-MRI after maltose injection from infected, inflamed, and normal muscle tissue models. We applied principal component analysis (PCA) to reduce dimensionality of entire CEST spectra and DCE signal evolutions, which were analyzed using standard classification methods. We extracted features from dimensional reduction as predictors for machine learning classifier algorithms. Normal, inflamed, and infected tissues were evaluated with H&E and gram-staining histological studies, and bacterial-burden studies. The T2 relaxation time constants and AUC of DCE-MRI after injection of maltose differentiated infected, inflamed, and normal tissues. %CEST amplitudes at -1.6 and -3.5 ppm differentiated infected tissues from other tissues, but these did not differentiate inflamed tissue from normal tissue. %CEST amplitudes at 3.5, 3.0, and 2.5 ppm, AUC of DCE-MRI for shorter time periods, and relative Ktrans and kep values from DCE-MRI could not differentiate tissues. PCA and machine learning of CEST-MRI and DCE-MRI did not improve tissue classifications relative to traditional analysis methods. Similarly, PCA and machine learning did not further improve tissue classifications relative to T2 MRI. Therefore, future MRI studies of infection models should focus on T2-weighted MRI and analysis of T2 relaxation times.

Machine learning improves classification of preclinical models of pancreatic cancer with chemical exchange saturation transfer MRI.

PURPOSE: We sought to assess whether machine learning-based classification approaches can improve the classification of pancreatic tumor models relative to more simplistic analysis methods, using T1 relaxation, CEST, and DCE MRI. METHODS: The T1 relaxation time constants, % CEST at five saturation frequencies, and vascular permeability constants from DCE MRI were measured from Hs 766 T, MIA PaCa-2, and SU.86.86 pancreatic tumor models. We used each of these measurements as predictors for machine learning classifier algorithms. We also used principal component analysis to reduce the dimensionality of entire CEST spectra and DCE signal evolutions, which were then analyzed using classification methods. RESULTS: The T1 relaxation time constants, % CEST amplitudes at specific saturation frequencies, and the relative Ktrans and kep values from DCE MRI could not classify all three tumor types. However, the area under the curve from DCE signal evolutions could classify each tumor type. Principal component analysis was used to analyze the entire CEST spectrum and DCE signal evolutions, which predicted the correct tumor model with 87.5% and 85.1% accuracy, respectively. CONCLUSIONS: Machine learning applied to the entire CEST spectrum improved the classification of the three tumor models, relative to classifications that used % CEST values at single saturation frequencies. A similar improvement was not attained with machine learning applied to T1 relaxation times or DCE signal evolutions, relative to more simplistic analysis methods.

A light-fluence-independent method for the quantitative analysis of dynamic contrast-enhanced multispectral optoacoustic tomography (DCE MSOT).

MultiSpectral Optoacoustic Tomography (MSOT) is an emerging imaging technology that allows for data acquisition at high spatial and temporal resolution. These imaging characteristics are advantageous for Dynamic Contrast Enhanced (DCE) imaging that can assess the combination of vascular flow and permeability. However, the quantitative analysis of DCE MSOT data has not been possible due to complications caused by wavelength-dependent light attenuation and variability in light fluence at different anatomical locations. In this work we present a new method for the quantitative analysis of DCE MSOT data that is not biased by light fluence. We have named this method the two-compartment linear standard model (2C-LSM) for DCE MSOT.

Preliminary Results that Assess Metformin Treatment in a Preclinical Model of Pancreatic Cancer Using Simultaneous [18F]FDG PET and acidoCEST MRI.

PURPOSE: We sought to determine if the synergy between evaluations of glucose uptake in tumors and extracellular tumor acidosis measured with simultaneous positron emission tomography (PET)/magnetic resonance imaging (MRI) can improve longitudinal evaluations of the response to metformin treatment. PROCEDURES: A standard 2-deoxy-2-[18F]fluoro-D-glucose ([18F]FDG) PET protocol that evaluates glucose uptake in tumors, and a standard acidoCEST MRI protocol that measures extracellular pH (pHe) in tumors, were simultaneously performed to assess eight vehicle-treated (control) mice and eight metformin-treated mice 1 day before treatment, 1 day after initiating daily treatment with metformin, and 7 days after initiating treatment. Longitudinal changes in SUVmax and extracellular pH (pHe) were evaluated for each treatment group, and differences in SUVmax and pHe between metformin-treated and control groups were also evaluated. RESULTS: MRI acquisition protocols had little effect on the PET count rate, and the PET instrumentation had little effect on image contrast during acidoCEST MRI, verifying that [18F]FDG PET and acidoCEST MRI can be performed simultaneously. The average SUVmax of the tumor model had a significant decrease after 7 days of treatment with metformin, as expected. The average tumor pHe decreased after 7 days of metformin treatment, which reflected the inhibition of the consumption of cytosolic lactic acid caused by metformin. However, the average SUVmax of the tumor model was not significantly different between the metformin-treated and control groups after 7 days of treatment, and average pHe was also not significantly different between these groups. For comparison, the combination of average SUVmax and pHe measurements significantly differed between the treatment group and control group on Day 7. CONCLUSIONS: [18F]FDG PET and acidoCEST MRI studies can be performed simultaneously. The synergistic combination of assessing glucose uptake and tumor acidosis can improve differentiation of a drug-treated group from a control group during drug treatment of a tumor model.

