Reverse Encoding Distortion Correction of Female Pelvic Diffusion-weighted Imaging on a 1.5-T MR System: Influence on Image Quality and Apparent Diffusion Coefficient Measurements for Both In Vitro and In Vivo Studies.

On comparison of SNRs, there was no significant difference in the SNRs between DWI with and without RDC (28.9 ± 13.1 vs. 28.5 ± 14.4, P  = 0.85). The results of the ADC measurement for each phantom are shown in Table 1 . When each measured ADC was correlated with the standard references, ADC values were significantly correlated between DWI modalities and the standard reference (ρ = 0.99, P  < 0.001). The mean difference in the phantom with PVP contrast at 40% and 50% was significantly different between DWI with and without RDC ( P  < 0.003). In addition, the limits of agreement between each DWI and the standard reference were sufficiently small (limits of agreement between DWI with RDC: 0.02 ± 0.04 × 10 −3 mm 2 /s; limits of agreement between DWI without RDC: 0.02 ± 0.04 × 10 −3 mm 2 /s). All patients had clinically or pathologically diagnosed uterine and ovarian diseases, including 25 benign uterine and ovarian lesions, 4 with malignant uterine and ovarian lesions, and 2 others (Nabothian cyst [n = 2]). Details of the patient characteristics are shown in Table 3 . A representative case is shown in Figure 3 . A comparison of the SNR, CNR, ADC value, and DR for DWI with and without RDC is shown in Table 4 . The correlation of each ADC between DWI with and without an RDC is shown in Fig. 4 . The DRs of DWI with RDC were lower than those of DWI without RDC on the uterine corpus, cervix, ovary, and lesion ( P  ≤ 0.0003), although there were no significant differences in other quantitative indexes between the two DWI. There were significant correlations between the ADC values of the myometrium, cervix, and ovary between DWI with and without RDC (myometrium: ρ = 0.84, P  < 0.001; cervix: ρ = 0.90, P  < 0.001; ovary: ρ = 0.82, P  < 0.001). Table 5 shows the findings for the inter-observer agreement and qualitative image quality index. Inter-observer agreements for all qualitative indexes were determined as ‘substantial’ or ‘almost perfect’ (0.78 ≤ κ ≤ 0.97, P  < 0.0001). The results of the comparison of the OIQ, DS, and DCL between DWI with and without RDC are shown in Table 5 ; OIQ and DCL of DWI with RDC (OIQ: Median 5, interquartile range [IQR] 4–5; DCL: Median 5, IQR 5–5) were significantly higher than those without RDC (OIQ: Median 4, IQR 3–5, P  = 0.0004; DCL: Median 5, IQR 4–5, P  = 0.03). The DS of DWI with RDC (median, 1; IQR, 1–2) was significantly lower than that without RDC (median, 2; IQR, 1–3; P  = 0.0004).

This study included both in vivo and in vitro studies. This in vivo retrospective study was approved by the Institutional Review Board (IRB) of Fujita Health University Hospital, and written informed consent was obtained from each subject. This study was technically and financially supported by Canon Medical Systems Corporation. Five of the authors are employees of the Canon Medical Systems Corporation (K.Y., N.Y., Y.S., M.I., and M.O.) but did not have control over any of the data used in this study. For the in vitro study to evaluate the SNR, an American College of Radiology (ACR) MRI phantom ( Fig. 1 ) was used (J.M. Specialty Parts, San Diego, CA, USA) in accordance with previously published findings. 17 The ACR phantom consisted of a hollow cylinder of acrylic plastic that was closed at both ends. The inside diameter and length of the phantom were 190 mm and 148 mm, respectively, and they were filled with a solution of nickel chloride and sodium chloride (10 mm NiCl 2 and 75 mm NaCl). Inside the phantom, various structures were designed to measure geometric accuracy, high-contrast spatial resolution, slice thickness accuracy, slice position accuracy, image intensity uniformity, percent signal ghosting, and low-contrast detectability. The “NOSE” and “CHIN” indications, corresponding to the positions on the patient’s head, are etched on the phantom’s surface to ensure proper positioning of the phantom. For the in vitro study to evaluate ADC measurement accuracy or DWI, a quantitative diffusion phantom (Mini Diffusion Phantom Model 138; CaliberMRI [formerly HPD/QalibreMD], Boulder, CO, USA) consisting of 13 vials filled with varying concentrations of polyvinylpyrrolidone (PVP) in an aqueous solution was used ( Fig. 1 ). The phantom was specifically designed to quantitatively map the isotropic Gaussian diffusion of water molecules and generate physiologically relevant ADC values (E1). 18 The PVP concentrations in the phantom vials are 0% (vials 1–3), 10% (vials 4–5), 20% (vials 6–7), 30% (vials 8–9), 40% (vials 10–11), and 50% (vials 12–13) in an aqueous solution. 17 The space between the vials within the phantom was filled with a water bath (26°C) to eliminate thermal variability in the ADC measurements. Between January and March 2022, 44 consecutive patients with suspected female pelvic diseases (mean age, 43 years; range, 16–85 years) underwent DWI with and without RDC using a 1.5-T MR system, and ADC maps were reconstructed for each modality. The patient selection flowchart for this study is shown in Fig. 2 . According to the inclusion and exclusion criteria, 31 patients (mean age, 41 years; range, 24–80 years) were ultimately included in this study. All MR images were acquired with spin-echo EPI pulse sequences with and without the RDC technique on a 1.5-T MRI system (Vantage Orian; Canon Medical Systems, Tochigi, Japan) using a 16-channel phased-array surface coil (16ch Flex SPEEDER and Atlas SPEEDER Spine; Canon Medical Systems). The details of the RDC technique have been described previously. 12 , 13 Each DWI dataset was reconstructed using deep learning reconstruction (DLR; Advanced Intelligent Clear-IQ Engine [AiCE]; Canon Medical Systems) for body imaging. 12 , 19 – 23 The details of the MR protocols for the phantom study using the quantitative diffusion phantom and the ACR Large MRI phantom for accessing ADC values are shown in Table 1 . All MR images were acquired with a 1.5-T MR system (Vantage Orian) using a 16-channel phased-array surface coil (Atlas SPEEDER Body and Atlas SPEEDER Spine; Canon Medical Systems). First, T2-weighted imaging (T2WI) was conducted using compressed sensing (Compressed SPEEDER; Canon Medical Systems) in the axial, sagittal, and coronal planes, covering the entire pelvis. Second, axial DWI with and without RDC was obtained at the corresponding slice locations for axial T2WI, and both DWI data sets were reconstructed with the same DLR ( AiCE ). Details of the RDC technique have been reported previously. 12 , 13 Each ADC map was then calculated for a pair of b-values at 0 and 1000 s/mm 2 by means of mono-exponential fitting for each patient. The details of the scanning protocols are listed in Table 2 . In this study, RDC was technically applicable only to DWI provided by a single vendor. Therefore, T2WI could not be acquired using RDC. Anticholinergic agents were not routinely administered during the MRI examinations in this study. To determine the influence of RDC and DLR on SNR and ADC measurements, a board-certified radiologist with 13 years of experience (T.U.) performed all ROI measurements to determine the SNR with an ACR Large MRI phantom and ADC values with each phantom using a commercially available workstation (Vitrea ver 7.4; Canon Medical Systems). Circular ROIs 10 mm in diameter were placed at the center slice and at 2 adjacent slices 5 times per phantom. As previously reported, 12 , 22 the SNR depends on the acquired signal and background noise, as shown by the following equation: (1) SNR = mean SI / SD Noise where SI and SD Noise are the mean signal intensity within a given