Evaluation of renal masses with CEST MRI: Protocol optimization and preliminary results

Abstract

Purpose

This study investigates the application of CEST MRI in human kidneys, with the overarching goal of developing CEST MRI for renal mass characterization. Here, the focus is on the development of a protocol and the proof‐of‐principle evaluation of CEST for the characterization of clear cell renal cell carcinoma (ccRCC).

Methods

Ten healthy volunteers and five patients diagnosed with ccRCC underwent CEST MRI scans at 3 T. The sequence performance was characterized in healthy volunteers and tested using three saturation power levels, B_1rms = 0.7, 1.2, and 1.6 μT. We compared the magnetization transfer ratio asymmetry (MTR_asym) values at 1.0, 2.0, and 3.5 ppm between the kidney cortex of healthy volunteers and the cortex/renal masses of renal cell carcinoma (RCC) patients. We also investigated the correlation of CEST with tumor characteristics, such as volume, necrosis percentage, and histopathological grade.

Results

Significant differences in MTR_asym values were observed between ccRCC tumor regions and healthy kidney cortex in all three compared frequency offset ranges (p < 0.05). At the same time, MTR_asym values in the cortex of the contralateral kidney in RCC patients were not significantly different when compared with the values of healthy kidney cortex values (p > 0.7).

Conclusion

CEST MRI demonstrates considerable potential as a non‐invasive imaging tool for the characterization of renal masses.

1. INTRODUCTION

Renal cell carcinoma (RCC) is among the top 10 most common cancers in both men and women. ¹ , ² The incidence of kidney cancer has increased in the last decades, primarily because of the widespread use of diagnostic imaging and the increased incidental detection of small renal masses ³ (SRMs) (up to 4 cm in size). Up to 20% of SRMs are benign ⁴ and many malignant tumors are indolent, limiting the potential benefit from therapy. ⁵ Furthermore, nephron loss secondary to surgery is associated with increased cardiovascular events and decreased long‐term survival. ⁸ , ⁹ Although percutaneous renal mass biopsy (RMB) offers an accurate diagnosis of cancer and histologic subtype, ⁶ it is not reliable in the determination of tumor grade, limiting its utility in determining the aggressiveness in renal masses. Ideally, surgical resection of renal masses should be avoided in patients with benign masses and in those with indolent malignancies and limited life expectancy, whereas patients with malignant aggressive tumors should be offered prompt treatment. Unfortunately, there are currently no reliable methods to predict the natural history of an incidentally discovered renal mass.

In recent years, there has been an increased emphasis on using non‐invasive imaging to characterize renal masses, the so called “virtual biopsy”. ⁷ The role of multiparametric MRI (mpMRI) in the histological subtyping of RCC and in the differentiation of benign from malignant renal masses has been reported. ⁸ , ⁹ However, a robust distinction between benign and malignant tumors (e.g., oncocytoma vs. clear cell RCC [ccRCC]) ⁹ is still not possible. Similarly, a method to differentiate indolent and aggressive malignant masses is lacking. A potential advantage of mpMRI is the combination of its inherent high‐resolution and excellent soft tissue contrast with quantitative maps of perfusion, tissue micro‐structure, and many other metabolic and molecular based parameters. ⁸ , ⁹

CEST imaging using endogenous sources of signal is gaining attention as a tool to characterize malignancy. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ CEST MRI relies on the detection of molecules with exchangeable protons, providing enhanced contrast sensitivity to molecular alterations. Most CEST studies have focus on detection of brain tumors ¹² , ¹³ , ¹⁴ , ¹⁶ , ¹⁷ and stroke, ¹⁸ although recent studies have illustrated the role of CEST in other body malignancies, such as determining breast cancer aggressiveness. ¹⁹ , ²⁰ , ²¹ CEST MRI for renal masses poses several technical challenges such as vulnerability to artifacts from physiological motion and from contaminating fat signals, which require advanced correction techniques. ²² , ²³ , ²⁴ This study focuses on the optimization of endogenous CEST MRI for evaluation of renal masses combining a three‐point Dixon CEST acquisition with a timed‐breathing approach, ²⁵ and post‐processing motion correction to reduce breathing‐related motion artifacts. We implemented and evaluated the acquisition protocol in healthy volunteers and conducted a proof‐of‐principle study in a small number of RCC patients.

2. METHODS

2.1. Subjects

The research was approved by the local institutional review board (IRB) and was conducted in compliance with the prescribed guidelines. Written informed consent was obtained from all subjects before enrollment. Ten healthy volunteers (5/5 male/female) without known kidney disease and five patients (4/1 male/female) with known renal masses were recruited for this study. The patients underwent abdominal MRI before radical nephrectomy. The clinical pathology report of the resected material served as the diagnostic reference standard. All renal masses were evaluated by a board‐certified uropathologist. For accurate colocalization of the excised tissue and MR images, the surgical specimens were positioned to match their in vivo anatomic orientation using fiducial markers placed during surgery by a urologist. ²⁶ , ²⁷

2.2. MRI

MRI was performed on a 3 T whole body scanner (Ingenia, Philips Healthcare) equipped with a dual‐transmit‐channel body coil for RF transmit and a 28‐channel phased‐array abdominal coil for signal reception.