Linearization improves the repeatability of quantitative dynamic contrast-enhanced MRI.

PURPOSE: The purpose of this study was to compare the repeatabilities of the linear and nonlinear Tofts and reference region models (RRM) for dynamic contrast-enhanced MRI (DCE-MRI). MATERIALS AND METHODS: Simulated and experimental DCE-MRI data from 12 rats with a flank tumor of C6 glioma acquired over three consecutive days were analyzed using four quantitative and semi-quantitative DCE-MRI metrics. The quantitative methods used were: 1) linear Tofts model (LTM), 2) non-linear Tofts model (NTM), 3) linear RRM (LRRM), and 4) non-linear RRM (NRRM). The following semi-quantitative metrics were used: 1) maximum enhancement ratio (MER), 2) time to peak (TTP), 3) initial area under the curve (iauc64), and 4) slope. LTM and NTM were used to estimate Ktrans, while LRRM and NRRM were used to estimate Ktrans relative to muscle (RKtrans). Repeatability was assessed by calculating the within-subject coefficient of variation (wSCV) and the percent intra-subject variation (iSV) determined with the Gage R&R analysis. RESULTS: The iSV for RKtrans using LRRM was two-fold lower compared to NRRM at all simulated and experimental conditions. A similar trend was observed for the Tofts model, where LTM was at least 50% more repeatable than the NTM under all experimental and simulated conditions. The semi-quantitative metrics iauc64 and MER were as equally repeatable as Ktrans and RKtrans estimated by LTM and LRRM respectively. The iSV for iauc64 and MER were significantly lower than the iSV for slope and TTP. CONCLUSION: In simulations and experimental results, linearization improves the repeatability of quantitative DCE-MRI by at least 30%, making it as repeatable as semi-quantitative metrics.

Multislice CEST MRI improves the spatial assessment of tumor pH.

PURPOSE: Multislice maps of extracellular pH (pHe) are needed to interrogate the heterogeneities of tumors and normal organs. To address this need, we have developed a multislice chemical exchange saturation transfer (CEST) MRI acquisition method with a CEST spectrum-fitting method that measures in vivo pHe over a range of 6.3 to 7.4. METHODS: The phase offset multiplanar (POMP) method was adapted for CEST fast imaging with steady-state free precession (FISP) MRI to acquire multiple image slices with a single CEST saturation pulse. The Bloch-McConnell equations were modified to include pH based on a calibration of pH and chemical exchange rate for the contrast agent iopamidol. These equations were used to estimate the pixel-wise pHe values throughout the multislice acidoCEST MR images of the tumor, kidney, bladder, and other tissues of a MDA-MB-231 tumor model. RESULTS: Multislice acidoCEST MRI successfully mapped a gradient of pHe from 6.73 to 6.81 units from the tumor core to rim, and also mapped a gradient of pHe 6.56 to 6.97 across the mouse kidney. The bladder was found to be pHe 6.3. CONCLUSION: AcidoCEST MRI with POMP acquisition and Bloch-McConnel analysis can map pHe in multiple imaging slices through the tumor, kidney, and bladder. This multislice evaluation facilitates assessments of spatial heterogeneity of tissue pHe. Magn Reson Med 78:97-106, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

QUESPOWR MRI: QUantification of Exchange as a function of Saturation Power On the Water Resonance.

QUantification of Exchange as a function of Saturation Power On the Water Resonance (QUESPOWR) MRI is a new method that can estimate chemical exchange rates. This method acquires a series of OPARACHEE MRI acquisitions with a range of RF powers for the WALTZ16(∗) pulse train, which are applied on the water resonance. A QUESPOWR plot can be generated from the power dependence of the % water signal, which is similar to a QUESP plot that is generated from CEST MRI acquisition methods with RF saturation applied off-resonance from water. A QUESPOWR plot can be quantitatively analyzed using linear fitting methods to provide estimates of average chemical exchange rates. Analyses of the shapes of QUESPOWR plots can also be used to estimate relative differences in average chemical exchange rates and concentrations of biomolecules. The performance of QUESPOWR MRI was assessed via simulations, an in vitro study with iopamidol, and an in vivo study with a mouse model of mammary carcinoma. The results showed that QUESPOWR MRI is especially sensitive to chemical exchange between water and biomolecules that have intermediate to fast chemical exchange rates and chemical shifts that are close to water, which are notoriously difficult to assess with other CEST MRI methods. In addition, in vivo QUESPOWR MRI detected acidic tumor tissues relative to normal tissues that are pH-neutral, and therefore may be a new paradigm for tumor detection with MRI.

