RPS 316 - Hot Topic: metastasis detection and oligometastasis

Author Block: J. Wu¹, A. De Araujo², S. Khozin³, M. J. Willemink¹; ¹Palo Alto, CA/US, ²São Paulo/BR, ³Cary, NC/US
Purpose: Oligometastatic disease (OMD) refers to a state of metastatic cancer, characterized by a limited number of metastatic lesions. Local treatments such as ablative therapy could potentially be curative to OMD. OMD-diagnosis is based exclusively on imaging findings. OMD challenges the older binary view of cancer as either localized or metastatic. Given the importance of imaging findings for diagnosing OMD, we explored the use of OMD in real-world radiological practice.
Methods or Background: We evaluated clinical radiology-reports from USA-located healthcare systems. Healthcare providers ranged from imaging centers, small- and medium-sized hospitals, to academic and oncology-focused providers. A real-world imaging data platform was used to search, filter, and select radiology reports (Segmed).
Results or Findings: A total of 33,747,977 reports of 11 radiology modalities (including X-ray, MRI, CT, and PET) from healthcare providers in 40 states were included in the search. OMD was mentioned in 164 radiology reports from 109 different patients. Clinicians provided OMD-information in the indication and history in 148 radiology reports of 95 patients, while radiologists mentioned OMD in 18 reports of 17 patients. OMD was mentioned in 20 different primary tumor-types, most commonly breast (n=39 reports of 22 patients), lung (n=33 reports of 26 patients), and prostate cancer (n=16 reports of 13 patients). Modalities with OMD-mention included CT (n=58), X-ray Angiography (n=44), PET/CT (n=35), MRI (n=26), and X-ray (n=1). The first mention of OMD was in 2017 (n=2 reports), which increased to n=33 in 2024.
Conclusion: In this large scale observational study, we found that despite the importance of imaging, the term OMD is rarely used in routine radiology practice in the USA. These results suggest the need to improve the definition and guidelines around OMD.
Limitations: Retrospective study with only US data, no EU data.
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: J. Chen, Z. Yuan, Y. Li, Y-L. He, Z. Jin, H-D. Xue; Beijing/CN
Purpose: To develop and validate the first nomogram that integrates Node-RADS with clinical parameters for predicting lymph node metastasis (LNM) in cervical cancer. We further aim to evaluate the clinical utility of this integrated model across various scenarios, particularly in fertility-sparing candidates and cases with indeterminate lymph nodes.
Methods or Background: This retrospective study included 465 cervical cancer patients. Two radiologists independently evaluated lymph nodes using Node-RADS on preoperative MRI. Diagnostic metrics (sensitivity, specificity, etc.) were calculated. A combined model incorporating Node-RADS and clinicopathological predictors was developed and presented as a nomogram. Diagnostic performance was evaluated via ROC analysis and decision curve analysis. Subgroup analyses were performed in fertility-sparing candidates (including age <45 years or tumor size ≤2 cm), lymph nodes with subcentimeter metastasis (short-axis diameter <1 cm), and those assigned a Node-RADS score of 3.
Results or Findings: Node-RADS score 3 provided optimal diagnostic balance (sensitivity 68.7%, specificity 88.2%, NPV 92.8%), while scores 4–5 showed high specificity (≥99%). The combined model outperformed Node-RADS alone (AUC 0.872 vs. 0.827), with good calibration (p = 0.44) and a greater net benefit at threshold probabilities of 10–40%. The nomogram enabled individualized risk estimation. Subgroup analysis confirmed strong performance in fertility-sparing candidates (AUC 0.884 for age<45; 0.851 for tumor ≤2 cm), identified patients with subcentimeter LNM（AUC 0.877）, and improved stratification of Node-RADS 3 nodes (AUC 0.838; LR+ 2.61).
Conclusion: Node-RADS is a robust independent predictor for preoperative lymph node metastasis assessment in cervical cancer. Node-RADS-integrated nomogram represents a powerful and clinically useful tool for the individualized preoperative prediction of LNM in cervical cancer, potentially enhancing decision-making for personalized treatment strategies.
Limitations: The generalizability of our findings may be limited by the study's retrospective design and single-center patient cohort.
Funding for this study: The CAMS Innovation Fund for Medical Sciences (grant no. 2023-I2M-C&T-B-020),
National High-Level Hospital Clinical Research Funding (grant no. 2025-PUMCH-C-029).