For healthy volunteers, planning was performed using coronal and transverse T₂‐weighted single‐shot multi‐slice breath‐hold fast‐spin echo images (SShTSE). The CEST acquisition was prescribed matching the resultant coronal T₂‐weighted image with the largest cross‐sectional kidney area. When possible, both kidneys were captured in a single acquisition. In RCC patients, same SShTSE sequence was used to generate T₂‐weighted images. The CEST acquisition plane, using either a single slice or two slices, corresponded to the maximum cross‐section of the tumor mass as identified on the T₂‐weighted images.

The CEST RF saturation consisted of 40 sinc‐Gaussian RF pulse elements, 50 ms each, resulting in a saturation pulse duration of 2.0 s, enabled by alternated parallel transmission. ²⁸ RF shimming based on 2‐channel B₁ maps was independently performed in two modes, to homogenize B₁ in alternated transmission for saturation pulses as well as for parallel two channel transmission for imaging (excitation) pulses. The CEST images were acquired using 2D 3‐point multi‐echo Dixon sequence with centric k‐space ordering, ²⁰ , ²⁹ using a single‐shot multi‐echo T₁‐weighted turbo field echo (TFE) sequence. We chose 3 echoes (3‐point Dixon method), as this represents the minimum number necessary to derive water‐only, fat‐only, and B₀ information from the MRI scans. For the TFE sequence, a flip angle of 10° was used for the excitation pulse.

For CEST imaging, the coronal FOV was 300 × 360 mm² with an acquisition voxel size of 3.0 × 3.0 mm². In the healthy control group, the slice thickness was 10 mm (volunteers 1–5) or 5 mm (volunteers 6–10) with the corresponding TR/TE1/∆TE = 3.4/1.03/0.6 ms or 3.9/1.10/0.8 ms, which were the shortest possible in all cases. The frequency offset range was set to ±6 ppm, with 15 Z‐spectral points uniformly distributed, along with an additional reference image acquired at an offset of −1560 ppm. For healthy subjects, three different B_1rms power levels were tested: 0.7, 1.2, and 1.6 μT. These corresponded to flip angles (FA) of 540°, 900°, or 1260° per RF saturation pulse element. The total scan time for the whole Z‐spectrum was about 2 min 40 s for one power level.

In patients, the slice thickness was 10 mm and B_1rms = 1.2 μT. CEST scans were performed as part of multiparametric research protocol and the CEST portion was limited to a total time of up to ˜15 min, including acquisition of the planning images. Therefore, only one power level was tested consistently in all patients. In two patients, two additional CEST scans were performed using B_1rms = 0.7/1.6 μT and B_1rms = 0.7/1.5 μT, respectively. The highest B_1rms tested in the second patient was reduced slightly compared to others to adhere to the patient‐specific SAR requirements.

To minimize motion artifacts, a timed‐breathing technique was used during the acquisition of the CEST datasets, similar to the method used in the previously reported arterial spin labeling (ASL) studies. ²⁵ Briefly, the subject is instructed to synchronize their breathing with the image acquisition by the MRI operator. The commonly used breath hold commands are used: “breathe‐in, breathe‐out, hold your breath.” However, in timed breathing, a short breath hold is required, and the overall breathing cycle is only slightly extended as compared to regular breathing. In this study, the timed breathing cycle was 10 s. The timed‐breathing period of 10 s was aligned with the CEST MRI acquisition including the RF saturation pulse train occurring during the “breathe‐in, breathe‐out” phase, and the single‐shot multi‐echo TFE readout being synchronized with a short breath hold (˜2 s). One Z‐spectrum point was acquired per breathing cycle.

2.3. Data analysis

Standard vendor‐provided Dixon post‐processing was used to generate six different types of images for each frequency offset: (1) the source images (a total of three magnitude images, one for each TE), (2) the water‐only image, (3) the fat‐only image, (4) the in‐phase (IP) image, (5) the out‐of‐phase (OP) image, and (6) the B₀ map.

Timed breathing alone is not sufficient to remove all motion artifacts present in the imaging data. ³⁰ To address residual motion, we used an approach termed structuralized mutual information (SMI)‐based retrospective motion correction. ³⁰ By considering both structural information and intensity characteristics of the images simultaneously, SMI offers a robust framework for accurately addressing motion artifacts in medical imaging datasets.

The water‐only CEST images were processed on a pixel‐by‐pixel basis using custom routines developed in MATLAB (The Mathworks, Inc). The Z‐spectrum points were interpolated to 151 and the minimum of Z‐spectrum approach was used for B₀ correction. ¹⁶ The magnetization transfer asymmetry ratio (MTR_asym) was used for the quantification of the CEST effect. MTR_asym maps were generated by pixel‐wise integration of the MTR_asym curves over three specific frequency ranges: (1) 1.0 ± 0.2 ppm, (2) 2.0 ± 0.2 ppm, and (3) 3.5 ± 0.2 ppm. These ranges are commonly associated, in a coarse approximation, with protons in hydroxyl (−OH), amine/guanidinium (−NH₂), and amide (−NH) functional groups, respectively.

