Apparent diffusion coefficient values on MRI as biomarkers for pre operative prediction of prognostic factors in endometrial cancer

Ashika Padival: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Bijal Manish Patel: Writing – review & editing, Validation, Supervision, Software, Resources, Investigation, Data curation, Conceptualization. Chetana D. Parekh: Writing – review & editing, Visualization, Investigation, Formal analysis, Conceptualization. Ruchi Arora: Writing – review & editing, Visualization, Investigation, Formal analysis, Conceptualization. Hitesh Rajpura: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Kedar Padhiyar: Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Methodology, Investigation, Formal analysis, Conceptualization.

Informed consent and patient details: Written informed consent to take part in the study and to publish the article has been obtained from all participants or their legal representatives. The privacy rights of participants have been observed. Studies in Human: This study was performed in compliance with relevant laws, regulatory frameworks and guidelines where the research took place. This study was approved by the Gujarat Cancer Research Institute. (Approval No. P-34/2026) Ethics & informed consent statement : This study was conducted in accordance with the Declaration of Helsinki. Ethics approval was obtained from the Institutional Review Board (IRB) of Gujarat Cancer Research Institute (GCRI), Ahmedabad, India IRB REFERENCE Number: IRC/2026/P-34; Date of Approval: 15/4/2026. Clinical trial registration: Not Applicable. This is a prospective observational study and does not qualify as a prospective clinical trial requiring registration.

A total of 51 patients with histologically proven endometrial carcinoma who underwent preoperative MRI were included in the study. The mean age of the study population was 60 ± 9.83 years (range: 32–77 years). Endometrioid adenocarcinoma was the most common histological subtype and was seen in 47 (92.2%) patients. Serous carcinoma and carcinosarcoma were each seen in 2 (3.9%) patients. Most patients had grade 2 disease, accounting for 23 (45.1%) cases. p53 mutant tumors were identified in 10 (19.6%) patients, while ER/PR positivity was seen in 28 (54.9%) patients. MMR deficiency was seen in 13 (25.5%) patients. Based on conventional MRI findings, >50% myometrial invasion was seen in 21 (41.2%) patients, while serosal involvement was noted in 1 (2%) patient. Cervical involvement was present in 13 (25.5%). Vaginal and parametrial involvement were seen in 4 (7.8%) and 2 (3.9%), respectively. Enlarged pelvic lymph nodes were noted in 9 (17.6%) patients. Fig. 1 demonstrates an example of an endometrial lesion with an ADC value calculation. Fig. 1 Representative preoperative pelvic MRI of endometrial carcinoma. (a) Axial oblique diffusion-weighted image demonstrating restricted diffusion within the endometrial tumor (arrow). The endometrial cavity is indicated by an arrowhead. (b) Corresponding apparent diffusion coefficient (ADC) map showing low signal intensity within the tumor. A focused region of interest was placed over the most diffusion-restricted solid tumor component, yielding a mean ADC value of 747.13 × 10⁻ 6  mm 2 /s. Fig. 1 Representative preoperative pelvic MRI of endometrial carcinoma. (a) Axial oblique diffusion-weighted image demonstrating restricted diffusion within the endometrial tumor (arrow). The endometrial cavity is indicated by an arrowhead. (b) Corresponding apparent diffusion coefficient (ADC) map showing low signal intensity within the tumor. A focused region of interest was placed over the most diffusion-restricted solid tumor component, yielding a mean ADC value of 747.13 × 10⁻ 6  mm 2 /s. On final histopathology findings, >50% myometrial invasion was observed in 24 (47.1%) patients, and serosal involvement was identified in 3 (5.9%) patients. Substantial LVSI was noted in 7(13.73%). Cervical involvement was noted in 12 (23.5%) patients, and parametrial involvement was noted in 1 (2%) patient. Adnexal involvement was seen in 5(9.8%).Pelvic lymph node involvement was seen in 6(11.8%).The clinical and demographic characteristics