10:30   Oncology Chair: Harald Groen Magnetic resonance thermometry performance during MR-guided hyperthermia treatment of soft-tissue sarcoma in the pelvis and lower extremities Spyridon Karkavitsas, Marianne Goeger-Neff, Lars Lindner, Maarten Paulides Abstract: Introduction: To evaluate magnetic resonance thermometry (MRT) during deep-regional hyperthermia therapy (HT) of soft-tissue sarcoma (STS) in the pelvis and lower extremities, and to investigate the influence of pre-treatment motion on MRT precision. Methods: According to our clinical standard protocol, patients with STS in the pelvis (17, 45 treatments) and lower extremities (16, 42 treatments) received HT and chemotherapy at the LMU Klinikum. The temperature increase was monitored non-invasively using MRT: two consecutive Double-Echo Gradient-Echo (DEGRE) sequences every 10 minutes. 3D temperature maps were created from the MR phase data using only the 2nd echo of the DEGRE sequence. MonAI Label (https://github.com/Project-MONAI) automatically segmented patient and body fat volumes, which were exploited for drift correction. The results were benchmarked against SigmaVision (Sennewald SMT, Germany), which utilizes both echoes and only the external oil reference tubes. The standard deviation of the voxel-wise temperature difference of the DE-GRE scans within the patient contour quantified temporal precision. Accuracy was assessed in three lower extremity patients (six treatments), using invasive tumor probes as the golden standard, and quantified by the mean absolute difference between probes and MRT. To investigate the impact of pre-treatment motion on MRT precision, we quantified pre-treatment gross motion and intestinal gas motion by the Jaccard Coefficient (JC). The association between JC and acceptable temporal precision (< 1°C) was analyzed using a ROC analysis. Results: Compared to the clinical standard dual-echo MRT, our post-processing algorithm (single-echo /body-fat + oil tubes) improved the mean temporal precision from 1.3±0.4°C to 1.1±0.3°C for the pelvis and from 1.0±0.3°C to 0.8±0.2°C for the lower extremities. Accuracy improved from 1.1°C to 0.8°C in the lower extremities. Pre-treatment gross motion was a good predictor of precision with AUC=0.80-0.86 (pelvis) and 0.81-0.83 (lower extremities), and better than intestinal gas motion (0.52-0.58). Conclusions: Single-echo MRT had significantly better precision than dual-echo MRT and body-fat-based field drift correction significantly improved MRT accuracy. Pre-treatment gross motion was a good predictor for MRT precision, in contrast to intestinal gas motion. This study shows that acceptable MRT accuracy is obtained in a subgroup while guiding research into correction strategies. Quantitative feature analysis of B-mode and contrast-enhanced ultrasound for the assessment of uterine fibroid vascularity Maria Papamichail, Ferenc Kandi, Catarina Dinis Fernandes, Massimo Mischi Abstract: Most women’s hysterectomy indications are due to uterine fibroids [1]. The selection and success of less drastic treatment options often depends on the vascularity of the fibroid. To provide a more objective assessment of fibroid vascularity, this study explores the use of quantitative features from 2D dynamic B-mode and contrast-enhanced ultrasound (CEUS) images acquired from a group of patients with a fibroid diagnosis. Using contrast-ultrasound dispersion imaging (CUDI), a framework that performs quantitative analysis of CEUS image data through convective dispersion modelling of the passage of ultrasound contrast agents, additional dispersion-indicating features are extracted [2]. CEUS and CUDI have not yet been explored for quantitative fibroid vascularity assessments. The fibroids were differentiated depending on the degree of their vascularity into low and high vascularity groups, based on patient case reports. Of the 12 fibroids investigated, 5 were classified as low, and 7 as high vascularity. Features related to spatial distributions such as standard deviation (SD), kurtosis and skewness were extracted for the B-mode and CEUS images, while features based on spatiotemporal dynamics such as coherence, correlation and mutual information (I) were extracted for the CUDI parametric maps. The features were aggregated for each group, then normalized. Finally, they were compared visually using box charts. While B-mode feature values showed no significant difference between groups, the low vascularity group in CEUS images had lower