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: X-P. Meng; Wuxi/CN
Purpose: To evaluate the diagnostic performance and prognostic value of the Node Reporting and Data System 1.0 (Node-RADS) for regional lymph node (LN) metastasis in pancreatic ductal adenocarcinoma (PDAC) using preoperative CT.
Methods or Background: This single-center retrospective study included patients with PDAC who underwent contrast-enhanced CT followed by surgical resection between January 2017 and July 2023. Each regional LN station was assessed using Node-RADS, in which nodal size and morphologic features were integrated into a five-point scoring system (1–5, from very low to very high suspicion), with histopathology serving as the reference standard. Diagnostic performance, including sensitivity and specificity, was evaluate at the patient, LN group, and station levels. Cox regression analyses were conducted to evaluate the prognostic impact of Node-RADS after adjusting for clinicopathological factors.
Results or Findings: A total of 216 patients (mean age ± standard deviation, 61.2 ± 10.0 years; 138 males) were included, with LN metastases detected in 154 of 216 patients (71.3%). A Node-RADS score of 3 or greater yielded the optimal cutoff for predicting lymph node metastasis, with sensitivity and specificity of 77.3% (119/154) and 83.9% (52/62) at the patient level, 74.8% (104/139) and 81.8% (63/77) for Group 1, and 77.3% (34/44) and 82.6% (142/172) for Group 2. Across regional nodal stations, Node-RADS achieved a sensitivity of 60.0–85.7% and a specificity of 79.5–98.8%. Higher Node-RADS categories were independently associated with worse overall survival (hazard ratio [HR], 1.34; 95% CI: 1.15–1.56; p = 0.012) and disease-free survival (HR, 1.45; 95% CI: 1.17–1.92; p = 0.007).
Conclusion: Node-RADS provides robust diagnostic accuracy for LN metastasis in PDAC and may serve as an effective preoperative prognostic predictor.
Limitations: Retrospective; Single-center study
Funding for this study: the foundation of the Wuxi Municipal Bureau on Science and Technology (Y20232002)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This single-center retrospective study was approved by the Ethics Review Committee of the Affiliated Wuxi People's Hospital of Nanjing Medical University, and the requirement for informed consent was waived due to the retrospective design.

Author Block: J. Berger, F. K. E. Mankertz, N. Maalouf, A. S. Brendlin, J. Herrmann, K. Nikolaou, S. Afat; Tübingen/DE
Purpose: To evaluate whether AI-based contrast boosting of single-phase CT improves the detection and visual conspicuity of subcentimetre liver metastases from choroidal melanoma, using liver MRI as the reference standard.
Methods or Background: This retrospective single-centre study included 68 patients (mean age, 69.6 ± 11.2 years; 46% men) with histologically confirmed choroidal melanoma and suspected liver metastases, enrolled between January 2019 and July 2022. All underwent portal venous phase contrast-enhanced whole-body CT and dedicated liver MRI. CT data were post-processed using a commercial AI-based contrast-boosting algorithm (ClariACE, ClariPi, Seoul, South Korea). MRI-identified metastases served as the reference standard and were stratified by size (<5 mm, 5-10 mm, >10 mm). Two radiologists independently assessed visibility on standard and AI-boosted CT, recording binary detection and conspicuity on a five-point Likert scale. Detection rates were compared using McNemar’s test, conspicuity using the Wilcoxon signed-rank test, and inter-reader agreement using Cohen’s kappa and intraclass correlation coefficient.
Results or Findings: A total of 104 liver metastases were identified on MRI, including 58 <5 mm. For sub-5 mm lesions, AI-boosted CT significantly improved sensitivity from 57% to 81% (p < 0.001). A modest, non-significant increase was observed for 5-10 mm lesions (82% to 89%, p = 0.09), while detection of >10 mm lesions remained 100%. Median conspicuity scores for sub-5 mm lesions increased from 3.0 to 4.3 (p < 0.001). Inter-reader agreement was excellent for detection (κ = 0.84) and conspicuity (ICC = 0.87). Increased conspicuity correlated with improved detection of small lesions.
Conclusion: AI-based contrast enhancement significantly improves sensitivity and conspicuity for detecting subcentimetre liver metastases, especially those <5 mm, without additional patient risk and may facilitate earlier detection, staging, and treatment planning.