2.4. ROI placement

In both healthy volunteers and patients, the placement of regions of interest (ROIs) was performed after the data processing, using the reference image (after all the Z‐spectrum images were co‐registered). The renal cortex was delineated manually by a fellowship‐trained radiologist avoiding the inclusion of the renal medulla. For acquisitions displaying both kidneys, a separate renal cortical ROI was drawn for each kidney. The average cortex ROI area in healthy volunteers was 1377 ± 300 mm². The analysis was limited to the renal cortex region because reliable identification and delineation of the medulla was challenging on the available water‐only and T₂‐weighted images. For patients, a fellowship‐trained radiologist drew an ROI including the entire renal mass. The ROIs were drawn on the water‐only images (either on the reference image or the image acquired at 6 ppm for increased contrast). The high‐resolution SShTSE images were used as reference to assist with renal mass delineation.

2.5. Statistical analysis

Unpaired t tests were used to assess the significance of the differences in CEST effects between different tissues of interest from healthy volunteers and RCC patients. Specifically, comparisons were made between the renal regions affected by tumor, the contralateral unaffected renal cortex in RCC patients, and the renal cortex of healthy volunteers. A significance level of p < 0.05 was considered to indicate a statistically significant difference.

3. RESULTS

3.1. Healthy volunteer results

Figure 1 demonstrates typical Z‐spectra in the renal cortex of a healthy volunteer. The corresponding reference image indicates the ROI placement in the renal cortex (Figure 1A). Figure 1B demonstrates the ROI‐averaged “raw” Z‐spectrum points post motion correction, but before pixel‐by‐pixel interpolation and B₀ correction, acquired using B_1rms = 1.2 μT. Figure 1C shows the corresponding B₀ map and Figure 1E displays the fully interpolated and corrected Z‐spectra and MTR_asym. From Figure 1D–F there is a noticeable increase in the Z‐spectra width and background magnetization transfer (MT) (as indicated by the overall intensity decrease) with the increasing saturation power level. Additional features include increased MTR_asym around 1 ppm noticeable here at B_1rms = 1.2 μT, and a peak around 3.5 ppm visible at B_1rms = 0.7 μT (Figure 1B,C, arrows). There might be an additional relayed nuclear Overhauser effect (rNOE) at the negative side of the spectrum that influences the asymmetry analysis, as indicated by slightly negative MTR_asym values at increasing off‐resonance (above ˜3 ppm). MT asymmetry might also contribute to the observation.

FIGURE 1.

CEST MRI in a healthy male volunteer. (A) The region of interest (ROI) placed over the renal cortex and shaded in red on the reference water‐only images. (B) ROI‐averaged Z‐spectrum points post motion correction, but before interpolation and B₀ correction, acquired using B_1rms of 1.2 μT. (C) B₀ map corresponding to acquisition shown in (A,B,E). (D–F) ROI averaged Z‐spectra (blue) and magnetization transfer ratio asymmetry (MTR_asym) curves (red) of the kidney cortex at different B_1rms of 0.7, 1.2, and 1.6 μT. Arrows indicate subtle features of the MTR_asym: Increase at ˜1 ppm (D), increase at ˜1 ppm and 3.5 ppm (E), and slightly negative values above ˜3 ppm (F).

Figure 2 shows MTR_asym maps in a different healthy volunteer at three different power levels. The MTR_asym values were close to zero and slightly negative, and the overall signal intensity across the entire kidney was relatively uniform. However, some “hot spots” were observed at lower off‐resonance values (1 ppm) at the edges of the kidneys and close to the spine (Figure 2, red arrows). There was a 10% to 20% variation in CEST effects between the two slices acquired, which might be associated with T₁ effects.

FIGURE 2.

Magnetization transfer ratio asymmetry (MTR_asym) maps in a healthy male volunteer at different saturation power levels, B_1rms = 0.7 μT (A–D), 1.2 μT (E–H), and 1.6 μT (I–L). (A,E,I) Show reference water‐only images, whereas MTR_asym maps at 1, 2, and 3.5 ppm are shown in (B,F,J), (C,G,K), and (D,H,L), respectively. Red arrows indicate “hot spots” of increased intensity.

Figure 3 and Table S1 display the dependence of MTR_asym values on the power level. The values were averaged across the cortices of all healthy volunteers. Notably, increasing B_1rms led to a decrease in MTR_asym at 1 ppm, but increase in the MTR_asym values for 2 and 3.5 ppm. The letter two became more positive as the saturation power level increased. At the same time, the MTR_asym values in healthy volunteers are very low for all power levels tested.

FIGURE 3.

Group mean magnetization transfer ratio asymmetry (MTR_asym) values and SD in the cortex of healthy volunteers as a function of saturation power, B_1rms. Blue circle, red squares, and green triangles correspond to MTR_asym at 1, 2, and 3.5 ppm, respectively.