are depicted in Table 1 . Table 1 Clinicopathological characteristics of the study population ( N  = 51). Characteristic Category n (%) Histological subtype Endometrioid adenocarcinoma 47 (92.2) Serous carcinoma 2 (3.9) Carcinosarcoma 2 (3.9) Tumor grade Grade 1 13 (25.5) Grade 2 23 (45.1) Grade 3 15 (29.4) Myometrial invasion Endometrium only 5 (9.8) 50% invasion 24 (47.1) Serosal involvement 3 (5.9) Cervical involvement Yes 13 (25.5) No 38 (74.5) Vaginal involvement Yes 4 (7.8) No 47 (92.2) Parametrial involvement Yes 2 (3.9) No 49 (96.1) Pelvic lymph node involvement Yes 9 (17.6) No 42 (82.4) Adnexal involvement Yes 6 (11.8) No 45 (88.2) LVSI Absent 37 (72.5) Focal 7 (13.7) Substantial 7 (13.7) Final FIGO 2023 stage I 24 (47.0) 1a1 5 (20.8) 1a2 7 (29.2) 1a3 2 (8.3) 1b 10 (41.7) II 16 (31.4) 2a 3 (18.8) 2b 1 (6.3) 2c 12 (75.0) III 11 (21.6) 3a1 1 (9.1) 3a2 1 (9.1) 3b(ii) 3 (27.3) 3c1 1 (9.1) 3c1(ii) 4 (36.4) 3c2(ii) 1 (9.1) FIGO 2009 stage I 24 (47.0) 1a 14 (58.3) 1b 10 (41.7) II 16 (31.4) III 11 (21.6) 3a 2 (18.2) 3b 3 (27.3) 3c 6 (54.5) MMR Status Deficient 13 (25.5) Proficient 38 (74.5) p53 Status Mutant 11 (21.6) Wild 40 (78.4) Note. FIGO, International Federation of Gynaecology and Obstetrics; LVSI, lymphovascular space invasion. Clinicopathological characteristics of the study population ( N  = 51). Note. FIGO, International Federation of Gynaecology and Obstetrics; LVSI, lymphovascular space invasion. The comparison between MRI stage and final histopathology stage is illustrated in the heat map analysis in Fig. 2 . Maximum concordance was noted for stages IB and IIC. Overall concordance between MRI staging and final histopathological staging showed a weighted kappa of 0.8666 indicated almost perfect agreement, demonstrating high accuracy of MRI in detecting >50% myometrial invasion. Fig. 2 Agreement heatmap showing concordance between MRI-based stage and final histopathological International Federation of Gynaecology and Obstetrics (FIGO) 2023 stage. Darker shades and higher cell counts indicate greater agreement. The highest concordance was observed for stages IB and IIC. Fig. 2 Agreement heatmap showing concordance between MRI-based stage and final histopathological International Federation of Gynaecology and Obstetrics (FIGO) 2023 stage. Darker shades and higher cell counts indicate greater agreement. The highest concordance was observed for stages IB and IIC. With the addition of functional MRI sequences such as DCE and DWI, 25 (49%) were seen to have >50% myometrial invasion, showing greater concordance with final histopathology findings. On DCE imaging, most tumors demonstrated heterogeneous enhancement in 21 (41.2%), and 24 (47.1%) showed mild post-contrast enhancement. The mean region of interest (ROI) size was 6.67 ± 4.48 cm 2 (range: 1.17–21.8 cm 2 ). Interobserver variability between the two radiologists was assessed using paired t -test and Bland–Altman analysis. Focused-ROI ADC measurements showed no statistically significant interobserver difference ( p  = 0.41), suggesting greater interobserver consistency. In contrast, whole-lesion ADC measurements demonstrated a statistically significant interobserver difference ( p  = 0.029), suggesting greater variability with whole-lesion segmentation. The Bland–Altman analysis comparing focused and whole-lesion ADC measurements between the two radiologists is shown in Fig. 3 . ANOVA analysis showed a statistically significant association between focused ADC values and tumor grade, histological subtype, and depth of myometrial invasion. Mean ADC values for grade 1, grade 2, and grade 3 tumors were 1001 ± 144.7 × 10 −6  mm 2 /s, 816 ± 131.9 × 10 −6  mm 2 /s and 683 ± 40.9 × 10 −6  mm 2 /s, respectively ( p  < 0.001). Fig. 3 Bland-Altman plots comparing ADC measurements between the two radiologists for (a) whole-lesion ADC and (b) focused-region ADC. Whole-lesion ADC showed significant interobserver variation, whereas focused-region ADC showed no significant interobserver difference. ADC values are expressed as x 10 −6  mm 2 /s. Fig. 3 Bland-Altman plots comparing ADC measurements between the two radiologists for (a) whole-lesion ADC and (b) focused-region ADC. Whole-lesion