Targeted Nanoparticles with High Heating Efficiency for the Treatment of Endometriosis with Systemically Delivered Magnetic Hyperthermia.

Endometriosis is a devastating disease in which endometrial-like tissue forms lesions outside the uterus. It causes infertility and severe pelvic pain in ≈176 million women worldwide, and there is currently no cure for this disease. Magnetic hyperthermia could potentially eliminate widespread endometriotic lesions but has not previously been considered for treatment because conventional magnetic nanoparticles have relatively low heating efficiency and can only provide ablation temperatures (>46 °C) following direct intralesional injection. This study is the first to describe nanoparticles that enable systemically delivered magnetic hyperthermia for endometriosis treatment. When subjected to an alternating magnetic field (AMF), these hexagonal iron-oxide nanoparticles exhibit extraordinary heating efficiency that is 6.4× greater than their spherical counterparts. Modifying nanoparticles with a peptide targeted to vascular endothelial growth factor receptor 2 (VEGFR-2) enhances their endometriosis specificity. Studies in mice bearing transplants of macaque endometriotic tissue reveal that, following intravenous injection at a low dose (3 mg per kg), these nanoparticles efficiently accumulate in endometriotic lesions, selectively elevate intralesional temperature above 50 °C upon exposure to external AMF, and completely eradicate them with a single treatment. These nanoparticles also demonstrate promising potential as magnetic resonance imaging (MRI) contrast agents for precise detection of endometriotic tissue before AMF application.

3D MR fingerprinting using Seiffert spirals.

PURPOSE: Seiffert spirals were recently explored as an efficient way to traverse 3D k-space compared to traditional 3D techniques. Several studies have shown the ability of 3D MR fingerprinting (MRF) techniques to acquire T1 and T2 relaxation maps in a short period of time. However, these sequences do not sample across a large region of 3D k-space every TR, especially in the way that Seiffert trajectories can. METHODS: A 3D MRF sequence was designed using 8 Seiffert spirals rotated in 3D k-space, with flip angle modulation for T1 and T2 sensitivity. The sequence was compared to an MRF sequence using a 2D spiral rotated in 3D k-space using the tiny golden angle acquisition with similar resolution/readout duration. Both sequences were evaluated using simulations, phantom validation, and in vivo imaging. RESULTS: In all experiments, the Seiffert spiral MRF sequence performed similar to if not better than the multi-axis 2D spiral MRF sequence. Strong intraclass correlation coefficients (> 0.9) were found between conventional and MRF sequences in phantoms, whereas the in vivo results showed slightly less aliasing artifact with the Seiffert trajectory. CONCLUSION: In this study, Seiffert spirals were used within the MRF framework to acquire high-resolution T1 and T2 relaxation time maps in less than 2.5 min. The reduced aliasing artifacts seen with the Seiffert sequence suggests that sampling over 3D k-space evenly each TR can improve quantification or shorten scan times.

T1ρ magnetic resonance fingerprinting.

T1ρ relaxation imaging is a quantitative imaging technique that has been used to assess cartilage integrity, liver fibrosis, tumors, cardiac infarction, and Alzheimer's disease. T1 , T2 , and T1ρ relaxation time constants have each demonstrated different degrees of sensitivity to several markers of fibrosis and inflammation, allowing for a potential multi-parametric approach to tissue quantification. Traditional magnetic resonance fingerprinting (MRF) has been shown to provide quick, quantitative mapping of T1 and T2 relaxation time constants. In this study, T1ρ relaxation is added to the MRF framework using spin lock preparations. An MRF sequence involving an RF-spoiled sequence with TR , flip angle, T1ρ , and T2 preparation variation is described. The sequence is then calibrated against conventional T1 , T2 , and T1ρ relaxation mapping techniques in agar phantoms and the abdomens of four healthy volunteers. Strong intraclass correlation coefficients (ICC > 0.9) were found between conventional and MRF sequences in phantoms and also in healthy volunteers (ICC > 0.8). The highest ICC correlation values were seen in T1 , followed by T1ρ and then T2 . In this study, T1ρ relaxation has been incorporated into the MRF framework by using spin lock preparations, while still fitting for T1 and T2 relaxation time constants. The acquisition of these parameters within a single breath hold in the abdomen alleviates the issues of movement between breath holds in conventional techniques.

Multi-parametric T2 * magnetic resonance fingerprinting using variable echo times.