Limitations: Retrospective, single-centre design and modest sample size may limit generalisability.
Funding for this study: None.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Local IRB board approval was obtained.

Author Block: C. Pontiggia, V. Magni, M. Vaiani; Milan/IT
Purpose: To compare Gd-EOB-DTPA hepatobiliary phase and DWI sequences with dynamic phase on contrast-enhanced MRI for detection of liver metastases from uveal melanoma, as assessed by two readers with different levels of expertise.
Methods or Background: This retrospective study included patients with liver metastases from uveal melanoma who underwent surgery between 2017 and 2025. Preoperative liver-focused MRI scans were independently reviewed by two readers with different expertise. Two imaging protocols were evaluated: T1/T2-weighted with dynamic sequences, and T1/T2-weighted with DWI/ADC and hepatobiliary phase sequences. Histopathological results from surgery and follow-up MRI for non-resected lesions were retrieved. Lesion counts were compared using the Wilcoxon signed-rank test. Inter-reader agreement was assessed with the ICC.
Results or Findings: 16 patients were included. Senior reader detected 68 lesions with DWI/hepatobiliary phase sequences and 25 with dynamic contrast-enhanced sequences, while junior reader detected 45 and 33 lesions, respectively. Of the 68 lesions detected by senior reader on DWI/hepatobiliary sequences, 60 had histopathological confirmation, seven were confirmed at follow-up MRI, and one was a false positive (positive predictive value of 98.5%). In 9/16 patients (56.3%), MRI detected all surgically proven lesions, whereas in 7/16 patients surgery revealed additional lesions not visible on MRI. Senior reader identified significantly more lesions with DWI/hepatobiliary than with dynamic sequences (median 3.5 versus 1.0, p = 0.0014). Junior reader also detected significantly more lesions with DWI/hepatobiliary than with dynamic sequences (median 2.0 versus 1.0, p = 0.017). Inter-reader agreement was moderate (ICC of 0.66 for dynamic and 0.64 for DWI/hepatobiliary sequences).
Conclusion: DWI/hepatobiliary MRI sequences improved lesion detection compared with dynamic sequences, particularly for the experienced reader; furthermore, they showed a high positive predictive value, with complete concordance to surgery in over half of patients.
Limitations: Small sample size.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: A. Choucair¹, J. Dupont², G. H. Jardali¹, L. Dawi¹, L. LAWRANCE³, S. Ammari², C. Massard², A. Italiano⁴, N. Lassau²; ¹Paris/FR, ²Villejuif/FR, ³Villepinte/FR, ⁴Bordeaux/FR
Purpose: Total tumor volume (TTV) on CT scans, circulating tumor DNA (CTDNA) from liquid biopsy (LB), known as tumor fraction (TF), as well as 3Danthropometric measurements, are indicators for predicting overall survival (OS) in metastatic patients.
Methods or Background: Patients with a baseline CT scan and LB (STING, NCT04932525, Gustave Roussy) were included from January 2021 to December 2023 in this study. CT scans were annonymized and manually annotated in 2D by radiologists, with a 3D TTV estimated (Tumor volume = 2/3 xMesh surface area x Minor axis). TF was reported as a percentage.
The Anthropometer3DNet software provided measurements for muscle, superficial, and deep fat masses; combined with TTV and with Cox model coefficients to generate a CT scan risk-score. Patients were grouped based on TF (≥0%) and divided into two subgroups according to the CT scan risk-score.
OS was the primary endpoint, and the analyses were carried out using Kaplan-Meier estimators and log-rankstatistics.
Results or Findings: Among 720 patients with 38,283 annotated metastases, the median OS was 13.1 months, with a median TF and TTV of 3.3% and 91.6 cm³. A total of 213 patients had TF=0% despite measurable tumor burden (median TTV 35.3 cm³), and their OS was shorter (10.2 months) compared to TF>0% (25.5 months). Within the TF=0% group, patients with a low CT-scan risk score lived longer (35.4 months) than those with a higher score (19.4 months). In the TF>0% group, OS was 13.2 months for low CT-scan risk-score but dropped to 7.5 months for those with TTV above the median.
Conclusion: TTV and anthropometric measurements, when combined with TF offer a more precise prediction of OS.
Limitations: This is a retrospective single-center study, relying on 2D TTV estimation with a heterogeneity in the cohort.
Funding for this study: No fundings.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: