RPS 2307 - Malignant lesions of the female pelvis: advances in imaging techniques, diagnostic and follow-up

Author Block: R. V. Ninkova, M. Gennarini, V. Miceli, A. Cupertino, F. Curti, C. Catalano, L. Manganaro; Rome/IT
Purpose: Aim of this study was to develop a machine learning model based on Magnetic Resonance Imaging (MRI) to stratify the single-subject risk of endometrial cancer (EC).
Methods or Background: From September 2023 to July 2024, we collected MRI images from 41 patients. Among these subjects, 15 patients (36.6%) belonged to class "microsatellite instability (MSI)" and 26 patients (63.4%) belonged to class "microsatellite stability (MSS)", according to histological diagnosis from biopsy. This image set was used for the training and cross-validation of different machine learning models. A robust radiomic approach was applied, under the hypothesis that radiomic feature could be able to capture the disease heterogeneity among the two groups.
Results or Findings: Three models consisting of 3 ensembles of machine-learning classifiers (random forests, support vector machines and k-nearest neighbor classifiers) were developed for the binary classification task of interest (“MSI” vs. “MSS”), based on supervised learning, using histological diagnosis from biopsy as reference standard. The best model showed ROC-AUC mean value of 84 % [78.4-89.8], accuracy mean value of 75.6% [69.6-81.7], sensitivity mean value of 68.9 % [59.3-78.5], specificity mean value of 79.5 % [74-85], PPV mean value of 71.66 %[58.1-73.9], and NPV mean value of 81.6% [76.3-86.9] (p <0.005 mean value).
Conclusion: The radiomics-based machine learning model achieved a high diagnostic performance in the molecular stratification of patients with EC, which could improve risk stratification and support clinical therapeutic decision.
This is a preliminary study which could provide a new perspective for patients with EC, allowing a complete and accurate identification of the disease and promoting personalized treatment.
Limitations: The primary limitation of this study is the small patient cohort. Expanding the sample size and investigating additional molecular pathological profiles will be necessary for further validation.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was performed in line with the principles of the Declaration of Helsinki.

Author Block: A. Rame¹, S. Bottazzi¹, G. Avesani¹, M. Bonatti², V. Celli¹, E. Perrone¹, T. Pasciuto¹, B. Gui¹, E. Sala¹; ¹Rome/IT, ²Bolzano/IT
Purpose: To investigate the role of MRI in the preoperative evaluation of endometrial cancer (EC) considering the new FIGO 2023 staging system.
Methods or Background: Patients diagnosed with EC from two institutions between 2019 and 2023 were retrospectively included. Inclusion criteria were the availability of preoperative MRI and biopsy, molecular data and definitive histopathological data obtained after surgery. Two radiologists retrieved MRI findings from reports and combined them with biopsy results (grading and histology) to determine the FIGO 2023 preoperative stage. Definitive histopathological and molecular data served as the gold standard (final staging). The preoperative evaluation was compared with final staging (FIGO stages I, II, III and IV as independent categories) and discrepancies were recorded.
Results or Findings: 231 patients were included. The agreement between preoperative evaluation and final staging was 74% (171/231), while the FIGO stage was discordant at 26% (60/231). Causes of discordance were: lymph node (LN) involvement in 26.7% (16/60), changes in grading or histological subtype between biopsy and surgical specimen in 23.3% (14/60), metastases and carcinosis in 15% (9/60: 5 in the upper abdominal peritoneum, 1 in the pelvic peritoneum and 3 distant), the presence of substantial lymphovascular space invasion (LVSI) in 18.2% (10/55). Less common causes of discrepancies were myometrial invasion (6.7%; 4/60), cervical stroma invasion (5%; 3/60), ovarian involvement (3.3%; 2/60) and vaginal invasion (1.7%; 1/60). Overall, parameters not assessable preoperatively (LVSI and definitive grading and histological subtype) accounted for staging discrepancies in 10.4% (24/231) of cases.
Conclusion: Our study showed that the inclusion of parameters not assessable preoperatively in the new FIGO 2023 staging system does not significantly diminish the role of MRI in the preoperative evaluation of EC, resulting in staging discrepancy in only approximately 10% of cases.
Limitations: No limitations.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: ID 6194, approved on 7/12/2023