Design and optimization of pulsed Chemical Exchange Saturation Transfer MRI using a multiobjective genetic algorithm.

Pulsed Chemical Exchange Saturation Transfer (CEST) MRI experimental parameters and RF saturation pulse shapes were optimized using a multiobjective genetic algorithm. The optimization was carried out for RF saturation duty cycles of 50% and 90%, and results were compared to continuous wave saturation and Gaussian waveform. In both simulation and phantom experiments, continuous wave saturation performed the best, followed by parameters and shapes optimized by the genetic algorithm and then followed by Gaussian waveform. We have successfully demonstrated that the genetic algorithm is able to optimize pulse CEST parameters and that the results are translatable to clinical scanners.

A single diamagnetic catalyCEST MRI contrast agent that detects cathepsin B enzyme activity by using a ratio of two CEST signals.

CatalyCEST MRI can detect enzyme activity by monitoring the change in chemical exchange with water after a contrast agent is cleaved by an enzyme. Often these molecules use paramagnetic metals and are delivered with an additional non-responsive reference molecule. To improve this approach for molecular imaging, a single diamagnetic agent with enzyme-responsive and enzyme-unresponsive CEST signals was synthesized and characterized. The CEST signal from the aryl amide disappeared after cleavage of a dipeptidyl ligand with cathepsin B, while a salicylic acid moiety was largely unresponsive to enzyme activity. The ratiometric comparison of the two CEST signals from the same agent allowed for concentration independent measurements of enzyme activity. The chemical exchange rate of the salicylic acid moiety was unchanged after enzyme catalysis, which further validated that this moiety was enzyme-unresponsive. The temperature dependence of the chemical exchange rate of the salicylic acid moiety was non-Arrhenius, suggesting a two-step chemical exchange mechanism for salicylic acid. The good detection sensitivity at low saturation power facilitates clinical translation, along with the potentially low toxicity of a non-metallic MRI contrast agent. The modular design of the agent constitutes a platform technology that expands the variety of agents that may be employed by catalyCEST MRI for molecular imaging.

Measuring extracellular pH in a lung fibrosis model with acidoCEST MRI.

PURPOSE: A feed-forward loop involving lactic acid production may potentially occur during the formation of idiopathic pulmonary fibrosis. To provide evidence for this feed-forward loop, we used acidoCEST MRI to measure the extracellular pH (pHe), while also measuring percent uptake of the contrast agent, lesion size, and the apparent diffusion coefficient (ADC). PROCEDURES: We developed a respiration-gated version of acidoCEST MRI to improve the measurement of pHe and percent uptake in lesions. We also used T2-weighted MRI to measure lesion volumes and diffusion-weighted MRI to measure ADC. RESULTS: The longitudinal changes in average pHe and % uptake of the contrast agent were inversely related to reduction in lung lesion volume. The average ADC did not change during the time frame of the study. CONCLUSIONS: The increase in pHe during the reduction in lesion volume indicates a role for lactic acid in the proposed feed-forward loop of IPF.

Comparison of analytical and numerical analysis of the reference region model for DCE-MRI.

This study compared three methods for analyzing DCE-MRI data with a reference region (RR) model: a linear least-square fitting with numerical analysis (LLSQ-N), a nonlinear least-square fitting with numerical analysis (NLSQ-N), and an analytical analysis (NLSQ-A). The accuracy and precision of estimating the pharmacokinetic parameter ratios KR and VR, where KR is defined as a ratio between the two volume transfer constants, K(trans,TOI) and K(trans,RR), and VR is the ratio between the two extracellular extravascular volumes, ve,TOI and ve,RR, were assessed using simulations under various signal-to-noise ratios (SNRs) and temporal resolutions (4, 6, 30, and 60s). When no noise was added, the simulations showed that the mean percent error (MPE) for the estimated KR and VR using the LLSQ-N and NLSQ-N methods ranged from 1.2% to 31.6% with various temporal resolutions while the NLSQ-A method maintained a very high accuracy (<1.0×10(-4) %) regardless of the temporal resolution. The simulation also indicated that the LLSQ-N and NLSQ-N methods appear to underestimate the parameter ratios more than the NLSQ-A method. In addition, seven in vivo DCE-MRI datasets from spontaneously occurring canine brain tumors were analyzed with each method. Results for the in vivo study showed that KR (ranging from 0.63 to 3.11) and VR (ranging from 2.82 to 19.16) for the NLSQ-A method were both higher than results for the other two methods (KR ranging from 0.01 to 1.29 and VR ranging from 1.48 to 19.59). A temporal downsampling experiment showed that the averaged percent error for the NLSQ-A method (8.45%) was lower than the other two methods (22.97% for LLSQ-N and 65.02% for NLSQ-N) for KR, and the averaged percent error for the NLSQ-A method (6.33%) was lower than the other two methods (6.57% for LLSQ-N and 13.66% for NLSQ-N) for VR. Using simulations, we showed that the NLSQ-A method can estimate the ratios of pharmacokinetic parameters more accurately and precisely than the NLSQ-N and LLSQ-N methods over various SNRs and temporal resolutions. All simulations were validated with in vivo DCE MRI data.