The use of quantitative imaging biomarkers in the imaging of various disease states, including cancer and neurodegenerative disease, has increased in recent years. T1 , T2 , and T2 * relaxation time constants have been shown to be affected by tissue structure or contrast infusion. Acquiring these biomarkers simultaneously in a multi-parametric acquisition could provide more robust detection of tissue changes in various disease states including neurodegeneration and cancer. Traditional magnetic resonance fingerprinting (MRF) has been shown to provide quick, quantitative mapping of T1 and T2 relaxation time constants. In this study, T2 * relaxation is added to the MRF framework using variable echo times (TE). To demonstrate the feasibility of the method and compare incremental and golden angle spiral rotations, simulated phantom data was fit using the proposed method. Additionally, T1 /T2 /T2 */δf MRF as well as conventional T1 , T2 , and T2 * acquisitions were acquired in agar phantoms and the brains of three healthy volunteers. Golden angle spiral rotation was found to reduce inaccuracy resulting from off resonance effects. Strong correlations were found between conventional and MRF values in the T1 , T2 , and T2 * relaxation time constants of the agar phantoms and healthy volunteers. In this study, T2 * relaxation has been incorporated into the MRF framework by using variable echo times, while still fitting for T1 and T2 relaxation time constants. In addition to fitting these relaxation time constants, a novel method for fitting and correcting off resonance effects has been developed.

Scoring hip osteoarthritis with MRI (SHOMRI): A whole joint osteoarthritis evaluation system.

BACKGROUND: To develop a semi-quantitative MR-based hip osteoarthritis (OA) evaluation system (Scoring hip osteoarthritis with MRI, SHOMRI), and to test its reproducibility and face validity. METHODS: The study involved 98 subjects with informed consent. Three-Tesla MR imaging of hip was performed in three planes with intermediate-weighted fat saturated FSE sequences. Two radiologists assessed cartilage loss, bone marrow edema pattern, subchondral cyst in 10 subregions, and assessed labrum in 4 subregions. In addition, presence or absence of ligamentum teres integrity, paralabral cysts, intra-articular body, and effusion in the hip joint were analyzed using the SHOMRI system. The reproducibility was assessed with intra-class correlation coefficient (ICC), Cohen's Kappa values and percent agreement. SHOMRI scores were correlated with radiographic Kellgren-Lawrence (KL) and OARSI atlas gradings, and clinical parameters, the hip osteoarthritis outcome score (HOOS) and hip range of motion (ROM), using Spearman's rank correlation and ordinal logistic regression. RESULTS: ICC values were in the excellent range, 0.91 to 0.97. Cohen's Kappa values and percent agreement ranged from 0.55 to 0.79 and 66 to 99%, respectively. SHOMRI demonstrated significant correlations with KL and OARSI gradings as well as with clinical parameters, HOOS and ROM (P < 0.05). Among the SHOMRI features, subchondral cyst and bone marrow edema pattern showed the highest correlation with HOOS and ROM. CONCLUSION: SHOMRI demonstrated moderate to excellent reproducibility and significant correlation with radiographic gradings and clinical parameters.

Validation of bone marrow fat quantification in the presence of trabecular bone using MRI.

BACKGROUND: To validate six-echo, chemical-shift based MRI with T2 * correction for the quantification of bone marrow fat content in the presence of trabecular bone. METHODS: Ten bone phantoms were made using trabecular bone cores extracted from the distal femur and proximal tibia of 20 human cadaveric knees. Bone marrow was removed from the cores and the marrow spaces were filled with water-fat gelatin to mimic bone marrow of known fat fractions. A chemical-shift based water-fat separation method with T2 * correction was used to generate fat fraction maps. The proton density fat fractions (PDFF) between marrow regions with and without bone were compared with the reference standard of known fat fraction using the squared Pearson correlation coefficient and unpaired t-test. RESULTS: Strong correlations were found between the known fat fraction and measured PDFF in marrow without trabecular bone (R(2) = 0.99; slope = 0.99, intercept = 0.94) as well as in marrow with trabecular bone (R(2) = 0.97; slope = 1.0, intercept = -3.58). Measured PDFF between regions with and without bone were not significantly different (P = 0.5). However, PDFF was systematically underestimated by -3.2% fat fraction in regions containing trabecular bone. CONCLUSION: Our implementation of a six-echo chemical-shift based MRI pulse sequence with T2 * correction provided an accurate means of determining fat content in bone marrow in the presence of trabecular bone.

Magnetic resonance analysis of loaded meniscus deformation: a novel technique comparing participants with and without radiographic knee osteoarthritis.