ADC showed significant interobserver variation, whereas focused-region ADC showed no significant interobserver difference. ADC values are expressed as x 10 −6  mm 2 /s. Among histological subtypes, serous tumors showed the lowest mean ADC value of 642 ± 39.58 × 10–6 mm 2 /s, compared to carcinosarcoma and endometrioid tumors, which showed mean ADC values of 689 ± 43.54 × 10–6 mm 2 /s and 837 ± 164.74 × 10–6 mm 2 /s, respectively ( p  = 0.046). However in view of very few patients in these aggressive histology groups, these results are not clinically significant. For depth of myometrial invasion, mean ADC values were 1136 ± 29.5 × 10–6 mm 2 /s for lesions confined to the endometrium, 830 ± 142.6 × 10–6 mm 2 /s for 50% myometrial invasion, and 687 ± 84.2 × 10–6 mm 2 /s for serosal involvement ( p  < 0.001). ROC analysis of focused ADC values showed a significant association with tumor grade, LVSI, pelvic lymph node involvement, and cervical involvement. The ROC curves along with AUC values, sensitivity, specificity, positive predictive value, negative predictive value, and ADC cutoff values are presented in Fig. 4 . Fig. 4 Receiver operating characteristic (ROC) curves showing the diagnostic performance of focused apparent diffusion coefficient (ADC) values for molecular markers: (a) prediction of tumor grade (b) prediction of lymphovascular space invasion (LVSI) (c) prediction of cervical involvement (d) prediction of pelvic lymph node involvement. Fig. 4 Receiver operating characteristic (ROC) curves showing the diagnostic performance of focused apparent diffusion coefficient (ADC) values for molecular markers: (a) prediction of tumor grade (b) prediction of lymphovascular space invasion (LVSI) (c) prediction of cervical involvement (d) prediction of pelvic lymph node involvement. The study also evaluated the correlation between ADC values and molecular classification of endometrial tumors through ROC curves. For p53 mutant tumors, the optimal ADC cutoff value was 728.71 × 10 −6  mm 2 /s with the highest sensitivity and specificity of 80% and 90.9%, respectively. ER/PR-positive tumors showed an ADC cutoff value of 779.7 × 10 −6  mm 2 /s with a sensitivity of 84.6% and a specificity of 68.4%. MMR-deficient tumors demonstrated an ADC cutoff value of 758.38 × 10 −6  mm 2 /s with a sensitivity of 52.8% and a specificity of 84.6%. To the best of our knowledge, this is among the first studies evaluating the association between ADC values and molecular classification in endometrial carcinoma. Fig. 5 shows ROC curves for tumors according to molecular classification. Fig. 5 Receiver operating characteristic (ROC) curves showing the diagnostic performance of focused apparent diffusion coefficient (ADC) values for molecular markers: (a) p53-mutant tumors and (b) estrogen/progesterone receptor (ER/PR)-positive tumors (c) MMR deficient tumors. ADC values are expressed as × 10 −6  mm 2 /s. Fig. 5 Receiver operating characteristic (ROC) curves showing the diagnostic performance of focused apparent diffusion coefficient (ADC) values for molecular markers: (a) p53-mutant tumors and (b) estrogen/progesterone receptor (ER/PR)-positive tumors (c) MMR deficient tumors. ADC values are expressed as × 10 −6  mm 2 /s.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

In brief, our study suggests that focused-ROI ADC assessment derived from DWI MRI is a standardized and reproducible quantitative imaging biomarker with potential clinical relevance for preoperative risk stratification. Because ADC measurements can be obtained from routinely acquired DWI sequences without requiring a separate imaging examination, additional contrast administration, or specialized post-processing, this approach may be particularly practical in resource-constrained settings where MRI is already available but access to advanced molecular testing is limited. However, MRI availability and cost remain important considerations, and the present study did not formally evaluate cost-effectiveness. Further prospective multicenter and health-economic studies are required to determine the clinical and economic value of incorporating ADC assessment into routine preoperative evaluation.