SD and higher kurtosis and skewness than the high vascularity group. Moreover, coherence, correlation and I on average exhibited lower values in the low vascularity group in contrast to the high one. Binary k-nearest neighbours classifiers based on CEUS and CUDI parametric map features were able to distinguish the two groups. The classifier based on CEUS achieved a sensitivity of 1.00 and a specificity of 0.40, while the classifier based on CUDI reached a sensitivity of 0.71 and a specificity of 0.80. The receiver operating characteristics AUC was 0.80 for CEUS and 0.83 for CUDI. The results of this study suggest that CEUS and CUDI feature values differ depending on the degree of vascularity of fibroids and could be of value as vascularity assessment methods. In-house device development in compliance with medical regulations: a case study of 3D-printed custom brachytherapy applicators Robin Straathof, Sharline van Vliet-Pérez, Linda Wauben, Ben Heijmen, Inger-Karine Kolkman-Deurloo, Remi Nout, Jenny Dankelman, Nick van de Berg Abstract: Purpose With the introduction of the more stringent Medical Device Regulation (MDR) 2017/745, in-house manufacturing of medical devices (MDs) is challenged. Hospitals can only develop or modify MDs if these devices among others fulfil safety and performance requirements, are manufactured and used under quality management systems (QMSs), fulfil needs that cannot be met by equivalent devices, and are supported by technical documentation. This work describes our efforts from a regulatory perspective in the development and evaluation of a novel 3D-printed brachytherapy applicator and accompanying software. Methods The workflow used investigational medical device dossier templates provided by the Medtech Innovation Support Office (MISO) of the Erasmus Medical Centre. To determine whether current needs were unmet by equivalent devices, the intended use was specified (IEC 62366-1:2015), and a market analysis was conducted. After justification, general safety and performance (Annex I, MDR), and design requirements were drawn up. The device and software were designed accordingly, incorporating risk analyses (ISO 14971:2019, and IEC 62304:2015). Applicator and software design iterations were documented in a design history file, and with git version control, respectively. Quality control of the hardware manufacturing was safeguarded by selecting a manufacturer with QMS certification (ISO 13485:2016). Software development and maintenance were controlled through use of an issue tracking platform. Several pre-clinical evaluations were performed: (1) material dose attenuation, (2) steam sterilisation measurements at 134°C and 3.04 bar, (3) biological evaluation (ISO 10993-1:2020), (4) needle deflections in a phantom, (5) applicator reconstructioning on MRI, and (6) virtual dose planning for 20 patients previously treated with a clinically used commercial applicator. Results Technical documentation of the design and manufacturing was compiled. Safety evaluations showed that: (1) material dose response was water-equivalent, (2) desired channel temperatures were reached, (3) biocompatibility risks were minimal, (4) needle deflections were within limits, and (5) library-based applicator reconstructioning was feasible. For all patients custom configurations resulted in acceptable plans that had similar or improved dose conformity in comparison with the clinically used configuration. Conclusions The development cycle of our custom applicator was systematic and well-documented. Researchers are encouraged to share best practices to provide guidance for the in-house development of novel MDs. Dynamic Predictive Models for Side Effects Following Cancer or Cancer Treatment: A Systematic Review Fatime Oumar Djibrillah, Rebecca Schipper, Agnes Berendsen, Gabriela F. Nane, Maurice van Keulen, Annemieke Witteveen, Arlene John Abstract: Background: Advances in cancer treatments such as surgery, radiotherapy and chemotherapy have increased patient survival rates. However, these treatments often result in complications or late side-effects in patients. Accurate predictions of these side-effects are important to optimize post-treatment care focused on targeted interventions for prevention or management. Dynamic predictive models, which are specifically tailored to adapt to longitudinal patient data, offer an innovative approach to updating patient risk in light of new data. This systematic review aims to summarize the application