Author Block: X. Liu, Z. Yuan, Y. Li, J. Ren, Y. He, H. Xue, Z. Jin; Beijing/CN
Purpose: To compare the diagnostic efficiency of enhanced T1 mapping MR imaging and dynamic contrast-enhanced (DCE) imaging for assessing the depth of myometrial invasion in patients with endometrial cancer.
Methods or Background: 46 women diagnosed with endometrial cancer underwent preoperative MR imaging. Two radiologists independently assessed the depth of myometrial invasion, categorized as no myometrial invasion, superficial myometrial invasion, or deep myometrial invasion, on T2WI+DWI+DCE MR imaging, followed by T2WI+DWI+enhanced T1 mapping MR imaging, four weeks later. The findings were then compared to histopathological examinations. The diagnostic performance comparison was conducted using the chi-square test.
Results or Findings: The overall accuracy for accessing depth of myometrial invasion on T2WI+DW+DCE and T2WI+DW+enhanced T1 mapping were 76.1%, 80.4% for reader 1, and 78.3%, 80.4% for reader 2, respectively. The increment was not statistically significant for either reader. While assessing the absence of myometrial invasion, the precision, sensitivity, and specificity achieved by both radiologists using DCE were 100%, 16.7%, and 100%, whereas for enhanced T1 mapping the precisions were 100%, with sensitivities of 50% and 33%, and specificities of 100%. In evaluating superficial myometrial invasion, these values using DCE were 82.4%/82.8%, 84.8%/87.9%, and 53.8%; using enhanced T1 mapping, these values were 87.5%/85.3%, 84.8%/87.9%, and 69.2%/61.5%. Assessment of deep myometrial invasion achieved these values of 54.5%/60%, 85.7%, and 87.2%/89.7% with DCE; 54.5%/60%, 85.7%, 87.2%/89.7% with enhanced T1 mapping. Inter-reader agreement, measured with kappa values, was 0.831 with DCE-MRI and 0.899 with enhanced T1 mapping. Both radiologists concurred that enhanced T1 mapping substantially improved diagnostic confidence over DCE-MRI.
Conclusion: Enhanced T1 mapping demonstrates superior diagnostic efficiency in the evaluation of myometrial invasion in endometrial cancer compared with DCE MR imaging.
Limitations: This was a pilot study with a small sample size.
Funding for this study: This work was supported by grants from Natural Science Foundation of China (grant No. 82271886).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the institutional review board, which waived the requirement of informed consent.

Author Block: A. Amerighi, M. Dolciami, A. Napoletano, I. Isufi, G. Avesani, E. Sala, B. Gui; Rome/IT
Purpose: Cervical cancer (CC) is increasingly diagnosed at earlier stages due to cancer screening
programs. Cone biopsy is often necessary to rule out invasive disease and can sometimes
suffice for treatment; however, determining residual disease is essential for treatment
planning.
Despite its crucial role in managing CC, the role of Magnetic Resonance Imaging (MRI) in
determining residual disease in patients after conization remains poorly understood, with
only a few studies focusing on this topic.
We aimed to assess MRI accuracy in detecting residual disease after cervical conization
for early-stage cancer and compare the diagnostic performance of MRI with and without a
contrast agent.
Methods or Background: We retrospectively enrolled all patients with early-stage CC who received conization
before MRI and then surgically treated (hysterectomy, trachelectomy, or re-conization).
Two radiologists evaluated MRI scans for residual disease in the cervix, blinded to surgical
outcomes.
Results or Findings: 119 patients were included in the study. MRI showed an accuracy of 78,18%, sensitivity of
65.45%, specificity of 90.91%, positive predictive value (PPV) of 87.80%, and negative
predictive value (NPV) of 72.46%. There was no significant change in the MRI
performance with and without contrast medium, in accuracy (81,05% vs 75,80%,
respectively), sensitivity (76,92% vs 55,17%), specificity (85,19% vs 96,43%), PPV
(83,33% vs 94,12%), and NPV (79,31% vs 67,50%).
Conclusion: MRI showed good accuracy in assessing residual tumor after conization with high
specificity and PPV; however, the main problem remains the high number of false
negatives. The use of contrast medium did not significantly affect the MRI's performance,
as it led to a slight increase in true negatives but also an increase in false positives.
Limitations: Retrospective study
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: ID prot. 7073 prot. 0022234/24