Endometrial carcinoma typically demonstrates diffusion restriction on DWI, appearing hyperintense relative to the surrounding myometrium and hypointense on ADC maps ( Nougaret et al., 2019 ; Moreira et al., 2023 ). As opposed to conventional T2-weighted imaging, DWI often provides clearer visualization of the tumor and its interface with the normal myometrium ( Nougaret et al., 2019 ; Beddy et al., 2012 ). Previous studies have reported both sensitivity and specificity of nearly 86% for the detection of deep myometrial invasion using DWI ( Beddy et al., 2012 ; Das et al., 2014 ). In the present study, ADC values derived from DWI MRI showed significant correlation with important prognostic factors in endometrial carcinoma, including tumor grade, histological subtype, and depth of myometrial invasion ( Inoue et al., 2015 ; Moreira et al., 2023 ; Saleh et al., 2024 ). These conclusions are in agreement with the growing literature supporting the role of DWI MRI as a quantitative imaging biomarker in endometrial carcinoma ( Inoue et al., 2015 ; Moreira et al., 2023 ; Saleh et al., 2024 ). Tamai et al. (2007), in a study involving 18 patients, observed that normal endometrium demonstrates higher ADC values compared to endometrial carcinoma ( Tamai et al., 2007 ). They also reported a progressive decline in ADC values with increasing tumor grade ( Tamai et al., 2007 ). Low-grade endometrial carcinomas may sometimes contain “normal-appearing” glands resembling benign endometrium and may coexist with endometrial hyperplasia ( Tamai et al., 2007 ). These tumors generally show lower cellularity with relatively increased water diffusion, resulting in higher ADC values ( Tamai et al., 2007 ). In contrast, high-grade tumors are characterized by increased cellular density and therefore tend to exhibit lower ADC values ( Tamai et al., 2007 ). Several studies have similarly demonstrated lower ADC values in high-grade endometrioid as well as non-endometrioid tumors, with reported mean ADC differences ranging between 100 and 200 × 10 −6  mm 2 /s when comparing low-grade and high-grade lesions ( Rechichi et al., 2011 ; Reyes-Pérez et al., 2020 ; Husby et al., 2015 ; Woo et al., 2014 ). In our study, ADC values demonstrated a significant inverse association with tumor grade, with clear separation on ANOVA analysis between grade 1 tumors (1001 ± 144.7 × 10 −6  mm 2 /s) and grade 3 tumors (683 ± 40.9 × 10 −6  mm 2 /s), indicating progressively lower ADC values with increasing tumor grade. However, notwithstanding these observations, the role of ADC values in differentiating tumor grades remains somewhat controversial because considerable overlap can still occur between low-grade and high-grade tumors ( Rechichi et al., 2011 ; Woo et al., 2014 ). Tumor grading is influenced by cellularity, nuclear atypia, and molecular pathological features, which cannot be fully assessed using ADC values alone ( Rechichi et al., 2011 ; Woo et al., 2014 ). Prior meta-analyses and multicenter studies report MRI accuracy rates of 75–85% for deep myometrial invasion, with kappa values from 0.6 to 0.8 ( Das et al., 2014 ; Andreano et al., 2014 ). In our cohort, MRI staging demonstrated excellent agreement with final histopathology, with a weighted kappa value of 0.866. This finding further supports the reliability of MRI as a preoperative staging modality and may also reflect the use of standardized imaging procedures at our center. Reyes-Perez et al. ( Reyes-Pérez et al., 2020 ) also evaluated the diagnostic performance of ADC values in predicting tumor grade and myometrial invasion ( Nougaret et al., 2015 ; Bonatti et al., 2018 ). Their study showed that both ADC mean and ADC minimum values were significantly lower in high-grade tumors and in cases with deep myometrial invasion, suggesting that diffusion-weighted imaging may reflect tumor cellularity and ultrastructural alterations associated with aggressive disease ( Reyes-Pérez et al., 2020 ; Nougaret et al., 2015 ; Bonatti et al., 2018 ; Ma et al., 2021 ). These data are comparable to the observations in our study, where ADC values considerably differentiated tumor grade and depth of invasion ( Inoue et al., 2015 ; Reyes-Pérez et al., 2020 ; Husby et al., 2015 ; Nougaret et al., 2015 ; Ma et al., 2021 ). Although FIGO 2023 incorporates molecular classification into contemporary risk stratification, conventional pathological factors such as tumor grade, histological subtype, myometrial invasion, LVSI, cervical involvement, and nodal status remain important determinants of treatment planning and are key targets of preoperative MRI assessment. Therefore, evaluating ADC values in relation to these established prognostic factors remains clinically relevant in the FIGO 2023 era. Molecular analyses involving p53 and MMR status