The Utility of Apparent Water Diffusion Coefficient Maps for Evaluating the Presence of Myometrial Invasion in Patients with Endometrial Cancer.

A total of 164 women (mean age, 56 years; range, 25–89 years) were included in this study. Table 2 shows the patient characteristics, pathological types of EC, and FIGO (2023) stages. All patients underwent hysterectomy, and the tumor stage and histological type were determined pathologically. Based on the diagnosis of preoperative MRI, pelvic lymph node dissection was not performed in patients in whom MI was not suspected; only a biopsy was performed, and lymph node dissection was performed in all patients with suspected MI. The median time between the MRI and surgery was 44 days. Regarding pathologic MI, 73 patients had no MI, and five of these cases had aggressive histology; hence, according to histological considerations aligned with FIGO 2023, they were classified as having stage IC. The remaining 91 patients had <1/2 MI; 10 of these patients had aggressive histology, and according to histological considerations aligned with FIGO 2023, they were classified as having stage IIC. Among the cases with MI in this study, the median depth of MI was 1.0 mm and standard deviation was 2.2 mm (note that if the pathology report indicated an invasion of less than 1 mm, the depth of MI was set as 1 mm). Table 3 shows the sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV) and AUC for the presence of MI of EC for three readers in each sequence and combination. Table 4 shows the differences in the AUC values for each sequence and combination. The average sensitivity of the ADC maps for the three readers was 78%, which showed the highest sensitivity among all sequences and combinations. Here are the items that showed significant differences concerning ADC maps. Regarding sensitivity, for one reader, ADC maps were significantly more sensitive than T2WI ( P  < 0.001) and DCE-T1WI ( P  = 0.018). Moreover, for two readers, ADC maps were significantly more sensitive than CE-T1WI ( P  = 0.045 and 0.043). Regarding specificity, for two readers, ADC maps were less specific than T2WI ( P  = 0.015 and <0.001) and CE-T1WI ( P  = 0.031 and 0.01). Additionally, for one reader, ADC maps were less specific than DCE-T1WI ( P = 0.031). Regarding AUC, for one reader, AUC of ADC maps was significantly higher than that of T2WI ( P  = 0.048) and DCE-T1WI ( P  = 0.049). In the combination study, the combination of T2WI + DWI + ADC maps showed higher PPV than all single sequences or the combination of T2WI + DCE-T1WI + CE-T1WI in all three readers. In addition, T2WI + DWI + ADC maps showed lower NPV than T2WI + DCE-T1WI + CE-T1WI for all three readers. For one reader, T2WI + DWI + ADC maps had significantly lower specificity ( P  < 0.001) and AUC ( P  = 0.003) than T2WI + DCE-T1WI + CE-T1WI, but otherwise there were no significant differences in sensitivity, specificity, or AUC between these combinations for each reader. Table 5 shows the agreement of the three readers on the presence of MI. κ-value for ADC maps was 0.45, which was highest among all sequences and combinations. Figure 2 shows a case with MI while Figs. 3 – 5 depict cases without MI. In Fig. 2 , the boundary between the EMJ and the tumor disappeared on the ADC maps, indicating the possible presence of MI. In the cases shown in Figs. 3 and 4 , the boundary between the EMJ and tumor was preserved on the ADC maps, suggesting the absence of MI. Fig. 5 represents a case with adenomyosis, and identifying the EMJ on ADC maps or the JZ on T2WI is challenging because of tumor infiltration along the adenomyosis. Cancer infiltration along with adenomyosis is not pathologically classified as MI, making this a pitfall. There were 19 false negatives on the ADC maps. In 11 of these cases, the tumor filling the endometrial cavity had broad contact with the EMJ, rendering the boundary between the EMJ and tumor indistinct. Furthermore, in four cases, small tumors localized to narrow sites, such as the lower part of the uterus, uterine fundus, or uterine horn, made it impossible to distinguish between one EMJ and the opposite EMJ as well as between the EMJ and the tumor. In two cases, deformation of the endometrial cavity due to compression by uterine leiomyomas obscured the deep EMJ. In one patient, the EMJ thickened, which impeded proper evaluation. In one case, the EMJ became invisible owing to the traction of the stalk of the polypoid tumor.

This retrospective study was approved by the Institutional Review Board of University of Tsukuba Hospital, which waived the need for written informed consent (approval number: R05-238). The inclusion criteria were as follows: patients diagnosed with EC that was pathologically confined to the uterus or womb with <1/2 MI or no MI and who underwent MRI before surgery between April 2016 and July 2023. The exclusion criteria were as follows: a) patients whose lesions were too small to be evaluated by MRI, b) patients without the required sequence, c) patients who were not imaged on the designated MRI machine, and d) patients whose MRI scans were of poor quality due to artifacts. A flowchart of the patient selection process is shown in Fig. 1 . We utilized the updated 2023 FIGO staging system to assess and classify patients with EC. 1 MRI was performed using 3T equipment (Ingenia; Philips Medical Systems, the Netherlands). T2WI, DWI with b-values of 0 and 1000, ADC maps, DCE-T1WI, and CE-T1WI sequences were evaluated. The evaluated images were acquired in oblique axial directions perpendicular to the long axis of the uterine body. Gadopentetate dimeglumine (Gadovist 1.0 metre; Bayer, Wuppertal, Germany) was used at a concentration of 5 mmol for CE-T1WI. The bolus intravenous contrast injection rate was 4 mL/s (2 mmol/s) diluted with saline, and four-phase images were acquired every 30s, starting 15s after contrast injection. Hyoscine butylbromide (20 mg; Buscopan; Sanofi, France) was injected intramuscularly to all patients immediately before the examination, unless contraindicated, to reduce motion artifacts caused by bowel peristalsis. Further details of these parameters are listed in Table 1 . Three expert radiologists (T. S., M. Y., and S. S.) with 21, 10, and 9 years of experience in abdominal radiology independently reviewed the MRI scans. For each patient, the reviewer was provided with an individual series of T2WI, DWI, ADC maps, DCE-T1WI, and CE-T1WI sequences exclusively to assess the presence of MI. They were blinded to the pathological and clinical findings, and there were two or more weeks between the interpretations of each sequence from the same patient. The criteria for determining the presence of MI in each image sequence were as follows: T2WI, irregular tumor-myometrial margins and unclear JZ; DWI, irregular tumor-myometrial margins; ADC maps, disappeared boundary between the EMJ and tumor; DCE-T1WI, disrupted subendometrial enhancement (SEE, a thin-layered enhancement between the endometrium and myometrium) continuity or irregular tumor-myometrium margin; and CE-T1WI, irregular tumor-myometrial margin. We also evaluated MI using the combination of T2WI, DWI, and ADC maps, as well as T2WI, DCE-T1WI, and CE-T1WI. An interval of at least two weeks was allowed between readings of each of the aforementioned sequences and combinations and between readings and combinations. Based on the results of the image evaluation for each sequence and combination, the sensitivity, specificity, and accuracy for diagnosing MI were calculated. Receiver operating characteristic (ROC) curve analysis was performed to assess the diagnostic performance. The diagnostic value was interpreted based on area under the ROC curve (AUC) as follows: 0–0.70, poor; 0.70–0.90, moderate; and 0.90–1.00, high. We estimated 95% confidence intervals (CIs), and the sensitivities and specificities for each sequence and combination were compared using McNemar’s test with exact binomials. The AUCs of all sequences were also compared with each other. The agreement of the three readers was assessed using Fleiss’ kappa (κ) statistics. The κ-statistic interpreted the agreement as follows: <0, no; 0–0.20, slight; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, substantial; and 0.81–1.00, almost perfect. All statistical analyses were performed using the SPSS software (SPSS Statistics 28.0; IBM, New York, NY, USA). Statistical significance was set at P <0.05.

The disappearance of the boundary between the EMJ and tumor on the ADC maps was a sensitive indicator diagnosing MI, with a higher inter-reader agreement rate. Although ADC maps have not been used for diagnosing MI, their use is effective for evaluating the presence of MI and plays an important role in integrated local evaluation, including other sequences.

In the present study, to determine the utility of the disappearance of the EMJ and tumor boundary on ADC maps for diagnosing MI, various sequences were independently and in combination compared. ADC maps showed the highest sensitivity, based on the average of three readers, among all sequences and combinations. This indicated that within the sequences assessed, the ADC maps had the lowest likelihood of missing an MI, making them useful for detecting MI. The low false-negative rate contributes to decisions regarding intraoperative lymph node dissection and determining the indication for fertility preservation. Additionally, the high inter-reader agreement rate for the ADC maps indicates the ease of evaluation. Furthermore, the combination of T2WI, DWI, and ADC maps demonstrated the highest PPV. According to the guidelines from various countries, 2 , 8 , 9 MRI is recommended for the initial staging of EC. Specifically, CE-T1WI is valuable for the assessment of MI and is recommended by various guidelines, 8 , 15 and conventionally, a combination of T2WI and DCE-T1WI is accepted as the best approach for the local staging of EC. 16 , 17 In the current comparison of combinations, the inferiority of T2WI + DWI + ADC maps to T2WI + DCE-T1WI + CE-T1WI in diagnostic performance was not observed in two out of three readers, and T2WI + DWI + ADC maps showed the highest PPV. Considering the disadvantages associated with contrast agents, such as the risk of allergies, longer examination times, and higher costs, the benefits of using contrast agents may be negligible. However, since this study has certain limitations, a more comprehensive evaluation is needed. JZ represents the innermost layer of the myometrium. On ADC maps, the EMJ is composed of a low-signal inner myometrium. If the high signal of the normal endometrium is not preserved between the EMJ and tumor, or if the signal difference between the EMJ and tumor is not clear and the two cannot be separated, there is a possibility of MI. It is a clear and easily accessible indicator. To the best of our knowledge, this is the first study to assess MI visually using ADC maps. In addition, no study has independently evaluated and characterized the different types of sequences that are considered useful for MI diagnosis. We recommend that if the EMJ and tumor boundary are not preserved on ADC maps, indicating the possible MI, then, furthermore, other sequences such as DWI or CE-T1WI should be used to evaluate the irregularity between the tumor and the myometrium. Simultaneously, T2WI should be used to assess the tumor and any coexisting conditions, such as leiomyomas or adenomyosis. In this study, with regard to cases with well-defined SEE, the innermost layer of SEE matched the EMJ on the ADC maps in all cases without MI. Even in cases with MI, there was consistency in areas without MI. Therefore, the EMJ indicates the innermost layer of the SEE and may reflect the effects of blood flow. Several points of caution should be exercised while evaluating this structure. In cases where adenomyosis is present, distinguishing between a normal endometrium and an overgrowth of endometrial glands in the myometrium is challenging and may result in poor EMJ delineation sometimes. 18 In addition, when EC coexist with endometrial hyperplasia, 19 the endometrium shows a heterogeneous signal throughout the ADC maps, making it difficult to distinguish it from the EMJ. Furthermore, in cases of uterine leiomyomas, distorted uterine structure, or extension of the tumor to the fundal side, assessing the relationship between the tumor and EMJs in the vertical section is not possible, thus reducing the diagnostic performance. If the tumor is in a region such as the cervix or uterine horn, separating the boundaries is difficult because of the narrowing between the EMJ. In such cases, other cross sections or combinations of other sequences are necessary. In the current study, the AUC of each sequence and combination was not high compared to that in previous studies. 20 , 21 This may be because previous reports have focused on the depth of MI, particularly whether it is deeper than 1/2, and have not evaluated the presence or absence of MI. The median MI depth in this study was very low. Because this very small amount of MI was diagnosed with a 3 mm slice MRI, it is not surprising that the AUC was not as high as that in previous reports involving lesions with >1/2 MI. Moreover, the fact that only oblique axial sections were evaluated to ensure the same conditions for each sequence may have contributed to the low AUC. This study had several limitations. First, the retrospective nature of this study may have caused a selection bias. Secondly, only oblique axial images were evaluated to compare sequences under the same conditions. Third, axial oblique images acquired perpendicular to the endometrial cavity are necessary to evaluate MI, but not all slices were necessarily perpendicular because of endometrial cavity deformation caused by leiomyoma or uterine adenomyosis. Fourth, this study focused only on the presence of MI and did not assess the degree of invasion. Fifth, the interval between MRI and surgery was long, with a median of 55 days and a maximum of 93 days. Finally, MRI was performed on a single 3T equipment. Since the image quality of ADC maps can vary depending on machine quality and magnet strength, further examination will be necessary to obtain more robust results for this study.

Endometrial cancer (EC) is the most common gynecological malignancy, and the International Federation of Gynecology and Obstetrics (FIGO) staging system is the most widely accepted method of EC staging. In 2023, a new FIGO classification has been proposed. 1 According to the 2009 system, Stage I was categorized solely on the basis of myometrial invasion (MI) depth. The current revision, however, includes the presence or absence of MI. 1 , 2 The presence of MI strongly correlates with outcomes, and EC without MI has a good prognosis as lymph node metastasis is less likely to occur. 3 – 7 The presence of MI affects treatment planning and surgical techniques, and in the absence of MI, a less invasive pelvic or para-aortic lymph node sampling can be selected instead of the more invasive radical lymph node dissection. 8 , 9 In addition, the absence of MI is a critical factor for preserving fertility. Therefore, accurately assessing the presence or absence of MI is essential, in women with EC and specifically those who wish to carry children. 10 MRI is the gold standard imaging technique for the preoperative assessment of EC because of its excellent soft-tissue contrast resolution. 11 – 13 Although apparent diffusion coefficient (ADC) maps are generally used to diagnose the presence or absence of EC but not to assess the extent of MI. Kido et al. reported that the endometrial-myometrial junction (EMJ) exhibits a thin layer of restricted water diffusion on ADC maps, similar to the junctional zone (JZ) observed on T2-weighted imaging (T2WI), 14 and in our hypothesis, the preservation of EMJ and tumor boundary on ADC maps indicates the absence of MI. In addition, for staging of EC, multiple sequences, including T2WI, diffusion-weighted imaging (DWI), dynamic contrast-enhanced T1-weighted imaging (DCE-T1WI), and contrast-enhanced T1-weighted imaging (CE-T1WI), are synthesized; however, it is unclear which sequence is more sensitive and specific or which sequence should be considered more important. Hence, in this study, we aimed to assess the utility of the lack of an EMJ and tumor boundary on ADC maps in diagnosing MI in EC by independently evaluating each of these sequences.