3.2. Patient results

3.2.1. Characteristics of patients

All patient participants underwent radical nephrectomy following their MRI scans. Table 1 presents the clinical and pathological characteristics of these participants. All patients presented with ccRCC, and the maximum tumor diameter ranged from 4.9 to 12.0 cm. The histological grades were 2 to 4, according to the nucleolar‐based World Health Organization (WHO) classification. Patient 3 presented with involvement of regional lymph nodes and distant metastasis, whereas no lymph node involvement or distant metastasis were identified in other patients.

TABLE 1.

Pathology data and CEST MRI results of the tumor lesions in patients.

Patient	Histological type	Tumor maximum size (cm)	Tumor volume (cm3)	Necrosis (%)	Histological grade 1–4	Pathological staging
1	Clear cell	12.0	438.8	20	3	T3a
2	Clear cell	11.0	290.1	0	2	T3a
3	Clear cell	6.4	110.6	5	3	T3a
4	Clear cell	5.5	28.8	0	3	T3a
5	Clear cell	4.9	40.4	5	4	T1b

3.2.2. CEST MRI results in patients

Figure 4 illustrates zoomed‐in high resolution SShTSE T₂‐weighted images, CEST MRI maps and macroscopic pathology images post radical nephrectomy in two ccRCC patients, patient 1 and 3 (Table 1). Figure 5 demonstrate Z‐spectra and MTR_asym from tumors and contralateral kidneys in the same patients 1 and 3. The images in Figure 5A,B show full FOV CEST reference images with the tumor and contralateral cortex ROIs highlighted. The MTR_asym maps of the contralateral kidneys can be found in the Figure S1. The tumor Z‐spectra (Figure 5) present elevated MTR_asym values in all three off‐resonance frequency ranges compared to the values in the contralateral cortex of the same patient. For patient 1, the tumor values are MTR_asym (1 ppm) = 6% ± 4%, MTR_asym (2 ppm) = 6% ± 3%, and MTR_asym (3.5 ppm) = 6% ± 3% versus contralateral cortex values of MTR_asym (1 ppm) = 4% ± 1%, MTR_asym (2 ppm) = 4% ± 1% and MTR_asym (3.5 ppm) = 4% ± 1% (Figure 5C,E). For patient 3, the tumor values are MTR_asym (1 ppm) = 3% ± 3%, MTR_asym (2 ppm) = 4% ± 3%, and MTR_asym (3.5 ppm) = 2% ± 4% versus contralateral cortex values of MTR_asym (1 ppm) = −1% ± 4%, MTR_asym (2 ppm) = 1% ± 4%, and MTR_asym (3.5 ppm) = 0% ± 4% (Figure 5D,F). There is heterogeneity in the MTR_asym maps for all frequency ranges, which we attribute to the corresponding tumor heterogeneity. This is especially pronounced in patient 1 (Figure 4B–D). To further investigate MTR_asym in malignancy versus healthy tissue we have conducted group comparisons presented and discussed below. In patients 3 and 4, we collected data at the lower B_1rms = 0.7 μT and at higher B_1rms = 1.6 (patient 3) and 1.5 μT (patient 4) (Figure S2).

FIGURE 4.

CEST MRI results in patient 1 (A–E) and patient 3 (F–J)acquired using B_1rms = 1.2μT. (A,F) Show high resolution T₂‐weighted images acquired using SShTSE. Note the slice position is within 1 mm of the CEST imaging slice plane. The panels display magnetization transfer ratio asymmetry (MTR_asym) maps at 1 ppm (B,G), 2 ppm (C,H), and 3.5 ppm (D,I). (E,J) Show macroscopic pathology images post radical nephrectomy with tumor areas outlined in yellow.

FIGURE 5.

CEST in tumors versus contralateral kidney cortex acquired using B_1rms = 1.2μT. Examples from patient 1 (A,C,E) and patient 3 (B,D,F). Full FOV CEST reference water only images (A,B) with shaded areas denoting specific region of interest (ROI) used to create Z‐spectra (C,D) and magnetization transfer ratio asymmetry (MTR_asym) (E,F) curves. Red box highlights kidney with tumor, blue box—contralateral kidney. Red and blue lines in (C–F) correspond to tumor and contralateral cortex results, respectively.

3.2.3. Comparisons of MTR_asym values among different target groups

Group comparisons were conducted for the MTR_asym values at 1 ppm, 2 ppm, and 3.5 ppm. ROI averaged results from three types of regions were compared: the renal masses in patients, the contralateral uninvolved (i.e., of the kidney without the renal mass) renal cortex in patients, and the renal cortex in healthy volunteers. Patient 2 contralateral kidney (both slices) and patient 4 contralateral kidney (one slice) was excluded from the analysis because of only partial visibility in images and prohibitive partial volume effects. The results are summarized in Table 2 and Figure 6. Figure 6 dots show individual data points, from each slice and each tumor. The mean MTR_asym values in the renal masses in patients were significantly higher than those in the renal cortex of healthy volunteers (p < 0.05). MTR_asym values were significantly higher in the renal masses versus values in the contralateral cortex in the patient group for MTR_asym (1 ppm) (p = 0.003) and MTR_asym (2 ppm) (p = 0.003), but the difference with MTR_asym (3.5 ppm) values was not statistically significant (p = 0.07). There were no significant differences in the MTR_asym values between the contralateral renal cortex in RCC patients and the renal cortex in healthy volunteers (p > 0.05).