were included as exploratory secondary endpoints, while POLE status was unavailable in our cohort. A recent prospective study by Saleh et al. (2024) ( Saleh et al., 2024 ) evaluated 68 patients with endometrial cancer and reported significantly lower ADC values in tumors associated with deep myometrial invasion, cervical stromal invasion, lymphovascular space invasion, and non-endometrioid histology ( Nougaret et al., 2015 ; Bonatti et al., 2018 ; Gil et al., 2019 ). They identified an optimal cut-off value of <0.84 × 10 −3  mm 2 /s for predicting deep myometrial invasion, and also presented a strong negative correlation between ADC values, tumor grade, and FIGO stage ( Saleh et al., 2024 ; Andreano et al., 2014 ; Nougaret et al., 2015 ; Gil et al., 2019 ). Similarly, Nougaret et al. and Wang et al. demonstrated a strong inverse correlation between ADC values and both myometrial invasion and lymphovascular invasion, although they did not establish a definitive cutoff value for these parameters ( Nougaret et al., 2015 ; Wang et al., 2022 ). Husby et al. also reported significantly lower mean ADC values in tumors with deep myometrial invasion (0.75 × 10 −3  mm 2 /s) compared with superficial invasion (0.85 × 10 −3  mm 2 /s), which is consistent with our findings ( Husby et al., 2015 ; Nougaret et al., 2015 ; Deng et al., 2015 ). However, studies by Rechichi et al. and Lin et al. did not observe statistically significant differences in ADC values between superficial and deep myometrial invasion, pointing to the variability that still exists in the literature ( Rechichi et al., 2011 ; Lin et al., 2009 ; Ozturk et al., 2021 ). Ozturk et al. ( Ozturk et al., 2021 ), in a study involving 83 patients, reported that ADC measurements were useful in differentiating histological subtypes of endometrial carcinoma, with significantly lower ADC values in non-endometrioid tumors compared with endometrioid histology ( Keles et al., 2022 ; Wang et al., 2022 ). These results were also observed in our study, however the number of patients in the non endometrioid subgroup are too small to draw any conclusions. Hence these results are hypothesis generating/exploratory but not definitive conclusions and larger studies are needed to address this cohort. The study by Ozturk et al. did not demonstrate significant correlation between ADC values and factors such as lymphovascular invasion or lymph node metastasis, unlike the findings in our study ( Ozturk et al., 2021 ). Such differences may be related as a result of variations in study design, MRI protocols, sample size, and ROI methodology ( Nougaret et al., 2015 ; Ma et al., 2021 ; Lin et al., 2009 ; Ozturk et al., 2021 ; Deng et al., 2015 ; Keles et al., 2022 ; Wang et al., 2022 ). In the present study, ADC quantitative variables were obtained using two different ROI approaches. Whole-lesion ROIs were used to account for intratumoral heterogeneity, whereas focused ROIs were placed over the areas demonstrating maximum diffusion restriction to better reflect tumor aggressiveness ( Nougaret et al., 2015 ; Ma et al., 2021 ; Padhani et al., 2009 ). Inter-reader reproducibility continues as a recognized challenge in ADC analysis and was an important aspect evaluated in our study ( Padhani et al., 2009 ; Koh et al., 2012 ; Xie et al., 2020 ). Several previous studies have suggested that focused or small ROIs placed in the most diffusion-restricted area provide better reproducibility and stronger association with tumor grade, whereas whole-lesion ROIs may better represent tumor heterogeneity but often reduce reproducibility unless highly standardized protocols are followed. ( Woo et al., 2014 ; Nougaret et al., 2015 ; Ma et al., 2021 ; Keles et al., 2022 ; Padhani et al., 2009 ) Measurement of ADC values is also technically challenging because ROI placement is performed on ADC maps that inherently have a relatively low signal-to-noise ratio ( Padhani et al., 2009 ; Koh et al., 2012 ; Xie et al., 2020 ). Accurate ROI delineation can be difficult in cases with indistinct tumor margins, small lesions, or tumors arising within an atrophic uterus ( Nougaret et al., 2019 ; Koh et al., 2012 ; Xie et al., 2020 ; Scepanovic et al., 2023 ). Uterine position may additionally influence lesion visualization and ROI placement ( Nougaret et al., 2019 ; Xie et al., 2020 ). These practical limitations are reflected in our findings, where focused ADC measurements showed no significant inter-reader difference, while whole-lesion ADC measurements presented statistically significant variation between observers ( Padhani et al., 2009 ; Koh et al., 2012 ; Xie et al., 2020 ; Jiang et al., 2018 ; Scepanovic et al., 2023 ). These findings suggest that reproducibility is strongly influenced by the method of ROI selection. The focused-ROI approach may offer greater consistency because it targets a visually identifiable area of maximal diffusion restriction rather than requiring complete delineation of the tumor margins. In contrast, whole-lesion segmentation is more susceptible to differences in tumor boundary definition and inclusion of heterogeneous tumor components. Nevertheless, the reproducibility demonstrated in our single-center cohort should not be interpreted as evidence of universal reproducibility. Differences in MRI field strength, acquisition parameters, scanner vendors, ADC map generation, tumor size and morphology, and radiologist experience may affect ADC measurements across institutions. Therefore, standardized acquisition and ROI-placement protocols and prospective multicenter validation are necessary before focused-ROI ADC measurements can be routinely adopted as a quantitative biomarker. While several larger studies and meta-analyses have established associations between ADC values and prognostic factors such as tumor grade, myometrial invasion, and histological subtype, our study provides additional insights in several areas ( Moreira et al., 2023 ; Saleh et al., 2024 ; Rechichi et al., 2011 ; Reyes-Pérez et al., 2020 ; Husby et al., 2015 ; Woo et al., 2014 ; Andreano et al., 2014 ; Nougaret et al., 2015 ; Bonatti et al., 2018 ; Ma et al., 2021 ). First, we directly compared whole-lesion and focused-ROI ADC measurement techniques and demonstrated superior interobserver reproducibility with the focused-ROI approach. Second, unlike many earlier studies, we incorporated an assessment of measurement reproducibility as a primary methodological objective, highlighting the practical challenges of ADC implementation in routine clinical practice ( Padhani et al., 2009 ; Koh et al., 2012 ; Xie et al., 2020 ). Third, our study evaluated ADC performance within the contemporary FIGO 2023 framework and included exploratory analyses of molecular markers, including p53 and MMR status ( Berek et al., 2023 ). Although the molecular analyses require validation in larger cohorts, our findings suggest that a simple focused-ROI approach may represent a clinically feasible and reproducible imaging biomarker for preoperative risk stratification, particularly in settings where advanced molecular testing is not universally available. Accurate preoperative staging is critical when tailoring surgical management. Differentiating superficial from deep myometrial invasion is critical since it is associated with high risk for LVSI, which relates directly to tumor grade, nodal metastases, and recurrence. Our findings further strengthen the role of MRI as a necessary tool in guiding surgical decisions, particularly in selecting patients for minimally invasive approaches and fertility-sparing therapies. Although MRI itself may not be readily available in all low-resource settings, particularly where access to advanced imaging infrastructure is limited, the potential utility of ADC assessment in resource-constrained settings relates to its integration into MRI examinations that are already being performed for routine preoperative staging. In our study, ADC measurements were derived from routinely acquired diffusion-weighted MRI sequences and did not require an additional imaging examination, additional contrast administration, or specialized post-processing techniques. Thus, in centers where MRI is already available, focused-ROI ADC assessment may provide additional quantitative information without substantially increasing imaging requirements or workflow complexity. This may be particularly relevant in settings where access to molecular testing is limited. However, we acknowledge that MRI availability and associated costs remain important barriers in low-resource environments, and our study did not perform a formal health-economic evaluation. Therefore, the cost-effectiveness and broader applicability of ADC-based assessment in resource-limited settings require confirmation in future prospective health-economic studies. The strengths of this study include a relatively homogeneous patient cohort, the use of a standardized MRI protocol, and prior assessment of the proficiency of both radiologists. The limitations of our study include its single-center design, which may limit generalizability, and the relatively small sample size compared to multicenter trials. Future prospective multicenter studies with expanded cohorts are required to validate the diagnostic thresholds identified in our study. Integration of multiparametric MRI, including diffusion and perfusion imaging, with radiomics and artificial intelligence may further enhance staging accuracy and prognostication in endometrial carcinoma.

Not Applicable: No generative AI tools were used in the preparation of this manuscript.