OBJECTIVE: To establish a novel method of quantifying meniscal deformation using loaded MRI. More specifically, the goals were to evaluate the (1) accuracy, (2) inter-rater reliability, (3) intra-rater reliability, and (4) scan-rescan reliability. The secondary purpose of this experiment was to evaluate group differences in meniscal deformation in participants with and without radiographic knee OA. MATERIALS AND METHODS: Weight-bearing 3-T MRIs of the knee in full extension and 30° of flexion were processed to create 3D models of meniscal deformation. Accuracy was assessed using a custom-designed phantom. Twenty-one participants either with or without signs of OA were evaluated, and another six participants (14 knees, one subject was scanned twice) underwent repeated imaging to assess scan-rescan reproducibility. Intraclass correlation coefficient (ICC), root-mean squared error (RMSE), and root-mean-square percent coefficient-of-variation (RMS%CV) analyses were performed. Exploratory comparisons were made between those with and without OA to evaluate potential group differences. RESULTS: All variables were found to be accurate with RMSE ranging from 0.08 to 0.35 mm and 5.99 to 14.63 mm(2). Reproducibility of peak anterior-posterior meniscal deformation was excellent (ICC > 0.821; p < 0.013) with RMS%CV for intra-rater ranging from 0.06 to 1.53 % and 0.17 to 1.97 %, inter-rater ranging from 0.10 to 7.20 % and 3.95 to 18.53 %, and scan-rescan reliability ranging from 1.531 to 7.890 % and 4.894 to 9.142 %, for distance and area metric, respectively. Participants with OA were found to have significantly greater anterior horn movement of both the medial (p = 0.039) and lateral meniscus (p = 0.015), and smaller flexed medial meniscus outer area (p = 0.048) when compared to controls. CONCLUSIONS: MRI-based variables of meniscus deformation were found to be valid in participants with and without OA. Significant differences were found between those with and without radiographic OA; further study is warranted.

Simultaneous acquisition of T1ρ and T2 quantification in knee cartilage: repeatability and diurnal variation.

PURPOSE: To develop a robust sequence that combines T1ρ and T2 quantifications and to examine the in vivo repeatability and diurnal variation of T1ρ and T2 quantifications in knee cartilage. MATERIALS AND METHODS: Six healthy volunteers were scanned in the morning and afternoon on 2 days using a combined T1ρ and T2 quantification sequence developed in this study. Repeatability of T1ρ and T2 quantification was estimated using root-mean-square coefficients-of-variation (RMS-CV). T1ρ and T2 values from morning scans were compared to those from afternoon scans using paired t-tests. RESULTS: The overall RMS-CV of in vivo T1ρ and T2 quantification was 5.3% and 5.2%, respectively. The RMS-CV of am scans was 4.2% and 5.0% while the RMS-CV of pm scans was 6.0% and 6.3% for T1ρ and T2 , respectively. No significant difference was found between T1ρ or T2 values in the morning and in the afternoon. CONCLUSION: A sequence that combines T1ρ and T2 quantification with scan time less than 10 minutes and is robust to B0 and B1 inhomogeneity was developed with excellent repeatability. For a cohort with low-level daily activity, although no significant diurnal variation of cartilage MR relaxation times was observed, the afternoon scans had inferior repeatability compared to morning scans.

Comparison of dynamic susceptibility contrast (DSC) using gadolinium and iron-based contrast agents in high-grade glioma at high-field MRI.

PURPOSE: To compare DSC-MRI using Gadolinium (GBCA) and Ferumoxytol (FBCA) in high-grade glioma at 3T and 7T MRI field strengths. We hypothesized that using FBCA at 7T would enhance the performance of DSC, as measured by contrast-to-noise ratio (CNR). METHODS: Ten patients (13 lesions) were assigned to 3T (6 patients, 6 lesions) or 7T (4 patients, 7 lesions). All lesions received 0.1 mmol/kg of GBCA on day 1. Ten lesions (4 at 3T and 6 at 7T) received a lower dose (0.6 mg/kg) of FBCA, followed by a higher dose (1.0-1.2 mg/kg), while 3 lesions (2 at 3T and 1 at 7T) received only a higher dose on Day 2. CBV maps with leakage correction for GBCA but not for FBCA were generated. The CNR and normalized CBV (nCBV) were analyzed on enhancing and non-enhancing high T2W lesions. RESULTS: Regardless of FBCA dose, GBCA showed higher CNR than FBCA at 7T, which was significant for high-dose FBCA (p < .05). Comparable CNR between GBCA and high-dose FBCA was observed at 3T. There was a trend toward higher CNR for FBCA at 3T than 7T. GBCA also showed nCBV twice that of FBCA at both MRI field strengths with significance at 7T. CONCLUSION: GBCA demonstrated higher image conspicuity, as measured by CNR, than FBCA on 7T. The stronger T2* weighting realized with higher magnetic field strength, combined with FBCA, likely results in more signal loss rather than enhanced performance on DSC. However, at clinical 3T, both GBCA and FBCA, particularly a dosage of 1.0-1.2 mg/kg (optimal for perfusion imaging), yielded comparable CNR.