A reference agent model for DCE MRI can be used to quantify the relative vascular permeability of two MRI contrast agents.

Dynamic Contrast Enhancement (DCE) MRI has been used to measure the kinetic transport constant, K(trans), which is used to assess tumor angiogenesis and the effects of anti-angiogenic therapies. Standard DCE MRI methods must measure the pharmacokinetics of a contrast agent in the blood stream, known as the Arterial Input Function (AIF), which is then used as a reference for the pharmacokinetics of the agent in tumor tissue. However, the AIF is difficult to measure in pre-clinical tumor models and in patients. Moreover the AIF is dependent on the Fahraeus effect that causes a highly variable hematocrit (Hct) in tumor microvasculature, leading to erroneous estimates of K(trans). To overcome these problems, we have developed the Reference Agent Model (RAM) for DCE MRI analyses, which determines the relative K(trans) of two contrast agents that are simultaneously co-injected and detected in the same tissue during a single DCE-MRI session. The RAM obviates the need to monitor the AIF because one contrast agent effectively serves as an internal reference in the tumor tissue for the other agent, and it also eliminates the systematic errors in the estimated K(trans) caused by assuming an erroneous Hct. Simulations demonstrated that the RAM can accurately and precisely estimate the relative K(trans) (R(Ktrans)) of two agents. To experimentally evaluate the utility of RAM for analyzing DCE MRI results, we optimized a previously reported multiecho (19)F MRI method to detect two perfluorinated contrast agents that were co-injected during a single in vivo study and selectively detected in the same tumor location. The results demonstrated that RAM determined R(Ktrans) with excellent accuracy and precision.

A linear algorithm of the reference region model for DCE-MRI is robust and relaxes requirements for temporal resolution.

Dynamic contrast enhanced MRI (DCE-MRI) has utility for improving clinical diagnoses of solid tumors, and for evaluating the early responses of anti-angiogenic chemotherapies. The Reference Region Model (RRM) can improve the clinical implementation of DCE-MRI by substituting the contrast enhancement of muscle for the Arterial Input Function that is used in traditional DCE-MRI methodologies. The RRM is typically fitted to experimental results with a non-linear least squares algorithm. This report demonstrates that this algorithm produces inaccurate and imprecise results when DCE-MRI results have low SNR or slow temporal resolution. To overcome this limitation, a linear least-squares algorithm has been derived for the Reference Region Model. This new algorithm improves accuracy and precision of fitting the Reference Region Model to DCE-MRI results, especially for voxel-wise analyses. This linear algorithm is insensitive to injection speeds, and has 300- to 8000-fold faster calculation speed relative to the non-linear algorithm. The linear algorithm produces more accurate results for over a wider range of permeabilities and blood volumes of tumor vasculature. This new algorithm, termed the Linear Reference Region Model, has strong potential to improve clinical DCE-MRI evaluations.

Imaging biomarkers to monitor response to the hypoxia-activated prodrug TH-302 in the MiaPaCa2 flank xenograft model.

TH-302, a hypoxia-activated anticancer prodrug, was evaluated for antitumor activity and changes in dynamic contrast-enhanced (DCE) and diffusion-weighted (DW) magnetic resonance imaging (MRI) in a mouse model of pancreatic cancer. TH-302 monotherapy resulted in a significant delay in tumor growth compared to vehicle-treated controls. TH-302 treatment was also associated with a significant decrease in the volume transfer constant (K(trans)) compared to vehicle-treated controls 1 day following the first dose measured using DCE-MRI. This early decrease in K(trans) following the first dose as measured is consistent with selective killing of the hypoxic fraction of cells which are associated with enhanced expression of hypoxia inducible transcription factor-1 alpha that regulates expression of permeability and perfusion factors including vascular endothelial growth factor-A. No changes were observed in DW-MRI following treatment with TH-302, which may indicate that this technique is not sensitive enough to detect changes in small hypoxic fractions of the tumor targeted by TH-302. These results suggest that changes in tumor permeability and/or perfusion may be an early imaging biomarker for response to TH-302 therapy.