ROI and the standard deviation, respectively. In the in vivo study, all quantitative and qualitative assessments of the image quality were performed using a commercially available workstation (Vitrea ver 7.4). For the quantitative image quality assessment, a board-certified female pelvic radiologist (T.U.) with 13 years of experience performed ROI measurements. Circular ROIs 5–10 mm in diameter were placed over the normal myometrium, cervix, and ovary of each patient. When both normal ovaries could be identified, an ROI was placed on the larger normal ovarian parenchyma. Moreover, when it was possible to place an ROI on the normal ovarian parenchyma portion of an ovarian lesion, the ROI was placed on that normal ovarian parenchyma. For the quantitative image index, the SNR and contrast-to-noise ratio (CNR) were calculated using the following equations based on past literature 12 , 22 , 23 : (2) SNR=S I Myometrium /S D glutealmuscle (3) SN R Cervix =S I Cervix /S D glutealmuscle (4) SN R Ovary =S I Ovary /S D glutealmuscle (5) CNR= ( SI Endometrium − SI Myometrium ) / SD Myometrium where SI and SD are the mean signal intensity within a given ROI and the standard deviation, respectively. Moreover, regarding the quantitative image index, the ADC values of the myometrium, cervix, and ovary were also determined in each patient by ROI measurement. In addition, the cases where a circular ROI of 5 mm or larger could not be measured for SNR, CNR, and ADC value were excluded. To determine the difference in deformation between DWI with and without RDC, the margin of each uterine corpus, cervix, ovary, and lesion was traced as a freehand ROI on the same slice for each sequence, after which the area of each lesion was calculated. When both normal ovaries could be identified, an ROI was placed on the larger normal ovarian parenchyma. Moreover, when multiple lesions are present in the uterus, the ROI was placed on the largest lesion. The deformation ratio (DR) was then defined as the difference in the freehand ROI area between each DWI and T2WI divided by the ROI area on T2WI, expressed as the following equation based on previous literature 13 : (6) DR=  | Area T2WI -Area DWI | /Area T2WI where Area DWI and Area T2WI represent the area of the lesion determined via DWI and T2WI, respectively. In addition, cases where the size was too small for freehand ROI measurement were excluded from the DR analysis. For the qualitative assessment of the image quality, the same female pelvic radiologist (T.U.) and a board-certified radiologist with 10 years of experience (T.M.) independently and visually evaluated the overall image quality (OIQ), deformation severity (DS), and diagnostic confidence level (DCL) using a 5-point scoring system. Both readers were blinded to all information, including the acquisition technique, and all MR evaluations were performed at different times, on different days, and in different reading rooms. In this study, the OIQ was scored as follows: 1, poor; 2, fair; 3, moderate; 4, good; and 5, excellent. The DS was scored as follows: 1, no artifacts; 2, slight artifacts; 3, some artifacts; 4, obvious artifacts; and 5, marked artifacts. The DCL was scored as follows: 1, ≤ 20% confidence (i.e. very unsure); 2, 21%–40% confidence; 3, 41%–60% confidence; 4, 61%–80% confidence; and 5, ≥ 81% confidence (i.e., highly confident). The final visual scores for each patient were determined by consensus between the two readers. To determine the influence of RDC and DLR on the DWI scan quality, SNRs for the DWI modalities were compared using a paired t -test. The relationship between the ADC derived from each DWI modality and the standard reference was determined using the Spearman’s rank correlation coefficient. The mean ADC value for each phantom was then compared among the DWI modalities using the paired sample t -test. Furthermore, a Bland–Altman analysis was used to assess the limits of agreement for ADCs from all phantoms between each DWI modality and the standard reference. 24 To compare the quantitative image quality index between DWI modalities, the SNR, CNR, ADC value, and DR were compared between DWI modalities by the paired t -test. The correlation of ADC values between DWI modalities was evaluated using Spearman’s rank correlation coefficient. ADC measurements for DWI modalities were compared using a paired t -test. Inter-observer agreement for each qualitative index was evaluated using weighted kappa statistics and χ 2 tests. Inter-observer agreements were considered as poor for κ < 0.21, fair for κ = 0.21–0.40, moderate for κ = 0.41–0.60, substantial for κ = 0.61–0.80, and excellent for κ = 0.81–1.00. 25 , 26 The qualitative indices for both methods were compared using Wilcoxon’s signed-rank test. A P -value < 0.05 was considered significant for all statistical analyses. All statistical analyses were performed using a commercially available statistical software program (JMP 14.2; SAS Institute Japan, Tokyo, Japan). All statistical analyses were performed by the same female pelvic radiologist with 13 years of experience (T.U.).

The present study showed that DWI with RDC improved both the quantitative and qualitative image quality by reducing image deformation on female pelvic DWI using a 1.5-T MR system. Furthermore, DWI with RDC had no significant effect on ADC evaluations in in vitro or in vivo studies. To our knowledge, no studies have assessed the capability of RDC to improve the image quality and influence ADC measurement accuracy in the female pelvic field using a 1.5-T MR scanner in in vitro and in vivo studies. A comparison of the quantitative image quality for each phantom and correlation of ADC between each DWI and standard reference in the in vitro study showed no significant difference in the SNR between DWI with and without RDC. Furthermore, there was a significant or excellent correlation of ADC values between DWI and the standard reference. However, mean differences in ADC values of DWI with RDC were significantly smaller than those without RDC using phantoms with PVP contrast 40% and 50%, although there were no significant differences in ADC values between each DWI using phantoms with PVP contrast 10%–30%. These results suggest that DWI with RDC may have the potential to improve ADC assessments without any influence on the image quality in in vitro studies, findings compatible with those reported in the literature. 13 In addition to the findings in the in vitro study, there was no significant difference in the SNR of the uterine myometrium, cervix, or ovary between DWI with and without RDC in our in vivo study. Furthermore, there was no significant difference in the CNR between the endometrium and myometrium on DWI. There was also a significant correlation between ADC values in the myometrium, cervix, and ovary among DWI modalities. However, the deformation ratios of DWI with RDC were lower than those of DWI without RDC in the uterine corpus, cervix, ovary, and lesion. These findings suggest that DWI with RDC improves image distortion while having little adverse effect on ADC quantification, findings that are compatible with those reported in the literature. 13 The assessment of the interobserver agreement on all qualitative indices for each method on DWI modalities resulted in kappa values ranging from 0.78 to 0.97. Therefore, our interobserver agreements were determined to be ‘substantial’ or ‘almost perfect’ and considered reproducible. When the OIQ, DS, and DCL were compared between DWI with and without RDC, the OIQ and DCL of DWI with RDC were significantly higher than those without RDC, and the DS of DWI with RDC was significantly lower than that without RDC. These results suggest that the image quality and deformation severity were improved using RDC and that DWI with RDC can be more useful than DWI without RDC. These findings are consistent with our quantitative and qualitative assessments of the image quality and a previous report. 12 – 14 This indicates that RDC can improve the image quality without any influence on the ADC evaluation of female pelvic via DWI using a 1.5-T MR system. In routine female pelvic MRI, accurate anatomical evaluation on DWI is essential for reliable lesion localization, correlation with T2-weighted imaging, and confident interpretation—even before quantitative tumor assessment is considered. In the present study, diagnostic confidence was formally evaluated using a 5-point diagnostic confidence level scale assessed by 2 blinded radiologists. Our findings demonstrate that RDC significantly reduces image distortion and improves overall image quality as well as diagnostic confidence without affecting ADC values. These results support RDC as a robust technical solution in daily clinical practice. By preserving quantitative reproducibility while improving image distortion accuracy, RDC may enable more consistent image interpretation, reduce distortion-related uncertainty, and enhance workflow efficiency in routine female pelvic MRI examinations. Some limitations of the present study should be acknowledged. First, the number of subjects was relatively small, and their underlying pathological and clinical conditions varied. In addition, the numbers of malignant and benign lesions in the study population were limited. Second, all DWI data were acquired using a 1.5-T MR system with phased-array surface coils, and no testing was performed on a 3.0T MR system in either in vitro or in vivo settings. Third, RDC was technically applicable only to DWI provided by a single vendor in this study. Therefore, T2WI could not be acquired using RDC. Although combined assessment of T2WI and DWI is essential in female pelvic MRI, T2-weighted imaging served as a reference standard for deformation ratio measurements rather than as a target for RDC evaluation in this study. Fourth, the diagnostic performance of each method was not compared. Fifth, the effects of metal objects, such as surgical clips, were not evaluated. Sixth, the reproducibility of the results was not assessed on other MRI systems or with RDC techniques from different vendors. Therefore, multicenter and multivendor studies with larger prospective cohorts are warranted. Seventh, although DWI is most commonly used for tumor evaluation in clinical practice, this study included a limited number of malignant lesions and focused mainly on normal anatomical structures in female pelvic organ. This design was intentional to isolate the fundamental technical effects of RDC on image distortion and ADC stability, independent of tumor heterogeneity. Future studies with larger observational study enriched for cervical and endometrial malignancies, including lesion-based ADC analysis and diagnostic performance assessment, are warranted to further establish clinical applicability. Eighth, a direct comparison between RDC without anticholinergic agents and anticholinergic agents without RDC was beyond the scope of this study. However, such a comparison could further clarify whether RDC reduces the need for pharmacologic bowel motion suppression and should be pursued in future prospective investigations. Ninth, DLR was applied uniformly to each DWI data to minimize noise-related variability across acquisitions. As DLR primarily improves image denoising and does not correct phase encoding–related image distortion, the observed reductions in deformation and distortion are attributed to RDC rather than DLR. Potential synergistic effects between RDC and DLR were not specifically evaluated and warrant investigation in future studies. Tenth, the limited number of malignant lesions reflects both the nature of this retrospective study and the primary technical focus of the present investigation, which aimed to evaluate the fundamental effects of RDC on distortion and ADC stability in female pelvic MRI. Consequently, additional tumor-focused analyses could not be performed at this stage. In conclusion, RDC is useful for improving image quality by reducing image deformation without any influence on the ADC evaluation of female pelvic DWI using a 1.5-T MR system in in vitro and in vivo studies.

Diffusion-weighted imaging (DWI) has many potential clinical applications in the pelvis and can be easily added to any routine MR protocol. In the female pelvis, DWI obtained using single-shot echo-planar imaging (EPI) is the most frequently used method because of its rapid readout sequence and is useful for the diagnosis of female pelvic diseases or staging of uterine and ovarian tumors. 1 – 6 However, a major disadvantage of DWI is that it is prone to artifacts, particularly susceptibility artifacts at tissue interfaces, and image blurring. 7 , 8 Therefore, several approaches for DWI, such as parallel transmit EPI or readout-segmented multi-shot EPI, reduced FOV in the phase-encoding direction, and 2D navigator phase correction, have been tested to improve the image quality and reduce artifacts owing to various causes. 9 – 13 Under the above-mentioned situations, reverse encoding distortion correction (RDC) is clinically applied for DWI in the male pelvis as well as for deep learning reconstruction (DLR), and its utility was demonstrated in in vitro and in vivo studies in 2022. 12 Subsequently, the reverse encoding direction has been tested with different approaches and suggested to be useful for reducing distortion artifacts and improving the image quality on DWI in not only the male pelvis but also central nervous system and head and neck imaging. 12 – 17 However, no major reports have described the capability of RDC to improve the image quality and influence the apparent diffusion coefficient (ADC) measurement accuracy of DWI on a 1.5-T MR system in in vitro and in vivo studies. We hypothesized that female pelvic DWI with RDC can improve the image quality with little influence on ADC measurement and reduce susceptibility, distortion, or chemical shift artifacts in in vitro and in vivo studies compared with DWI without RDC. The present in vitro and in vivo study examined the influence of RDC on the quantitative and qualitative image quality and ADC evaluation of female pelvic DWI using a 1.5-T MR system.