Joost Nederend

Over half of patients with locally recurrent rectal cancer (LRRC) are ineligible for curative-intent multimodality treatment, typically including neoadjuvant treatment followed by surgery, due to unresectable disease, distant metastases, and/or frailty. Non-operative management for unresectable LRRC may include (chemo)(re-)irradiation, but evidence remains scarce. This study explores the current clinical practice of radiotherapy for LRRC in the Netherlands within this setting. A web-based survey was designed and distributed to members of the Dutch Society of Radiation Oncology between December 2024 and February 2025. The survey comprised 28 items, addressing clinical indications, target delineation, treatment planning, dose and fractionation, radiotherapy techniques and knowledge gaps in current clinical practice. 23 radiation oncologists from 18 out of 21 Dutch institutes completed the survey. While agreement existed regarding indications for radiotherapy in the non-operative management for LRRC, marked variability was observed in dose and fractionation schedules and target volume delineation. There seemed to be no consensus on response evaluation or follow-up schedules. Reported clinical challenges included patient heterogeneity, absence of evidence-based guidelines, limited knowledge on radiation-induced toxicity, difficulties estimating prognosis and management of therapy-resistant symptoms. This national survey highlights the clinical heterogeneity and lack of consensus regarding radiotherapy for unresectable LRRC. Variations in dose and fractionation schedules, target volume delineation and follow-up underscore the need for prospective studies. Future research initiatives should prioritize optimal dose and fractionation schedules and the role of dose escalation, while integrating quality assurance programs and patient-reported outcomes to guide development of evidence-based recommendations for this complex patient group.

Objectives: Accurate assessment of peritoneal tumour burden is crucial for treatment selection and prognosis. The Peritoneal Cancer Index (PCI), traditionally assessed by laparoscopy, is increasingly applied to imaging. This study investigates whether pre-segmented radiological PCI (rPCI) regions on CT improve interobserver variability, diagnostic accuracy, and user experience. Materials and methods: This prospective, comparative reader study included nine contrast-enhanced CT scans from patients with peritoneal metastases of ovarian, gastric, or colorectal origin. Radiologists, surgeons, and gynaecologists independently scored rPCI on two scans: one with and one without pre-segmented rPCI regions shown as colour overlays. Interobserver variability, mean absolute error compared with surgical PCI (sPCI), and reader confidence were compared between both conditions. User experience was evaluated through structured feedback. Results: In total, 82 clinicians from 19 countries participated in this study. Each CT was assessed 8–10 times with and without overlaid segmentations. Pre-segmented region overlays did not significantly improve interobserver variability (p = 0.123), PCI accuracy (mean absolute error, p = 0.374), or reader confidence (p = 0.593). However, with use of segmentations, rPCI values were closer to sPCI in 7 of 9 scans. In regions 9–12, the proportion of participants reporting moderate or high confidence increased from 48–53% without overlays to 59–63% with overlays. User feedback was positive, with 83% of participants reporting the tool as useful, 82% as easy to use, and 71% indicating they would use it in clinical practice. Conclusion: Pre-segmented region overlays for CT-based rPCI assessment were positively received by clinicians. Although no statistically significant improvements in interobserver variability or accuracy were observed, the findings of this exploratory study support further evaluation in larger studies to determine their clinical value.

A computer-aided diagnosis model for preoperative classification of ovarian tumors, developed using a multicenter CT dataset, demonstrated comparable performance to US-based models and radiologists on the internal and external test sets, illustrating challenges due to the heterogeneous nature of the disease and multicenter data.

Purpose: Accurate interpretation of CT scans after pancreatic resection is crucial for detecting abnormalities, including postoperative complications and cancer recurrence. This study investigates the feasibility and clinical utility of a novel MKNet-family deep learning architecture for auto-segmentation of the residual pancreas on postoperative CT imaging, in comparison to previous approaches. Method: Novel MKNet, MSKNet and MAKNet architectures were developed. Two datasets were used: the National Institutes of Health (NIH) dataset, comprising 82 annotated normal preoperative CT scans, and the IMPACT Consortium dataset (NCT06055010; https://github.com/IMPACTconsortium/IMPACT ), comprising 81 annotated postoperative CT scans obtained < 4 weeks after pancreatectomy. Performance was assessed by Hausdorff Distance (HD), 95th-percentile-HD (HD95) and Normalized Surface Distance (NSD), and secondarily by Dice Similarity Coefficient (DSC), and compared with self-implemented existing models for preoperative pancreas auto-segmentation. Qualitative evaluation was conducted by ten abdominal radiologists. Results: In the postoperative setting, the MAKNet architecture showed the best performance, with an HD and HD95 of 17.3 ± 11.2 mm and 11.5 ± 10.2 mm, respectively. DSC (64.9 ± 14.8%) and NSD (27.2 ± 8.2%) were comparable to the Attention-U-Net (DSC 66.0 ± 13.8%; NSD 27.8 ± 8.4%). Clinical evaluation indicated that the MKNet-family accurately defined the postoperative pancreas (i.e., requiring minimal or no modifications) in 64 of 81 segmentations (79%). Conclusion: This study demonstrates the effectiveness of novel MKNet-family architectures to accurately segment the residual pancreas on postoperative CT imaging over previous approaches. This advances the state-of-the-art in pancreas auto-segmentation and may be beneficial for medical application and education, acceleration of data annotation, and future research.