of dynamic predictive models in predicting cancer treatment-related complications and to synthesize techniques and algorithms used in developing and validating these models. Methods: This review was conducted following the PRISMA guidelines. A systematic search was performed across multiple databases including Scopus, Web of Science, PubMed, and IEEE Xplore to identify studies that have employed dynamic predictive models for cancer or cancer treatment related complications or side-effects. The keywords used included “dynamic prediction”, “predictive models”, “treatment side effects”, “cancer treatment” and “over time”. Studies were included if they have employed longitudinal or time-varying data to update predictions over time, reflecting the incorporation of new data at multiple time points. Preliminary Results: A total of 506 studies were initially screened, resulting in the inclusion of 13 articles. Modelling techniques varied, including statistical models such as Cox proportional hazards and machine learning-based models like Long Short-Term Memory (LSTM) employed for handling time series data. The included studies were found to cover various cancer types, with prostate and head and neck cancers being the most common. Treatment types included surgery, radiation therapy, chemotherapy and hormone therapy. Predicted complications ranged from biochemical recurrence to patient-reported outcomes such as voice impairment. Each model employed different strategies for dynamically incorporating follow-up data. Most studies are from 2020–2024, reflecting a recent focus on dynamic models. Conclusion: Despite their versatility, dynamic models are not often used in oncology applications. This review highlights the diverse applications of dynamic models in predicting cancer or cancer treatment-related complications or side effects over time. These models showcase significant potential for improving post-treatment care by updating predictions as more data becomes available. MRI-adaptive regional hyperthermia in the pelvic region – benefit and hurdles Kemal Sumser, Iva Vilas-Boas Ribeiro, Ellen van Wesel, Gerard van Rhoon, Sergio Curto, Maurice Heemels, Maarten Paulides Abstract: Introduction: The development of magnetic resonance (MR) compatible hyperthermia devices enabled real-time dosimetry using MR temperature imaging (MRTI). Unfortunately, MRTI is affected by scanner drift, low signal to noise, and the impact of patient outer and internal motion (cardiac, respiratory, air motion). We recently showed that treatment modelling can address these gaps when smartly combining model and measurement. Herein we studied if this approach can be improved using MR imaging (MRI) of the patient in the treatment position. Methods: Fourteen healthy female volunteers underwent MRI scanning in two setups: without the applicator (“planning setup”) and inside the MR-compatible hyperthermia device (BSD-2000-3D MR system by Pyrexar Medical), integrated with a 1.5T GE Optima 450W scanner (referred to as “treatment setup”). Using both power absorption (SAR) and temperature-based optimization approaches, we created two hyperthermia treatment plans (HTP) from these setups. Temperature prediction acquired using the planning setup’s steering settings were compared to the "MR-adapted plan," which represents the optimized plan based on the actual patient anatomy and position inside the device (“treatment setup”). Applying the steering settings from the standard plan to the treatment setup led to an "apparent hyperthermia dose”. The temperature achieved in at least 50% of the target volume (T50) was used to assess the predicted difference between the planning and treatment setup. Results: Patient anatomy proved crucial; the absolute error between standard and apparent hyperthermia dose was 0.4±0.3°C when including both optimization approaches. For temperature-based optimization, the absolute errors were substantially higher than for SAR-based optimization. A clear improvement in tumour temperatures from 0.4°C to 0.7°C in ΔT50 was found for MRI-based replanning when temperature-based optimization approaches were used, which was not the case for SAR-based optimization (<0.2°C). Conclusions: Our study suggests that patient anatomy and position during treatment should be considered when aiming for accurate thermal simulations; we found improvement between 0.4°C and 0.7°C in T50. The benefit of MR-adapted HTP was most relevant for temperature-based optimization.

end %-->