Deletion of the mRNA stability factor ELAVL1 (HuR) in pancreatic cancer cells disrupts the tumor microenvironment integrity.

Stromal cells promote extensive fibrosis in pancreatic ductal adenocarcinoma (PDAC), which is associated with poor prognosis and therapeutic resistance. We report here for the first time that loss of the RNA-binding protein human antigen R (HuR, ELAVL1) in PDAC cells leads to reprogramming of the tumor microenvironment. In multiple in vivo models, CRISPR deletion of ELAVL1 in PDAC cells resulted in a decrease of collagen deposition, accompanied by a decrease of stromal markers (i.e. podoplanin, α-smooth muscle actin, desmin). RNA-sequencing data showed that HuR plays a role in cell-cell communication. Accordingly, cytokine arrays identified that HuR regulates the secretion of signaling molecules involved in stromal activation and extracellular matrix organization [i.e. platelet-derived growth factor AA (PDGFAA) and pentraxin 3]. Ribonucleoprotein immunoprecipitation analysis and transcription inhibition studies validated PDGFA mRNA as a novel HuR target. These data suggest that tumor-intrinsic HuR supports extrinsic activation of the stroma to produce collagen and desmoplasia through regulating signaling molecules (e.g. PDGFAA). HuR-deficient PDAC in vivo tumors with an altered tumor microenvironment are more sensitive to the standard of care gemcitabine, as compared to HuR-proficient tumors. Taken together, we identified a novel role of tumor-intrinsic HuR in its ability to modify the surrounding tumor microenvironment and regulate PDGFAA.

Comprehensive analysis of the Cramer-Rao bounds for magnetic resonance temperature change measurement in fat-water voxels using multi-echo imaging.

OBJECT: The aim of this paper is to characterize the noise propagation for MRI temperature change measurement with emphasis on finding the best echo time combinations that yield the lowest temperature noise. MATERIALS AND METHODS: A Cramer-Rao lower-bound (CRLB) calculation was used to estimate the temperature noise for a model of the MR signal in fat-water voxels. The temperature noise CRLB was then used to find a set of echo times that gave the lowest temperature change noise for a range of fat-water frequency differences, temperature changes, fat/water signal ratios, and T2* values. CRLB estimates were verified by Monte Carlo simulation and in phantoms using images acquired in a 1.5 T magnet. RESULTS: Results show that regions exist where the CRLB predicts minimal temperature variation as a function of the other variables. The results also indicate that the CRLB values calculated in this paper provide excellent guidance for predicting the variation of temperature measurements due to changes in the signal parameters. For three echo scans, the best noise characteristics are seen for TE values of 20.71, 23.71, and 26.71 ms. Results for five and seven echo scans are also presented in the text. CONCLUSION: The results present a comprehensive analysis of the effects of different scan parameters on temperature noise, potentially benefiting the selection of scan parameters for clinical MRI thermometry.

Pharmacological reduction of coagulation factor XI reduces macrophage accumulation and accelerates deep vein thrombosis resolution in a mouse model of venous thrombosis.

BACKGROUND: Deep vein thrombosis (DVT) and post-thrombotic syndrome (PTS) remain highly prevalent despite modern medical therapy. Contact activation is a promising target for safe antithrombotic anticoagulation. The anti-factor XI (FXI) monoclonal antibody 14E11 reduces circulating levels of FXI without compromising hemostasis. The human recombinant analog, AB023, is in clinical development. The role of FXI in mediation of inflammation during DVT resolution is unknown. OBJECTIVES: Investigate the effects of pharmacological targeting of FXI with 14E11 in an experimental model of venous thrombosis. METHODS: Adult wild-type CD1 mice were treated with subcutaneous anti-FXI antibody (14E11, 5 mg/kg) versus saline prior to undergoing surgical constriction of the inferior vena cava (IVC). Mice were evaluated at various time points to assess thrombus weight and volume, as well as histology analysis, ferumoxytol enhanced magnetic resonance imaging (Fe-MRI), and whole blood flow cytometry. RESULTS: 14E11-treated mice had reduced thrombus weights and volumes after IVC constriction on day 7 compared to saline-treated mice. 14E11 treatment reduced circulating monocytes by flow cytometry and macrophage content within thrombi as evaluated by histologic staining and Fe-MRI. Collagen deposition was increased at day 3 while CD31 and smooth muscle cell actin expression was increased at day 7 in the thrombi of 14E11-treated mice compared to saline-treated mice. CONCLUSION: Pharmacologic targeting of FXI enhances the early stages of experimental venous thrombus resolution in wild-type CD1 mice, and may be of interest for future clinical evaluation of the antibody in DVT and PTS.

An Advanced Thermal Decomposition Method to Produce Magnetic Nanoparticles with Ultrahigh Heating Efficiency for Systemic Magnetic Hyperthermia.

Due to the limited heating efficiency of available magnetic nanoparticles, it is difficult to achieve therapeutic temperatures above 44 °C in relatively inaccessible tumors during magnetic hyperthermia following systemic administration of nanoparticles at clinical dosage (≤10 mg kg-1 ). To address this, a method for the preparation of magnetic nanoparticles with ultrahigh heating capacity in the presence of an alternating magnetic field (AMF) is presented. The low nitrogen flow rate of 10 mL min-1 during the thermal decomposition reaction results in cobalt-doped nanoparticles with a magnetite (Fe3 O4 ) core and a maghemite (γ-Fe2 O3 ) shell that exhibit the highest intrinsic loss power reported to date of 47.5 nH m2 kg-1 . The heating efficiency of these nanoparticles correlates positively with increasing shell thickness, which can be controlled by the flow rate of nitrogen. Intravenous injection of nanoparticles at a low dose of 4 mg kg-1 elevates intratumoral temperatures to 50 °C in mice-bearing subcutaneous and metastatic cancer grafts during exposure to AMF. This approach can also be applied to the synthesis of other metal-doped nanoparticles with core-shell structures. Consequently, this method can potentially be used for the development of novel nanoparticles with high heating performance, further advancing systemic magnetic hyperthermia for cancer treatment.

Noninvasive temperature mapping with MRI using chemical shift water-fat separation.

Tissues containing both water and lipids, e.g., breast, confound standard MR proton reference frequency-shift methods for mapping temperatures due to the lack of temperature-induced frequency shift in lipid protons. Generalized Dixon chemical shift-based water-fat separation methods, such as GE's iterative decomposition of water and fat with echo asymmetry and least-squares estimation method, can result in complex water and fat images. Once separated, the phase change over time of the water signal can be used to map temperature. Phase change of the lipid signal can be used to correct for non-temperature-dependent phase changes, such as amplitude of static field drift. In this work, an image acquisition and postprocessing method, called water and fat thermal MRI, is demonstrated in phantoms containing 30:70, 50:50, and 70:30 water-to-fat by volume. Noninvasive heating was applied in an Off1-On-Off2 pattern over 50 min, using a miniannular phased radiofrequency array. Temperature changes were referenced to the first image acquisition. Four fiber optic temperature probes were placed inside the phantoms for temperature comparison. Region of interest (ROI) temperature values colocated with the probes showed excellent agreement (global mean +/- standard deviation: -0.09 +/- 0.34 degrees C) despite significant amplitude of static field drift during the experiments.

Correction of breathing-induced errors in magnetic resonance thermometry of hyperthermia using multiecho field fitting techniques.

PURPOSE: Breathing motion can create large errors when performing magnetic resonance (MR) thermometry of the breast. Breath holds can be used to minimize these errors, but not eliminate them. Between breath holds, the referenceless method can be used to further reduce errors by relying on regions of nonheated fatty tissue surrounding the heated region. When the surrounding tissue is heated (i.e., for a hyperthermia treatment), errors can result due to phase changes of the small amounts of water in the tissue. Therefore, an extension of the referenceless method is proposed which fits for the field in fatty tissue independent of temperature change and extrapolates it to the water-rich regions. METHODS: Nonheating experiments were performed with male volunteers performing breath holds on top of a phantom mimicking a breast with a tumor. Heating experiments were also conducted with the same phantom while mechanically simulated breath holds were performed. A nonheating experiment was also performed with a healthy female breast. For each experiment, a nonlinear fitting algorithm was used to fit for temperature change and B0 field inside of the fatty tissue. The field changes were then extrapolated into water-rich (tumor) portions of the image using a least-squares fit to a fifth-order equation, to correct for field changes due to breath hold changes. Similar results were calculated using the image phase, to mimic the use of the referenceless method. RESULTS: Phantom results showed large reduction of mean error and standard deviation. In the non-heating experiments, the traditional referenceless method and our extended method both corrected by similar amounts. However, in the heating experiments, the average deviation of the temperature calculated with the extended method from a fiber optic probe temperature was approximately 50% less than the deviation with the referenceless method. The in vivo breast results demonstrated reduced standard deviation and mean. CONCLUSIONS: In this paper, we have developed an extension of the referenceless method to correct for breathing errors using multiecho fitting methods to fit for the B0 field in the fatty tissue and using measured field changes as references to extrapolate field corrections into a water-only (tumor) region. This technique has been validated in a number of situations, and in all cases, the correction method has been shown to greatly reduce temperature error in water-rich regions. The method has also been shown to be an improvement over similar methods that use image phase changes instead of field changes, particularly when temperature changes are induced.

Accuracy of real time noninvasive temperature measurements using magnetic resonance thermal imaging in patients treated for high grade extremity soft tissue sarcomas.

PURPOSE: To establish accuracy of real time noninvasive temperature measurements using magnetic resonance thermal imaging in patients treated for high grade extremity soft tissue sarcomas. METHODS: Protocol patients with advanced extremity sarcomas were treated with external beam radiation therapy and hyperthermia. Invasive temperature measures were compared to noninvasive magnetic resonance thermal imaging (MRTI) at 1.5 T performed during hyperthermia. Volumetric temperature rise images were obtained using the proton resonance frequency shift (PRFS) technique during heating in a 140 MHz miniannular phased array applicator. MRTI temperature changes were compared to invasive measurements of temperature with a multisensor fiber optic probe inside a #15 g catheter in the tumor. Since the PRFS technique is sensitive to drifts in the primary imaging magnetic field, temperature change distributions were corrected automatically during treatment using temperature-stable reference materials to characterize field changes in 3D. The authors analyzed MRT images and compared, in evaluable treatments, MR-derived temperatures to invasive temperatures measured in extremity sarcomas. Small regions of interest (ROIs) were specified near each invasive sensor identified on MR images. Temperature changes in the interstitial sensors were compared to the corresponding ROI PRFS-based temperature changes over the entire treatment and over the steady-state period. Nonevaluable treatments (motion/imaging artifacts, noncorrectable drifts) were not included in the analysis. RESULTS: The mean difference between MRTI and interstitial probe measurements was 0.91 degrees C for the entire heating time and 0.85 degrees C for the time at steady state. These values were obtained from both tumor and normal tissue ROIs. When the analysis is done on just the tumor ROIs, the mean difference for the whole power on time was 0.74 degrees C and during the period of steady state was 0.62 degrees C. CONCLUSIONS: The data show that for evaluable treatments, excellent correlation (deltaT < 1 degrees C) of MRTI-ROI and invasive measurements can be achieved, but that motion and other artifacts are still serious challenges that must be overcome in future work.

Hyperthermia MRI temperature measurement: evaluation of measurement stabilisation strategies for extremity and breast tumours.

PURPOSE: MR thermometry using the proton resonance frequency shift (PRFS) method has been used to measure temperature changes during clinical hyperthermia treatment. However, frequency drift of the MRI system can add large errors to the measured temperature change. These drifts can be measured and corrected using oil references placed around the treatment region. In this study, the number and position of four or more oil references were investigated to obtain a practical approach to correct frequency drift during PRFS thermometry in phantoms and in vivo. MATERIALS AND METHODS: Experiments were performed in a 140 MHz four antenna mini-annular phased array (MAPA) heat applicator (for treatment of extremity tumours) and an applicator for heating of the breast, with symmetric and asymmetric positioning of the oil references, respectively. Temperature change PRFS images were obtained during an hour or more of measurement with no application of heat. Afterwards, errors in calculating temperature change due to system drift were quantified with and without various oil reference correction arrangements. RESULTS: Results showed good temperature correction in phantoms and in a human leg, with average errors of 0.28 degrees C and 0.94 degrees C respectively. There was further improvement in the leg when using eight or more oil references, reducing the average error to 0.44 degrees C, while the phantoms showed no significant improvement. CONCLUSIONS: These results indicate that oil reference correction performs well in vivo, and that eight references can improve the correction by up to 0.5 degrees C compared to four references.