Author Block: M. Gennarini, R. V. Ninkova, V. Miceli, A. Cupertino, S. Riccardi, F. Curti, C. Cutonilli, C. Catalano, L. Manganaro; Rome/IT
Purpose: This study evaluates the diagnostic accuracy of the Node-RADS score in Magnetic Resonance Imaging (MRI) and assesses the significance of the Apparent Diffusion Coefficient (ADC) in identifying actual lymph node neoplastic involvement in patients with cervical cancer (CC).
Methods or Background: A retrospective analysis was performed on cervical cancer patients who underwent preoperative MRI and radical surgery with lymphadenectomy from February 2018 to July 2024. Lymph node involvement risk was evaluated for the primary pelvic lymph nodes, assigning scores ranging from 1 to 5: 1 (very low), 2 (low), 3 (unclear), 4 (high), and 5 (very high). The mean ADC, relative ADC (rADC), and corrected ADC (cADC) for lymph nodes rated as Node-RADS 3, 4, and 5 were measured and compared against lymph node histology results.
Results or Findings: In this study, 156 lymph nodes from 54 patients, with a Node-RADS score greater than 2, were included. Of these, 108/156 (69.2%) were histologically confirmed as positive, while 48/156 (30.8%) were negative. The mean ADC value proved most statistically significant, showing a sensitivity of 87.0%, specificity of 82.6%, a positive predictive value (PPV) of 92.2%, and a negative predictive value (NPV) of 73.1%. The area under the curve (AUC) values for Node-RADS >2 and Node-RADS 3 were 0.892 and 0.677, respectively, with ADC thresholds of <0.963×10^-3 mm^2/s and <0.983×10^-3 mm^2/s.
Conclusion: The ADC measurement of lymph nodes provides crucial data that aids in the accurate classification of patients with cervical cancer. Utilizing an ADC cut-off of 0.963×10^-3 mm^2/s, the MRI demonstrated high diagnostic sensitivity. Thus, ADC serves as a valuable tool for enhancing the diagnostic process in cervical cancer management.
Limitations: Single center retrospective design study.
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: N/A

Author Block: X. Ou, Y. Li, Y. Pei, W. Li; Hunan/CN
Purpose: To investigate the feasibility of MAGiC (sy-T2WI; sy-T1, sy-T2, and sy-PD maps) to determine treatment plan and predict recurrence risk factors (RRF) ，short-term treatment efficacy (STE) in cervical cancer (CC) patients using high-resolution T2-weighted (hr-T2WI）and diffusion-weighted imaging (DWI) as reference standards.
Methods or Background: 194 patients suspected of CC were enrolled. For CC underwent CCRT, hrT2WI performed in 2 months for evaluating STE (completed response and no-CR ). MAGiC can generate synthetic morphologic images (syT2WI) and quantitative synthetic images (sy-T1, T2 and PD maps). For syT2WI, The evaluation of image quality and staging using sy-T2WI and hr-T2WI was conducted. The accuracy, sensitivity and specificity of sy-T2WI were analyzed for making treatment strategies (IB-IIA: surgery; IIB-IVA: CCRT). The AUC was used to predict RRF and STE, use quantitative sy-T1, T2, PD maps.
Results or Findings: 69 out of 119 CC(IIB－IVA) received CCRT . 50 out of 119 CC(IA-IIA) received surgery. SyT2WI was no significant differences with hrT2WI in four aspects (P＞0.05). The accuracy, sensitivity and specificity of sy-T2WI was 0.908, 0.908 ,0.999 for differentiating IB-IIA from IIB-IVA, and an excellent agreement between them (k = 0.935; p < 0.001). T2, T1 and ADC values had a significant differences to identified CR from no-CR and identified RRF from no-RRF (P < 0.05) but PD.The diagnostic performance of ADC was inferior to T2 for STE, which was similar to T2 for RRF. Furthermore, T1 ＋T2 was superior to ADC for predicting RRF (AUC: 0.980 vs. 0.776; p = 0.005) and forecasting STE (AUC: 0.982 vs. 0.737; p < 0.001) .
Conclusion: MAGiC is a promising technique for deciding therapeutic planning and predicting RRF and STE in CC , which is similar and even superior to hr-T2WI and DWI.
Limitations: Not applicable
Funding for this study: Not applicable
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: the local Research Ethics Committee

Author Block: C. Vercelli¹, F. Rosa², C. Martinetti¹, D. Schettini¹, A. Gastaldo², N. Gandolfo¹; ¹Genova/IT, ²Savona/IT
Purpose: In 2022, an important Consensus Statement and diagnostic algorithm for assessing the risk of uterine leiomyosarcoma (LMS) using MRI was published (Hindman et al.). The aim of our study is to compare the diagnostic performance of the recently proposed algorithms available in the literature for the differential diagnosis of uterine mesenchymal tumors on MRI.
Methods or Background: After a literature review, three diagnostic algorithms were identified: Hindman et al. (2022), Rosa et al. (2023), and Wahab et al. (2020). A retrospective evaluation was conducted on 50 MRIs performed for suspected uterine mesenchymal masses. Each lesion was categorized in a double-blind manner by an expert Radiologist in gynecological imaging according to the three algorithms. Histological examination or appropriate follow-up was used as reference standard for the final diagnosis.
Results or Findings: The study included 52 uterine masses: 37 benign lesions (leiomyomas) and 15 malignant lesions (7 LMS, 2 STUMP, 2 endometrial stromal sarcomas, 1 adenosarcoma, and 3 non-sarcomatous malignant lesions). The three algorithms demonstrated equal specificity in diagnosing LMS (97.3%). However, when analyzing the performance in the differential diagnosis between malignant and benign lesions (including rarer histotypes), the Rosa et al. algorithm showed superior sensitivity (93.33% vs 73.33%) and diagnostic accuracy (96.15% vs 90.38%).
Conclusion: The differential diagnosis of uterine mesenchymal lesions presents a diagnostic challenge with significant implications for management and outcome. The three algorithms demonstrated high diagnostic accuracy; in particular, the Rosa et al. algorithm proved more effective in identifying also rarer malignant histotypes
Limitations: Retrospective study; Small cohort with high prevalence of sarcomas.
Funding for this study: No funds
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: N. Registro CER Liguria: 78/2023 - DB id 12883
N. CET - Liguria: 104/2024 - DB id 13756