TABLE 2.

Group mean comparison results of MTR_asym values at 1 ppm, 2 ppm, and 3.5 ppm in ROIs over three types of regions: (1) the tumor lesion in RCC patients, (2) the contralateral non‐lesioned renal cortex in RCC patients, and (3) the renal cortex in healthy volunteers. All values are given as percentages (%).

Mean MTRasym at offset	Renal cortex in healthy volunteers (n = 10)	Tumor lesions in RCC patients (n = 5)	p Value
1 ppm	1 ± 3	4 ± 3	0.008
2 ppm	0 ± 3	4 ± 2	0.000
3.5 ppm	−2 ± 3	2 ± 3	0.012

Mean MTRasym at offset	Renal cortex in healthy volunteers (n = 10)	Contralateral renal cortex in RCC patients (n = 5)	p Value
1 ppm	1 ± 3	−1 ± 3	0.784
2 ppm	0 ± 3	−0 ± 3	0.911
3.5 ppm	−2 ± 3	−1 ± 4	0.784

Mean MTRasym at offset	Tumor lesions in RCC patients (n = 5)	Contralateral renal cortex in RCC patients (n = 5)	p Value
1 ppm	4 ± 3	−1 ± 3	0.003
2 ppm	4 ± 2	−0 ± 3	0.003
3.5 ppm	2 ± 3	−1 ± 4	0.072

Abbreviations: MTR_asym, magnetization transfer ratio asymmetry; RCC, renal cell carcinoma, ROIs, regions of interest.

FIGURE 6.

Comparison between group averaged magnetization transfer ratio asymmetry (MTR_asym) values in regions of interest (ROIs) over: Tumor areas in renal cell carcinoma (RCC) patients (red), the contralateral non‐lesioned renal cortex in RCC patients (green), and the renal cortex in healthy volunteers (blue). Stars mark statistically significant differences as defined by p $\ge$0.05, “ns” marks non‐significant (p < 0.05). Dots show individual data points.

3.2.4. CEST versus tumor volume

MTR_asym was correlated with tumor volume in patients (Table 1). Figure 7 displays MTR_asym versus tumor volume. The MTR_asym increased with the tumor volume. To investigate the dependence further we have used least square fits of semi‐logarithmic curves to the results (Figure 7, solid lines) with R² of 0.45, 0.74, and 0.77 for 1, 2, and 3.5 ppm, confirming the observed trends. The MTR_asym (2 ppm) and MTR_asym (3.5 ppm) values (red squares and green triangles, respectively, Figure 7) are higher in larger tumors with increased R². The fitting confidence of MTR_asym (1 ppm) values with tumor size is not as strong, but it still follows the same trend.

FIGURE 7.

ROI and patient‐averaged magnetization transfer ratio asymmetry (MTR_asym) values versus tumor volume . (A) MTR_asym values versus tumor volume with blue circles (1 ppm), red squares (2 ppm), and green triangles (3.5 ppm). The solid lines show semi‐logarithmic curve fitted to the growths, with R² of 0.45 (for 1 ppm), 0.74 (for 2 ppm), and 0.77 (for 3.5 ppm).

3.2.5. CEST for assessment of tumor heterogeneity

Figure 8 shows Z‐spectra from different areas in a heterogeneous ccRCC (patient 1) exhibiting viable areas of tumor, necrosis, scar/hyalinization, hemorrhage, and adjacent uninvolved renal parenchyma at pathology. The insert in Figure 8 displays the ROI‐averaged values of MTR_asym for the different regions. Necrosis and renal parenchyma have markedly decreased values compared to viable tumor, scar, and hemorrhagic areas. Hemorrhagic areas show the highest MTR_asym values in combination with narrow Z‐spectra and increased M_z/M₀ (which serves as a surrogate measure of background MT, with lower values indicating stronger MT effects).

FIGURE 8.

Z‐spectra from selected areas in the tumor (patient 1) acquired using B_1rms = 1.2μT demonstrating tissue and CEST heterogeneity: (A) tumor, (B) scar/hyalinization, (C) necrosis, (D) residual benign kidney, and (E) hemorrhage. The figure inserts show region of interest (ROI) placement over the image with the RF off‐resonance 6 ppm.

4. DISCUSSION

In this study, we initiated optimization of a CEST protocol for evaluation of renal masses at 3 T MRI. The combination of a Dixon acquisition to remove spurious fat signals and a timed‐breathing technique provided a robust methodology to evaluate the cortical signal in kidneys of healthy volunteers and explore differences in renal masses. The Z‐spectra in healthy volunteers are relatively narrow at 3 T (Figure 1). The width increased with increasing B_1rms, as expected. Overall, the MTR_asym values in the cortex were low (Table S1) and similar to those reported by Stabinska et al. ²² that used a 2‐point Dixon method. In some cases, as shown in Figure 1, peak‐like features were observed at distinct off‐resonance values (arrows). In Figure 1, there is a small peak at 3.5 ppm most evident at the lowest power level used here (0.7 μT), which might be attributed to amide groups and associated cytosolic proteins, similar to brain and other tissues. ¹⁶ , ¹⁷ There are additional features at the frequency ppm range of 0.5 to 2 ppm, most pronounced at 0.7 and 1.2 μT. Previous studies indicated that the 2 ppm signal was associated with urea, ³¹ , ³² , ³³ although more recently, urea infusion was correlated with the 1 ppm signal. ³⁴ , ³⁵ Notably, prior studies focused on urea signal in the urine within the collecting system, whereas this study analyzed the cortex where urea is expected to contribute less to the overall CEST signal. Our observations are in agreement with the animal model study by Shin et al., ³⁴ where an MTR_asym close to zero was reported in healthy renal cortex. Overall, MTR_asym was slightly negative at higher off‐resonance values, above ˜3 ppm (Figures 1 and 2), which might be indicative of asymmetry in the underlying MT, similar to previous observations in the brain. ³⁶ Additional source of negative asymmetry might be rNOE, which is hard to distinguish at relatively low field of 3 T.

The CEST sequence resulted in slightly negative values of MTR_asym at 2 ppm and 3.5 ppm for higher saturation powers (Figure 2) (B_1rms = 1.2 and 1.6 μT). Quantitative maps of MTR_asym (1 ppm) showed mostly positive values, especially at the lowest saturation power (Figure 2). Residual spatial inhomogeneities of the effects were observed at all off‐resonance values, especially at MTR_asym (1 ppm). This off‐resonance value is close to the water resonance and the maps are more susceptible to imperfections in B₀ and B₁.

We also observed a 10% to 20% variation in CEST effects between the two slices acquired. This might be associated with the residual T₁ effect. The first slice was acquired after prolonged delay followed by saturation, whereas the second slice was acquired following 2 s of RF saturation, without an additional delay. This might lead to variations in reference intensities for structures with longer T₁, such as the kidney papilla.

Although timed breathing mediated some of the motion, it was not enough for all of the cases, and additional retrospective motion correction in the processing was needed. Figures S3 and S4 demonstrate two cases of motion correction. One is for the patient 1 where residual motion influence was moderate: almost non‐observant in the large tumor ROI, but noticeable in the residual tissue ROI. Another example is from a healthy volunteer subject, where motion non‐corrected Z‐spectra is non‐suitable to analysis (Figure S4). Motion correction using SMI can eliminate the artifacts (Figure S4). Approximately half of the healthy volunteer cases and patients needed post‐processing motion correction. Therefore, the motion correction was used in all the cases.

The MTR_asym (1 ppm) decreased with increasing power level, while MTR_asym (2 ppm) and MTR_asym (3.5 ppm) increased (Figure 3). Because MTR_asym (2 ppm) and MTR_asym (3.5 ppm) are negative, the increase reflects decrease in the absolute value of MTR_asym (i.e., |MTR_asym|). We hypothesize that this is the result of increased direct saturation combined with decrease in rNOE effects and asymmetry in the semi‐solid MTR background. Although no specific peaks were reliably identified at the negative off‐resonances, overall negative MTR_asym at 2 and 3.5 ppm indicate presence of rNOE and/or overall MT background asymmetry, as explained above.

In healthy volunteers, lower saturation power enhanced the detection and differentiation of MTR_asym at different off‐resonances. Because of the overall scan time, we could not test multiple saturation power levels in all patients. We compromised prioritizing B_1rms = 1.2 μT to conduct patient studies although the optimal power level for renal CEST acquisitions may require further research, similar to what was conducted for brain tumors and strokes. ³⁷ In two patients (3 and 4), we collected data at the lower B_1rms = 0.7 μT and at higher B_1rms = 1.6 μT (patient 3) and 1.5 μT (patient 4). The data are shown in Figure S2. Overall, patient data for different power levels shows trends similar to those in healthy volunteers.

The two examples shown in detail in Figures 4 and 5 demonstrate that the CEST measurements are heterogeneous, corresponding to tumor heterogeneity. Moreover, although the Z‐spectra may appear higher or lower in intensity (Figure 5C,D), MTR_asym values were consistently higher in the tumors compared to contralateral kidneys (Figure 5E,F, red lines vs. blue). Because of large tumor sizes and high possibility of infiltration in the adjacent tissue, no analysis was done comparing values in the ipsilateral remaining cortex.

There were different MTR_asym values in renal masses and renal cortex (Figure 6), with exception of MTR_asym (3.5 ppm) between renal masses and contralateral cortex. Although individual measurements could overlap, the difference was statistically significant for 1 and 2 ppm. MTR_asym (3.5 ppm), or amide proton transfer (APTw) imaging, is considered the most sensitive for glioblastoma (GBM) characterization and staging. ¹⁵ , ³⁸ However, a previous study in breast cancer indicated that MTR_asym (1 ppm) might be most sensitive to malignant alterations. ²⁰ These differences may reflect differences in the underlying tumor biochemistry. Wang et al. ²⁴ observed significant differences between RCC and renal parenchyma using APTw (i.e., MTR_asym [3.5 ppm]). The discrepancy in observations between the study by Wang et al. ²⁴ and this study may stem from the different acquisition sequences and saturation powers used. Although a gradient recalled echo‐based Dixon sequence was used in this study to eliminate lipid signals, Wang et al. ²⁴ used a fast spin‐echo (FSE) acquisition with spectral presaturation with inversion recovery (SPIR) for fat suppression. With SPIR fat suppression, APTw imaging is particularly sensitive to partial fat contents from the aliphatic fat at −3.4 ppm, leading to artificially increased MTR_asym(3.5 ppm). ³⁹ Clear cell RCC is characterized by increased lipid and glycogen content. ⁴⁰ Therefore, FSE with SPIR may artificially detect higher increase in MTR_asym(3.5 ppm) in these tumors. More extensive studies are needed to verify these observations and to fully assess different acquisition and fat removal methods for imaging renal masses.

ccRCC exhibited statistically significant higher MTR_asym values compared to renal cortex, both in the contralateral kidney at 1 and 2 ppm and that of healthy controls for all off‐resonance frequencies (Figure 6). The similar values observed in the contralateral renal cortex of patients with renal masses and healthy volunteers suggest that these differences are indeed related to the metabolic alterations in ccRCC. Potential explanations for this finding include increased cellular density, heightened metabolic activity, and altered chemical environment typical of malignant tissues. Moreover, heterogeneity in MTR_asym maps correlated with tumor heterogeneity at the histopathologic level (Figures 4 and 8). Although larger patient studies are needed, these early results suggest the potential utility of MTR_asym as a tool for characterization of renal masses.

A positive trend was observed correlating MTR_asym and calculated tumor volume that could be well approximated by a semi‐logarithmic curve (Figure 7). The increase in CEST effect with increased volume is in line with prior studies where increased MTR_asym values were observed in larger tumors, increasing with the tumor aggressiveness. ²⁴ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ Partly, the increase can be associated with larger heterogeneity comprising compartments with high fluidity (e.g., hemorrhage) known to show increased MTR_asym.

Statistical comparison between results from different histological grades (Table 1) was prohibitive because of small sample sizes (n = 1, 3 and 1 for histological grades 2, 3, and 4, respectively), but the data is summarized and can be seen in the Figure S5. More patient data is needed to verify these results and establish trends or correlations.

The exact origins of the increased MTR_asym in renal tumors versus healthy volunteers are challenging to identify yet and further investigation is needed. Generally, the CEST effect is influenced by multiple factors including changes in molecular concentration, T₁, exchange range (pH) and macromolecular components. The increase in MTR_asym (3.5 ppm) associated with cancer is widely reported in brain tumors ¹⁶ and forms the basis for successful application of APTw imaging to studies of GBMs, including differentiation molecular and molecular signatures. ⁴⁶ Changes in MTR_asym (2 ppm) have been associated with multiple sources. For example, changes in creatine concentration (CrCEST) are associated with this frequency and CrCEST was shown to decrease with brain tumor aggressiveness. ⁴⁷ However, guanidine amino protons also resonate approximately 2 ppm and have been shown to contribute to the CEST changes at 2 ppm in tumors. ⁴⁸ As noted above, 1 and 2 ppm were previously associated with urea in different contexts. ³¹ , ³² , ³³ , ³⁴ , ³⁵ In addition, increase in MTR_asym (1 ppm) has been associated with the increase in hydroxyl protons, such as glycogen ⁴⁹ and glucose. ⁵⁰ , ⁵¹ Importantly, ccRCC is known to accumulate large amounts of glycogen in the cytoplasm, whereas other subtypes of RCC and benign renal masses do not. ⁵² , ⁵³ Last, our previous study in breast cancer had indicated increased MTR_asym (1 ppm) in more aggressive Estrogen Receptor‐negative breast tumors as well as strongly correlating with the Ki‐67 proliferation index. ²⁰ Overall, our observations in Figures 5 and 6 are in line with these studies, indicating increased MTR_asym in tumors.