This study prospectively evaluated patients with endometrial carcinoma who underwent preoperative pelvic MRI at Gujarat Cancer Research Institute between 1 June 2023 and 1 June 2025. Patients were included if histopathological confirmation of endometrial carcinoma, complete preoperative MRI examinations including DWI, and surgical histopathology reports were available. Patients who had received treatment such as pelvic radiotherapy, chemotherapy, or hormonal therapy were excluded. Cases with impaired renal function, contraindications to MRI, imaging performed outside the institute, and incomplete clinical or imaging records were also excluded from the analysis. Pelvic MRI was performed as part of the routine preoperative assessment and locoregional staging of patients with histologically confirmed endometrial carcinoma. MRI was used to evaluate myometrial invasion, cervical stromal involvement, extrauterine extension, and pelvic lymph node status to assist in surgical planning. Clinical and demographic information, including age, menopausal status, FIGO stage, histological subtype, and tumor grade, was obtained from electronic medical records. Surgical and pathological details such as depth of myometrial invasion, lymphovascular space invasion (LVSI), and lymph node status were also recorded for analysis. The study protocol was approved by the Institutional Review Board of Gujarat Cancer Research Institute (IRC/P-34). Informed consent was obtained from all the study participants. All patients underwent preoperative pelvic MRI using a 3.0 T Siemens scanner with a body coil. MRI sequences were planned according to the European Society of Urogenital Radiology guidelines. The protocol included sagittal T2-weighted images, coronal T2-weighted non-fat suppressed and fat-suppressed images, axial T2-weighted fat-suppressed images, axial T1-weighted images, and non-contrast axial T1 fat-suppressed sequences. The DWI was performed in the axial plane using multiple b-values ranging from 0 to 800 s/mm 2 (50, 400, and 800 s/mm 2 ) (5). Dynamic contrast-enhanced MRI was performed following intravenous gadolinium administered at a dose of 0.1 to 0.2 mmol/kg body weight. Pre- and post-contrast T1 fat-saturated spoiled gradient (live acquisition with volume acquisition sequence) images were acquired in axial plane every 30 s over a duration of 5 min. Additional post-contrast sagittal and axial T1 fat-suppressed sequences were also taken. The DWIs were correlated with conventional magnetic resonance (MR) images for better anatomical localization of the endometrial lesion. The entire examination took 20 min. ADC maps were generated automatically using the Advantage Workstation (ADW) software. DWI findings were interpreted in conjunction with conventional MR images for better anatomical localization of the endometrial lesion. Diffusion sequences were planned according to the site of the lesion within the uterus. The following parameters were assessed: depth of myometrial invasion, cervical stromal involvement, serosal or parametrial invasion, and pelvic lymphadenopathy. Pelvic lymph nodes were considered suspicious if the short-axis diameter was greater than 10 mm and/or if restricted diffusion was present. Endometrial cancer was suspected when the lesions demonstrated high signal intensity on DWI images with a b-value of 800 s/mm 2 and corresponding low signal intensity on ADC maps. The lesion showed high signal intensity on DWI, but a high ADC value was not considered to represent true restricted diffusion, as this may occur through the T2 shine-through effect. Apparent Diffusion Coefficient (ADC) maps were reviewed independently by two radiologists with at least five years of experience who were blinded to the histopathological findings. Regions of interest (ROIs) were placed within the solid component of the primary tumor while avoiding necrotic or cystic areas. ROI placement was performed after correlation with T2-weighted, T1-weighted, and diffusion-weighted images. Two ROI techniques were evaluated for ADC quantification using a representative axial oblique slice of the lesion. One ROI encompassed the whole lesion, while the second focused on the area showing maximum diffusion restriction (focused ROI). Mean ADC values were measured manually by drawing circular two-dimensional ROIs over the darkest part of the lesion on ADC maps, with reference to T2-weighted and post-contrast images to avoid inclusion of necrotic areas. All surgical specimens underwent detailed histopathological examination by dedicated gynaecological pathologists. Tumor type and histological grade were assigned according to the WHO classification system. The depth of myometrial invasion was measured as the percentage of tumor infiltration into the myometrial thickness. The presence or absence of lymphovascular space invasion (LVSI) and lymph node metastases was also recorded from the final histopathology reports. p53 status was determined by immunohistochemistry and categorized according to the final pathology report as p53-mutant (abnormal expression pattern) or p53-wild type (normal expression pattern). Statistical analyses were performed using R programming software. Descriptive statistics were used to summarize patient demographics and clinical characteristics. Categorical variables were expressed as frequencies and percentages, while quantitative variables were presented as mean with standard deviation or mean with range, wherever appropriate. Interobserver agreement between the two radiologists was assessed using paired t -tests along with Bland–Altman analysis. Correlation between ADC values and histopathological parameters was evaluated using Pearson's coefficient analysis. Concordance between radiological staging and final histopathological staging was assessed using weighted Cohen's kappa coefficient. The strength of agreement for weighted Cohen's kappa was interpreted as follows: poor agreement or less than chance (κ < 0), slight agreement (κ = 0.01–0.20), fair agreement (κ = 0.21–0.40), moderate agreement (κ = 0.41–0.60), good agreement (κ = 0.61–0.80), and very good agreement (κ = 0.81–0.99). Receiver operating characteristic (ROC) curve analysis was performed to determine optimal ADC cutoff values for predicting various prognostic factors. Sensitivity, specificity, positive predictive value, negative predictive value, and overall diagnostic accuracy. A p -value of <0.05 was considered statistically significant.

