Using amide proton transfer-weighted MRI to non-invasively differentiate mismatch repair deficient and proficient tumors in endometrioid endometrial adenocarcinoma

This study was approved by the Institutional Review Board and complied with ethical committee standards. Written informed consent was obtained from all participants. The sample size was estimated by considering the difference in APTw values between the dMMR and pMMR groups as the primary outcome. The error was set at 0.05, and the power level was set to 80%. Based on data from the pathology department, a standard deviation of 0.4 was expected and a proportion of 0.3 for the dMMR group was proposed. Therefore, a total sample size of 30 was estimated using the G*Power 3 (version 3.1.9.7) sample size calculation program. A higher number of patients was enrolled in consideration of potential dropouts and poor image quality.

From January 2018 to October 2020, 78 consecutive female patients aged 25–78 years (mean age, 47.0 ± 14.9 years) with suspicious endometrial lesions and without MR scanning contraindications were prospectively enrolled in this pelvic APTw MRI study. Twenty-six patients without postoperative pathological diagnosis were excluded. Fifty-two patients who underwent staging surgery within 2 weeks after MR scanning were included for further analysis (Fig. 1).

An empty bladder was needed. 10 mL of glycerin enema was administered into the rectum 30 min before pelvic MRI examination to reduce air in the rectum and sigmoid. No pre-medications were used to control uterine peristalsis. Pelvic scans were performed with a clinical whole-body 3.0-T MRI unit (Ingenia 3.0 T; Philips Healthcare, Best, the Netherlands). A 16-channel dS Torso coil, which enabled parallel imaging with embedded coils, was applied above the patients. T2-weighted, T1-weighted, and DWI sequences were obtained following with an APTw sequence. A detailed overview of the MRI parameters is listed in Table 1. An ADC map was generated by referring to the signal intensities of DWI with b values of 0 and 1000 s/mm².

A 3D APTw imaging sequence was used in this study, with long radiofrequency (RF) pulses at 2-μT amplitude applied to saturate the amide proton spin signals. In order to generate a continuous 2-s long saturation RF, two RF transmit coils were used and each was turned on for 500-ms alternatively to last four sections. To reduce CEST artifacts from the presence of fat in the pelvis, an asymmetric frequency-modulated pulse (chemical-shift-selective) was applied to suppress fatty tissue MR signals. The APTw sequence was repeated nine times at seven RF saturation offsets (for convenience, the water resonance frequency was set at 0 ppm in the z-spectrum for APT-related offset definitions). Saturation RF pre-pulses were applied selectively at 3.5 ppm which is proven to be the amide proton frequency offset [14]. The APT effect is proportional to the difference of MRI signals with and without saturation pulses at 3.5 ppm. However, the RF magnetic transfer effects have to be compensated for APTw quantification, so the APTw signal is defined as Magnetization Transfer Ratio (MTR) Asymmetry at ± 3.5 ppm. Moreover, B0 inhomogeneity strongly impacts the accuracy because Larmor frequencies of amide protons are shifted if B0 inhomogeneity exists [31]. Hence, two additional acquisitions with different echo shifts of 0.5 ms were performed with the same saturation pre-pulses, and the image volumes with three different echo times (all at offset of + 3.5 ppm) were used to derive the B0 field map. To compensate for B0 inhomogeneity, four more saturation frequency offsets (3.5 ± 0.8 ppm and − 3.5 ± 0.8 ppm) in z-spectrum were measured. The z-spectrum was aligned for each voxel, and the signals at targeting offsets, \(S\left( { - 3.5\,{\text{ppm}}} \right)\) and \(S\left( { + 3.5\,{\text{ppm}}} \right),\) were calculated using Lagrange interpolation [31]. In summary, APTw values were calculated according to the following equation:where the \(S\left( { - 1560{\text{ ppm}}} \right)\) was the signal from one acquisition with saturation RF at extra-large offset (− 1560 ppm) as the control for APTw quantification. The APTw specific absorption ratio (SAR) value was 1.1 W/kg, which fell within the U.S. Food and Drug Administration guidelines. The middle slice of the APTw images was identified based on the largest cross-section of the lesions present on conventional MR images selected by radiologists with 10–17 years of experience in interpreting MR images of the female pelvis.

The APTw calculations, including z-spectrum shift and interpolation, were performed online using a MR control console. Raw image datasets were transferred to a workstation (Intellispace Portal; Version 10.1.0.64190; Philips Healthcare, Best, the Netherlands) for post-processing. All MR images were reviewed by three radiologists (Xue HD, He YL, and Lin CY; observers 1, 2, and 3 with 17, 10, and 5 years of experience in interpreting pelvic MR images, respectively), who had previously evaluated over 300 APTw images and were blinded to the patients’ clinical and histopathologic data. Based on the APTw image quality evaluation criteria for uterine cervical cancer [32], the three observers independently ranked the APTw images relative to image quality and measurement confidence on a 5-point Likert scale with respect to image blur, distortion, motion and ghosting artifacts, lesion recognition, and contour delineation. Table 2 summarizes the marking scale and the APTw image scores assessed by the three readers.