As a very preliminary assessment of the potential to study renal tumor heterogeneity with CEST, we have investigated different tumor areas in the largest tumor (patient 1). Figure 8 and Table S2 demonstrate differences in the Z‐spectra and MTR_asym values in different regions. Notably, the Z‐spectrum in the scar/hyalinization tissue (Figure 8B) is narrower and shows less MT background as can be approximately assessed from the Z‐spectrum width and the M_z/M₀ value at the higher ppm values (>6 ppm). At the same time, MTR_asym values in tumor tissue (Figure 8A) are slightly less than the scar values (Table S2). Interestingly, the Z‐spectrum of necrotic region (Figure 8C) shows similar width and background MT as the tumor region. However, the MTR_asym values are lower in necrosis versus tumor. The values in the necrotic area are comparable to the values observed in the remaining healthy kidney tissue (Figure 8D). Interestingly, previous studies of radiation necrosis versus viable tumors in GBM patients and animal models also demonstrated decreased APTw signal in necrosis versus tumor regions, while the MT effect was observed to be lower. ¹² , ⁵⁴ , ⁵⁵ The proposed explanation of the observation is that necrotic lesions contain reduced concentration of mobile proteins because of the loss of cytoplasm and organelles, leading to the reduced APTw effect. Although very preliminary, our results might be indicative of the similar behavior in necrosis versus viable renal tumor, although the high MT effect indicates intact boundaries or fibrotic structures. Finally, Figure 8E shows a Z‐spectrum from a hemorrhagic area. The Z‐spectrum is narrow, with decreased MT effect and increased MTR_asym values. Although very preliminary, this observation is in general agreement with previous observations in liquids and blood, ⁵⁶ where the absence of background MT leads to higher CEST effects. Overall, these early results are encouraging for future use in the studies of tumor heterogeneity. We can hypothesize that advanced analysis methods such as radiomics or deep learning may use different Z‐spectral features to assist with the identification of pathology heterogeneity within tumors.

The CEST MRI for renal masses is still in early stages and additional development is needed. Although we did not perform test–retest study here, the scan was repeated in patient 2 and the results are shown in Table S3 and Figure S6. There is a good agreement between the two scans. Such test–retest studies could be challenging because of motion and complete 3D coverage of the kidney would be advantageous, allowing co‐registration and pixel‐by‐pixel analysis of the scans.

Our study has several limitations. First, the number of patients with renal masses was small. However, the reported data was sufficient to identify statistical differences between ccRCCs and renal parenchyma. Furthermore, opportunities to further improve the acquisition protocol were identified. The technical development nature of this study required alteration to acquisition strategies (i.e., change in slice numbers) and future prospective clinical studies in patients with renal masses will require a standardized acquisition protocol. Third, the long acquisition time of the CEST sequence results in limited anatomic coverage and the need for optimization to fit within the breathing cycle. More advanced acquisition methods, including full 3D coverage of the kidney, are currently being developed. ⁵⁷ Note that future studies and protocol standardization would also require test–retest assessment that was not performed in this early proof‐of‐principle step. Fourth, the magnitude and specificity of the observed CEST effects may depend on multiple factors that are not fully understood, including saturation power, acquisition sequence, and post‐processing methods. Finally, only MTR_asym analysis was used in this preliminary study. More advanced methods, such as multi‐peak Lorentzian analysis should allow better delineation of various contributing factors, specifically rNOE quantification. Future studies will use fluid suppression techniques to avoid signal increase in scar/hyalinization or hemorrhagic regions of heterogeneous tumors.

5. CONCLUSIONS

In this study, we have combined a CEST‐mDixon sequence with guided timed breathing to assess the potential of CEST‐MRI to characterize renal masses. The results illustrated differences in MTR_asym between renal cortex and malignant renal masses. Further studies in large cohorts are needed to continue technical optimization and to fully explore the CEST MRI application in renal cancer.

CONFLICT OF INTEREST STATEMENT

Dr. J. Keupp and Dr. I. Dimitrov are employees of Philips.

Supporting information

Figure S1. CEST MRI results in Patient 1 (A–D) and Patient 3 (E–H), contralateral kidney. Panels A and E show water only images. Both patients had large liquid filled cysts. The panels display MTR_asym maps at 1 ppm (B, F), 2 ppm (C, G), and 3.5 ppm (D, H).

Figure S2. RF saturation power dependence of MTR_asym at 1 ppm (blue circles), 2 ppm (red squares) and 3.5 ppm (green triangles) averaged over the cancer tissue ROIs for Patient 3(A) and 4(B).

Figure S3. Example of moderate motion artifacts and improvement using retrospective SMI motion correction. Data from Patient 1, the results with the motion correction are also shown in Figure 8A,D.

Figure S4. Example of severe motion artifacts and improvement using retrospective SMI motion correction. Healthy Control Volunteer 6.

Figure S5. Patient averaged MTR_asym results vs. Hystological grade. Open blue circles, open red squares and open green inversed triangles corresponding to MTR_asym at 1, 2 and 3.5 ppm respectively.

Figure S6. Z‐spectra and MTR_asym curves measured in two different scans repeated within the same session.

Table S1. Group mean MTR_asym values and standard deviation in renal cortex of healthy volunteers as a function of saturation power, B_1rms. The values are shown in Figure 2 and are displayed as percentages (%).

Table S2. Example of CEST values reflecting tumor heterogeneity. The values correspond to ROIs shown in Figure 8 and are displayed as percentages (%).

Table S3. MTR_asym values measured in two different scans repeated within the same session. The are displayed as percentages (%).

Wang X, Keupp J, Dimitrov IE, et al. Evaluation of renal masses with CEST MRI: Protocol optimization and preliminary results. Magn Reson Med. 2025;94(6):2374‐2387. doi: 10.1002/mrm.30641