Author Block: E. Zlotykamien Taieb; Vincennes/FR
Purpose: To externally validate a previous MRI-based expert consensus algorithm and evaluate the potential improvement of an MR-scoring system's accuracy in diagnosing uterine mesenchymal tumors.
Methods or Background: A bicentric retrospective observational cohort study was conducted from January 2018 to December 2023 including women with a pathological diagnosis of myometrial tumor following a pelvic MRI within six months. Clinical and MR criteria were recorded blindly by two radiologists. Continuous variables were analyzed using a Mann–Whitney test, and categorical variables using Fisher’s exact test. Odds ratios for predicting malignancy were calculated with 95% confidence intervals and p-values.
Results or Findings: The cohort included 455 women with mesenchymal tumors: 437 leiomyomas, 2 STUMPs, and 16 leiomyosarcomas. Using initial criteria (pelvic hypertrophic lymph nodes, T2W signal intensity, DW signal intensity compared to endometrium, and ADC cutoff value of 0.9 × 10⁻³ mm²/sec), the model accurately classified 420 out of 455 cases (Accuracy: 80.9%, sensitivity was 61.1% and specificity 93.6%. A refined approach added “irregular tumor margins” and menopausal status, modified DW signal compared to bladder, and an ADC cutoff value of 1.23 × 10⁻³ mm²/s, improving classification to 445 out of 455 cases (Accuracy: 92.5%; sensitivity: 83.3% ; specificity: 98.4%. The refined algorithm significantly improved accuracy, allowing the development of a 5-category scoring system.
Conclusion: MR imaging effectively differentiates leiomyosarcoma from other uterine tumors. The new algorithm increases diagnostic accuracy, helping prevent morcellation risks in women with uterine leiomyosarcoma
Limitations: Our study is retrospective, which does not permit avoidance of all biases.
Even though the cohort was bicentric, the prevalence of malignancy remained low.
Due to the sample size, differentiation between STUMP and frankly invasive UMT was not possible.
No external validation of our modified score is yet available, necessitating further studies.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Institutional ethics committee approval and waiver of informed consent (CRM-2405-410)

Author Block: D. Garg, R. Kaur, R. Bedi, R. Gupta, B. Goel, U. Handa; Chandigarh/IN
Purpose: Proton magnetic resonance spectroscopy (¹H-MRS) is a non-invasive imaging technique that offers insights into biochemical metabolism. While its role in brain and prostate malignancies is well-established, its application in evaluating adnexal masses is still in its early stages. This study investigates the role of ¹H-MRS in characterising adnexal masses and its value, in conjunction with dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI), in enhancing the diagnostic accuracy of conventional MRI for differentiating adnexal masses.
Methods or Background: We conducted a prospective study including 62 histologically confirmed adnexal masses (19 benign and 43 malignant), which were indeterminate on ultrasound. Patients underwent conventional MRI, DWI (apparent diffusion coefficient), DCE-MRI (time intensity curves), and ¹H-MRS. Single-voxel spectroscopy analysed resonance peaks for choline, N-acetyl aspartate (NAA), creatine, lactate, and lipids. Choline-to-creatine ratios were compared between benign and malignant tumours, and ROC curves were used to define optimal thresholds.
Results or Findings: Conventional MRI showed sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 97.67%, 57.89%, 84%, and 91.67%, respectively. We detected choline peak in 100% of malignant and 47.4% of benign masses, NAA in 79.1% of malignant and 31.6% of benign masses, as well as lipid peaks in 36.8% of benign and 20.9% of malignant masses. The mean choline-to-creatine ratio is 1.05+0.55 in benign and 12.18+12.38 in malignant tumours, statistically significant (p<0.05). With a choline-to-creatine threshold of 2.01, sensitivity, specificity of 100% is achieved. The addition of ¹H-MRS, DWI, and DCE-MRI improved the diagnostic accuracy of conventional MRI, with 100% sensitivity, 94.74% specificity, 97.73% PPV, and 100% NPV.
Conclusion: 1H-MRS has promising role in characterising adnexal masses and in conjunction with DWI and DCE-MRI, enhances the diagnostic accuracy of conventional MRI.
Limitations: The sample size is limited.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The ethics committee notification can be found under the number ECR/658/Inst/PB/2014/RR-20