Two observers (Xue HD and He YL; observers 1 and 2) independently measured the APTw values with image quality scores of no less than 3. With reference to conventional MR images, the observers selected a single APTw image and ADC map slice with the maximum lesion area and drew a smooth contour region of interest (ROI) to cover the lesion. The mean APTw and ADC values in areas of the ROIs were recorded.

Surgically resected specimens stained with hematoxylin and eosin, and MMR immunohistochemical staining were reviewed by a pathologist (Chen B with 10 years of experience in gynecological pathology), who was blinded to the clinical and imaging data. The aggressiveness of each EEA specimen was categorized into three groups based on the FIGO grading system criteria: grade 1, well-differentiated EEA; grade 2, moderately differentiated EEA; and grade 3, poorly differentiated EEA [33]. The expression of MMR proteins MLH1, MSH2, MSH6, and PMS2 was detected by immunohistochemistry using an FFPE tissue microarray and a Ventana Benchmark XT autostainer (Ventana Medical Systems Inc., Tucson, AZ) according to the manufacturer’s protocols. The absence of nuclear staining in tumor cells was considered a “loss of expression” with intervening stromal positivity serving as an internal control. The complete expression of all four MMR proteins was considered a case of pMMR. The loss of at least one MMR protein was considered a case of dMMR [34].

Statistical analysis was performed using standard statistical software (Prism 8, GraphPad Software, San Diego, CA; SPSS Statistics 23, IBM, NY, USA). For image quality assessment, Kendall’s W test was used to evaluate the inter-observer agreement. Inter-class correlation coefficients (ICCs) were computed to evaluate the inter-observer agreement of the APTw value measurements. Kendall’s W and ICC values of less than 0.4, 0.41–0.75, and greater than 0.75, were considered to indicate positive but poor, good, and excellent agreement, respectively. The Shapiro–Wilk test was performed to evaluate the normality of the distribution of APTw and ADC values. APTw and ADC values are presented as the mean ± standard deviation. For normally distributed data, Student’s t-test was performed to compare APTw and ADC values between the dMMR and pMMR groups. APTw and ADC values were compared among the three grades using a one-way analysis of variance with Scheffe’s post hoc test. Statistical significance was set at p < 0.05. Receiver operating characteristic (ROC) analysis was performed to determine the feasible threshold value with assessment of sensitivity and specificity.

In total, 49 patients with EEA aged 26–77 years (mean age, 50.9 years) were enrolled for evaluation of APTw image quality. Table 2 summarizes the APTw image scores assessed by the three observers, which exhibited excellent agreement (Kendall’s W = 0.867, p < 0.001). Most cases were ranked with a score of 4. Poor image quality was observed in 28.6% of the cases due to distortion and artifacts.

APTw values were obtained in 35 cases of EEA aged 26–76 years (mean age, 50.0 years). Patient demographics are presented in Table 3. The mean ROI area was 315.7 (21.1–1288.0) mm². The APTw values were normally distributed (p = 0.890). The mean APTw value was 2.9 ± 0.5% with an inter-observer ICC of 0.985 (95% confidence interval [CI]: 0.971–0.993).

For ADC value measurements, the mean ROI area of endometrial lesions was 300.2 (32.3–1021) mm². ADC values were normally distributed (p = 0.335). The mean ADC value was 0.895 ± 0.101 × 10⁻³ mm²/s with an inter-observer ICC of 0.976 (95% CI: 0.954–0.988).

The dMMR and pMMR groups comprised 9 and 26 cases, respectively. The mean APTw value was significantly higher in the dMMR group than in the pMMR group (3.2 ± 0.3% and 2.8 ± 0.5%, respectively; p = 0.019; Fig. 4). The area under the curve of ROC analysis for differentiating the dMMR and pMMR groups was 0.778 (Fig. 5). The feasible threshold value was determined to be 3.0%, with a sensitivity of 88.9% and specificity of 69.2%. No significant differences were observed in mean ADC values between the dMMR group (0.874 ± 0.104 × 10⁻³ mm²/s) and pMMR group (0.903 ± 0.100 × 10⁻³ mm²/s; p = 0.476, Fig. 4).

The mean APTw values of grade 1, grade 2, and grade 3 were 2.9 ± 0.5%, 2.9 ± 0.4% and 3.1 ± 0.6%, respectively; the mean ADC values of grade 1, grade 2 and grade 3 were 0.922 ± 0.119 × 10⁻³ mm²/s, 0.876 ± 0.070 × 10⁻³ mm²/s and 0.845 ± 0.110 × 10⁻³ mm²/s, respectively (Fig. 4). No significant differences in APTw or ADC values were observed among the three histologic grades (p = 0.766 and p = 0.295, respectively).