Endometrial carcinoma is among the most commonly diagnosed malignancies among women in developed countries, and its incidence continues to rise worldwide, partly due to increasing obesity rates, metabolic disorders, and longer life expectancy ( Crosbie et al., 2022 ). Although many patients present with an early-stage disease and achieve favorable outcomes, accurate preoperative assessment remains essential for tailoring surgical management, identifying candidates for fertility-preserving treatment, planning adjuvant therapy, and estimating prognosis ( Crosbie et al., 2022 ). Magnetic resonance imaging (MRI) has become the preferred imaging modality for locoregional staging of endometrial carcinoma because of its superior soft-tissue contrast and multiplanar capabilities ( Nougaret et al., 2019 ). MRI provides detailed information regarding the tumor extent, depth of myometrial invasion, cervical stromal involvement, extrauterine spread, and lymph node status ( Nougaret et al., 2019 ). Previous studies have reported better locoregional staging accuracy when compared to ultrasound and computed tomography (CT) ( Freeman et al., 2012 ). However, accurate detection of the depth of myometrial invasion in conventional MRI may be challenging due to coexisting adenomyosis, leiomyomas, and myometrial compression by large tumors ( Nougaret et al., 2019 ). Functional MRI techniques have markedly enhanced tissue characterization in gynecologic imaging over time. Among these, diffusion-weighted imaging (DWI) has become a valuable adjunct to conventional MRI, providing information on the microscopic motion of water molecules within tissues ( Inoue et al., 2015 ). Quantitative analysis using apparent diffusion coefficient (ADC) measurements provides additional information about tumor cellularity and biological behavior ( Inoue et al., 2015 ). Tumors with high cellular density generally demonstrate greater diffusion restriction and lower ADC values ( Inoue et al., 2015 ; Moreira et al., 2023 ). Because of this, ADC measurements have increasingly been investigated as potential imaging biomarkers for predicting aggressive disease features before surgery ( Moreira et al., 2023 ). Recent studies show that ADC-based imaging biomarkers are becoming more important for risk assessment in endometrial carcinoma ( Moreira et al., 2023 ; Saleh et al., 2024 ). Lower ADC values have been linked to worse prognostic features, including high-grade histology, deep myometrial invasion, lymphovascular space invasion, and nodal metastasis ( Saleh et al., 2024 ). Updated imaging guidelines and the FIGO 2023 staging system also highlight the need for accurate preoperative evaluation to help tailor treatment ( Nougaret et al., 2019 ; Berek et al., 2023 ). However, differences in MRI protocols, ROI placement, and interpretation between observers make it hard to get consistent results across different hospitals ( Nougaret et al., 2019 ; Moreira et al., 2023 ). In this context, we evaluated the diagnostic utility of focused and whole-lesion ADC measurements for diagnosing endometrial carcinoma and how consistent the results are between different observers. We also studied how ADC values relate to important prognostic factors like tumor grade, histological subtype, depth of myometrial invasion, lymphovascular space invasion, nodal metastasis, and the final FIGO 2023 stage. By testing focused ROI analysis, we aimed to see if ADC measurements could be reliable and useful imaging biomarkers for preoperative assessment of endometrial carcinoma.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.